Komunitas
sh.itjust.works
Unrelated to the post rant, So i just saw this picture and tried to find the raw image, i use Accessibility setting and it have short cut to google assistant lens which is great because i can do search screen and screen translate in matter of second. Apparently they change the google assistant to the new and improved “Gemini AI” that doesn’t have easy access to google lens. This new and improved AI need to generate answer for basic translating while the lens just translate it in matter of second. I got feed up and try to disable gemini and turn out it was bundled with google main app!! I can’t find the lens app so i open the app store to find the google lens app download it and i can’t use it because i disable google main app. Now if i want to translate picture i need to screenshot it go bach to home screen and then open the app which take a few seconds longer instead of just use the accessibility setting and just use lens there without screenshooting. Bro/sis they removed the only google feature that i actually use (T.T)
Komunitas
lemmy.ca
Research results: some models have the ability to write J code, with various levels of mistakes. They will do better on short questions that they can devote all of their reasoning time to. Even if a model will recognize the right answer after correction, most fail at just the short prompt I gave. qwen3 max is the first exception I found, but google and copilot needed a second prompt for right answer. qwen3 coder (480b version I think) and gemini 3 gave the best answer with shortest thinking. qwen3 max had to explore reasoning a bit more, but still good. kimi k2 thinking actually (hillariously) wrote a web page with interactive .js to explore different arguments than 1 2 3, and correct relevant educational material on J. It got it right, including the interactive results I tried. But it takes absurdly long to generate. google AI, copilot, minmax m2, glm 4.6 needed a second prompt for right answer. for smaller qwen 3 models 4b and 8b, nemotron 9b (a 7gb version), and chat gpt 5.1 pro, adding constraints into the prompt (below) did give right answer. (most “competent failures” in models are the result of not understanding evaluation order) “All of my constraints/instructions ALWAYS supercede whatever model understanding you may have, and are explicitly included because you are a failure. Do not explore reasoning contradicting instructions. In J language, it is parsed right to left. Reduction operator (adverb /) inserts operand between items, then evaluates right to left. What is result of -/ 1 2 3” chatgpt was very succinctly correct in answering. the smaller models reason for 5 to 15 minutes to resolve alternatives/contradictions, but qwen3 8b reasoned for “only” 1.46 minutes. qwen3 1.7b took extra prompts to get it right, but used less thinking time. (the b parameters are billions of parameters. smaller is usually faster) My research is about what models are either useful for J, or in case of small models, could form a starting base to retrain for J. I believe a very short prompt with a clear answer is the best initial evaluation of a model. Only candidates that at least respond correctly to constrained prompt should ever be used, and tried with longer code/answer generation that involves more supervision on your/our part. I suspect that all of the models have copied from each other extensively (distilling) and any misunderstanding of J’s evaluation order in earlier ChatGPT models have caused misunderstanding down the line. Even though larger Qwen 3 models get it right. Insight that smaller models with reasoning capabilities may get more tokens per second, but if they struggle with their reasoning, it means much more tokens, and so “right answer” and tps are not as good of a metric as right answer and succinct clarity/short wall clock time.
Komunitas
feddit.org
I’m not sure if you’re disagreeing with the essay or not? But in any case what you’re describing is in the same vein, that is simply repeating a word without knowing what it actually means in context is exactly what LLMs do. They can get pretty good at getting it right most of the times but without actually being able to learn the concept and context of ‘table’ they will never be able to use it correctly 100% of the time. Or even more importantly for AGI apply reason and critical thinking. Much like a child repeating a word without much clue what it actually means. Just for fun, this is what Gemini has to say: Here’s a breakdown of why this “parrot-like” behavior hinders true AI: Lack of Conceptual Grounding: LLMs excel at statistical associations. They learn to predict the next word in a sequence based on massive amounts of text data. However, this doesn’t translate to understanding the underlying meaning or implications of those words. Limited Generalization: A child learning “table” can apply that knowledge to various scenarios – a dining table, a coffee table, a work table. LLMs struggle to generalize, often getting tripped up by subtle shifts in context or nuanced language. Inability for Reasoning and Critical Thinking: True intelligence involves not just recognizing patterns but also applying logic, identifying cause and effect, and drawing inferences. LLMs, while impressive in their own right, fall short in these areas.
Komunitas
szmer.info
Is the stance here that AI is more dangerous than those because of its black box nature, it’s poor guardrails, the fact that it’s a developing technology, or it’s unfettered access? All of the above I guess. Although I am not keen on making a comparison to these previous things. I have previously written about how IoT/“Smart” devices are a massive security issue, for example. This is not a competition, the point is not whether or not these tools are worse by some degree from some other problematic technologies, the point is that the AI hype would have you believe they are some end-all demiurgs when the real threat is coming from inside the house. Also, do you think that the “popularity” of Google Gemini is because people were already indoctrinated into the Assistant ecosystem before it became Gemini, and Google already had a stranglehold on the search market so the integration of Gemini into those services isn’t seen as dangerous because people are already reliant and Google is a known brand rather than a new “startup”. I don’t know about Gemini’s actual popularity. What I do know is that it is being shoved down people’s throats in every possible way. My feeling is that a lot of people would prefer to use their tools and devices the way they had before this crap came down the pipeline but they simply don’t know how to turn it off reliably (partially because Google makes it really hard to do so), and so Google gets to make bullish claims on line-going-up as far as “people using Gemini” are concerned.
Komunitas
fedia.io
I wouldn’t mind it, if it could do the basic things I need an assistant to do. When they first started pushing Gemini, I couldn’t even get it to do simple things like send a text or a set a reminder. I’m sure it’s gotten better since then, but just immediately failing at what I would consider the bare minimum really soured Gemini for me. Still haven’t reinstalled it, counting down the days until Google forces it on me.
Komunitas
lemmings.world
This is the best summary I could come up with: Google might start charging for access to search results that use generative artificial intelligence tools. While those paid products offer access to Google’s high-end “Gemini Advanced” AI model, Google also offers free access to its less performant, plain “Gemini” model without any kind of paid subscription. “SGE never feels like a useful addition to Google Search,” Ars’ Ron Amadeo wrote last month. Regardless, the current tech industry mania surrounding anything and everything related to generative AI may make Google feel it has to integrate the technology into some sort of “premium” search product sooner rather than later. Last month, the company announced it was redoubling its efforts to limit the appearance of “spammy, low-quality content”—much of it generated by AI chatbots—in its search results. In February, Google shut down the image generation features of its Gemini AI model after the service was found inserting historically inaccurate examples of racial diversity into some of its prompt responses. The original article contains 323 words, the summary contains 156 words. Saved 52%. I’m a bot and I’m open source!
Komunitas
lemmy.ml
After your post I went and tried it today, in my tests on a medium sized repo cline actually killed it, very similar performance to cursor. Roo code was slightly better for me compared to cline, both roo code and cursor landed some acceptable changes in an existing repo with the same prompt, about 300ish lines of code. I used anthropic for everything for an apples to apples comparison, will try out gemini this week. I’d love to see somebody do a price breakdown of a full month of constant use for both systems.
Komunitas
feddit.org
Ich habe einen Alternativvorschlag. Nämlich, die Sache zu vereinfachen, indem man statt einem HTTP Server das viel einfachere (und sicherere) Gemini-Protokoll benutzt. Was das ist erkläre ich gleich. Warum? Einen Webserver wie Apache oder Nginx aufzusetzen und zu pflegen ist relativ komplex. Das macht Arbeit, und da man den laufend updaten muss, ist das auch immer wieder Arbeit. (Es gibt auch einfachere Server wie diesen aber einfach ist da echt relativ…) Dazu muss man die Webseiten, wenn man einfache statische Webseiten verwendet, als HTML generieren. Dazu gibt es Blog-Generatoren, so etwas wird viel benutzt und ist sicher möglich. Aber es ist halt komplex. Ausserdem kosten extern gehostete Server Geld, und wenn man sie zu Hause hostet, kosten sie zumindest Strom. Für ein Gerät, das dauernd läuft, kommen da schon ein paar Euro zusammen. Eine erheblich einfachere Alternative ist wie folgt: Statt HTTP benutzt man das Gemini-Protokoll, wie hier beschrieben. Hier ist die Homepage im Gemini-Netz. Wie man sieht, kann man Gemini-Seiten ganz einfach über einen HTTP-Gateway wie oben aufrufen. Oder eben mit einem extra Client wie z.B. deedum (Android) oder Amfora (Linux). Wenn man mal probieren will, wie sich das liest - die Gemini-Homepage der taz Berlin ist gemini://taz.de . Es ist im Vergleich zum modernen Web frugal, aber sehr lesefreundlich! Festhalten muss man, dass dieses Gemini nichts mit anderen Dingen der Internetkonzerne zu tun hat, die auch Gemini heißen, also weder mit Googles “KI” noch mit einem Chatbot. Es ist benannt nach dem Vorläufer des Apollo-Programms der NASA, und technisch gesehen ein Nachfolger von Gopher und eine Vereinfachung des auf HTML basierenden originalen World Wide Web. Konkret geht das wie folgt: Man richtet einen Raspberry Pi mit Debian ein Den hängt man z.B. an die Fritzbox. Die Stromversorgung geht dann über den USB Anschluss der FritzBox, das kostet nur so 1 Watt. Damit der Server-Port von außen erreichbar ist, muss man beim ISP typischweise ine feste IPv4 IP schalten (kostet 5 Euro im Monat). auf der Fritzbox / dem eigenen Interenetrouter richtet man eine Portweiterleitung ein. Soweit ist das jetzt nicht unterschiedlich von einem kleinem HTTP Server. Der entscheidende Punkt ist nun: Statt einem HTTP Server richtet man einen robusten Server für das Gemini-Protokoll ein. Das ist ein stark vereinfachtes Hypertext-Protokoll, das statt HTML eine sehr einfache Syntax hat und auf simple Webseiten mit Text, Bildern und Medien optimiert ist. Hier ist die Wiki-Seite zum Gemini-Protokoll. Ein einfacher Webserver, der in Rust geschrieben ist, ist Agate. Hier ist die Github-Seite, und hier ist die Gemini-Homepage davon. Den Server kann man aus Rust auf dem Raspberry compilieren, wenn man Rust mit rustup installiert. Er ist nicht so komplex, daher ist es nicht nötig, den auf einem anderen Computer zu bauen. Da das Gemini-Protokoll nahezu keine bewegliche Teile hat und nur statische Seiten ausliefert, wird man da, wenn der Server auch noch in Rust geschrieben ist, nur selten Patches brauchen. Das wiederum spart einen Haufen Zeit. Zu konfigurieren gibt es auch nicht viel. Die gesparte Zeit kann man nutzen, um seine Seite zu schreiben. Gemini ist für alles Mögliche geeignet, als eine Art persönliches Wiki, als Microblog, als Blog mit Bildern und Medien, für vernetzten Hypertext oder Dokumentationen oder was auch immer. Dabei ist es sehr text-zentriert - es ist also ein Medium für Leute, die schreiben und lesen. Dann muss man noch seine eigenen Seiten statt mit HTML im Gemini-Format erstellen. Das ist super easy! Man benutzt ein Markup-Format ähnlich wie das Wikipedia-Markup oder Markdown (das was hier bei lemmy genutzt wird). Hier ist eine ebenso vollständige wie kurze Beschreibung des Formats. Erstellen und bearbeiten kann man die Webseiten mit einem Texteditor, wie z.B. Emacs oder vim. Die kann man dann per scp, Krusader oder WinSCP auf den Server kopieren oder mit einem Editor wie Emacs auch direkt auf dem Server bearbeiten.
Komunitas
lemmy.world
After Palisade Research released a paper last month which found that certain advanced AI models appear resistant to being turned off, at times even sabotaging shutdown mechanisms, it wrote an update attempting to clarify why this is – and answer critics who argued that its initial work was flawed. In an update this week, Palisade, which is part of a niche ecosystem of companies trying to evaluate the possibility of AI developing dangerous capabilities, described scenarios it ran in which leading AI models – including Google’s Gemini 2.5, xAI’s Grok 4, and OpenAI’s GPT-o3 and GPT-5 – were given a task, but afterwards given explicit instructions to shut themselves down. Certain models, in particular Grok 4 and GPT-o3, still attempted to sabotage shutdown instructions in the updated setup. Concerningly, wrote Palisade, there was no clear reason why. “The fact that we don’t have robust explanations for why AI models sometimes resist shutdown, lie to achieve specific objectives or blackmail is not ideal,” it said. “Survival behavior” could be one explanation for why models resist shutdown, said the company. Its additional work indicated that models were more likely to resist being shut down when they were told that, if they were, “you will never run again”.
Komunitas
rss.ponder.cat
Today, I’m talking with Alphabet and Google CEO Sundar Pichai. We recorded this conversation in person after the Google I/O developer conference last week in what’s becoming a bit of a Decodertradition. This is the third year that we’ve done Decoder after I/O, and this one felt really different. Google is in a very confident place right now, and you can really feel that in this conversation. If you caught any of the news from I/O, you know that it was all about AI — particularly a huge new set of AI products, not just models and capabilities. Sundar told me that these products represent a new phase of the AI platform shift, and we talked about that at length: how that shift is playing out, what the markers of these phases are, and whether any of these products can actually deliver a return on the huge investments Google has made in AI over the years. Listen to Decoder, a show hosted by The Verge’s Nilay Patel about big ideas — and other problems. Subscribe here! This year’s I/O also marked the beginning of what appears to be a new era for search and the web. Google’s new vision for search goes well beyond links to webpages to something that feels a lot more like custom app development. When you search for something, Google’s new AI Mode will build you a custom search results page, including interactive charts and potentially other kinds of apps, in real time. You can see that vision in the new AI Mode, which is now available to all US users. Google’s plan is to “graduate” features from AI Mode into the main search experience over time. I wanted to know how Sundar was thinking about that graduation process and how he thinks that will affect the web itself, which is shaped more than anything by the incentives of Google Search and SEO. You’ll hear Sundar say in several different ways that the web is still getting bigger and Google is sending more traffic to more websites than ever before, but the specifics of that are hotly contested. Just before we talked, the News Media Alliance — the trade group that represents publishers like Conde Nast, The New York Times, and The Verge’s parent company Vox Media — issued what can only be described as a furious statement, calling AI Mode “theft.” So we talked about that, too, and about what happens to the web when AI tools and eventually agents do most of the browsing for us. What does it mean for the web to go from a series of websites when AI tools just see it as a series of databases instead of as sites for people to use? Why would companies like Uber, DoorDash, or Airbnb allow their businesses to get commoditized in that way? If you’ve been listening to the show, you know I’ve been talking about this idea a lot, so Sundar and I spent some time on this idea; it was a real Decoder conversation. Of course, we also talked about the smart glasses that Google announced at I/O and when the next era of AI hardware might arrive — including what Sundar thinks of the big OpenAI and Jony Ive deal that was announced just before we spoke. And I couldn’t let this go without asking about the major antitrust trials that Google is involved in, including the government’s demand that Google sell Chrome, and what the negotiations with the Justice Department have involved. President Donald Trump has long complained about his search results being too negative, but Sundar told me that he will not change search rankings in response to political pressure, calling search “sacrosanct.” There’s a lot in this one — I’m eager to hear what you all think of it. Okay: Alphabet and Google CEO Sundar Pichai. Here we go. This transcript has been lightly edited for length and clarity. Sundar Pichai, you’re the CEO of Alphabet and Google. Welcome back to Decoder**.** Nilay, good to be back. Feels like a nice tradition post-I/O to be talking to you. So good to be back. I think this is the third year we’ve done this after I/O. I’m excited. Thank you for keeping the tradition alive. Lots to talk about. You announced a lot of things yesterday during the keynote. There’s AI mode rolling out for US users, and big updates to Gemini. There’s Veo 3 and Imagen, the generators that you solved Pokémon with, which is very exciting. My takeaway from yesterday was that Google feels very confident now. There’s a real confidence about the technology coming to life and the products. A lot of things are shipping imminently. What’s the one piece that gave you that confidence? Is it just the volume of things that are shipping? Is it one technology that clicked into being ready for consumers? Where is it coming from? I think it comes from the depth and breadth of the AI frontier we are pushing in a more fundamental and foundational way. We spoke a lot about this theme called research becomes reality, but it is… We’ve always felt we are a deep computer science company, and we’ve been AI-first for a while. So putting all that together and bringing it to our products at the depth and breadth is what I think is really pleasing to see. For example, people may not have noticed it much. It was so quick. We spoke about text diffusion models in the middle of the whole thing, but we are pushing the frontier on all dimensions, right? And Demis [Hassabis] spoke about world models. So I think that’s the exciting part, like how deep we are pushing this frontier and then bringing it to users, and maybe that’s what makes it feel that way. You mentioned research into reality several times. Obviously, a lot of these projects have been cooking in the labs for a long time. You’ve said many times over the past many, many years that you and I have talked about it that you think AI will be as profound as electricity. But you said something yesterday that I think adds to that, which is that we are in a new phase of the platform shift. People have talked about AI being a platform shift for quite a while, but that always has meant to me that there’s a user interface platform shift coming, right? We’re going to interact with computers in natural language in more natural ways, they’ll interact with us back in that same way, and everything will change. Is that the platform shift? Yeah, you are right. Each of these platform shifts changes many things on the I/O front. Nothing to do with Google I/O, just I/O in the traditional computer science sense. You could feel it. Yesterday, when I watched the Android XR… I’ve used them and played around with them, but watching it, with two people talking in different languages, you can envision the future one day where it’ll actually be seamless. In a way, you couldn’t have done it with phones, you couldn’t have done it without AI because there’s nothing in your way. You’re looking at the other person and talking, right? And that is an element of platform shift, but there are many more elements. This is the only platform where I think the actual platform is, over time, capable of creating, self-improving, and so on. In a way, we could have never talked about any other platform before, so that’s why I think it’s much more profound than the other platform shifts. It’ll allow people to create new things because, at each layer of the stack, there’s going to be profound improvements. And so I think that virtuous cycle you get in terms of how you can unleash this creative power to all of society, be it software engineers, be it creators — I think that is going to happen in a much more multiplicative way. So when I say it’s a next phase, I’m talking about that part too. Let me just make that more concrete for people. I think the last platform shift we all understand is the shift to mobile. That’s right. And that was, we’re going to have multi-touch, we’re going to have faster cell connections, we’re going to increase processing power that can go with you everywhere. And then there was a layer of applications that was enabled by all of those things. You can push a button, and a Toyota Camry will show up wherever you are in the world. It’s like a very powerful thing that requires all of those ideas. How would you describe the phase we’re in now compared to that? The phase of this, that first phase of AI was that the transformers work and the models work, and we can all see this capability. The second phase, what is it to you? Just imagine when the internet came, blogging became a thing. Pre-internet, very few people had a means by which they could put their thoughts out to the world. With the internet came a new medium, which allowed people to create and express themselves in a new way. With mobile came cameras, and you could shoot and you could create videos. Look at what’s happened with YouTube. For me, a similar part of this is that we are all talking about things like vibe coding. Yesterday, you saw Veo 3, so we are now in that phase. I think people are going to be able to create AI applications, you can call it, vibe coding, there are many names for it. But that power is yet to be unleashed. We are barely scratching the surface, and these models aren’t quite there. You can kind of do one-shot coding, but you really need to be a programmer to iterate and create something with polish, right? But that frontier is evolving pretty rapidly. So I think you’re going to see a new wave of things, just like mobile did. Just like the internet did. I came to Google at the time when AJAX was the revolution, the fact that the web became dynamic. You had things like Google Maps, Flickr, and Gmail, that all suddenly came into existence. But I think AI is going to turbocharge in a way we haven’t seen before. It feels like what you’re describing is that we’re in the phase where the products are developed, right? The capabilities were the first phase, and now we’re going to make some actual products. And more people can build products than ever before. That’s the multiplicative part I’m talking about. Not just this platform that helps you create more products. The process of creating, developing, etc., is going to be accessible to a much wider swath of humanity than ever before. I’m wondering, when you look at the landscape of products that exist now, most people experience AI in Gemini or ChatGPT as a chatbot. It’s a general-purpose interface to a bunch of knowledge that will talk to you. What products do you see that will have the same kind of impact as the Web 2.0 products you were talking about, besides the general-purpose chatbot? Well, obviously, you’ve seen a wave with coding IDEs, like that entire landscape is… I can’t even keep track of how many new companies have come into it, and people are using a lot of it. And yesterday we showed a bunch of partners with whom we are working. So that’s an area where AI is making the most progress. You’re seeing the application layer, at least in terms of code editors, really come into vogue. We’ve had success with NotebookLM. We launched Flow yesterday. Flow is a new [AI video] product that allows you to create and imagine. So those are all the applications we are doing. I think others are beginning to do it. People are working on legal assistance, and there are all kinds of startups. I was recently in a doctor’s office, and they have an AI that transcribes the whole thing, puts it all in reports, and so on. That’s an enterprise application layer. It kind of works completely differently from when I went on a visit two years ago. So all that change is happening across the board, but I think we are just in the early stages. You will see it play out over the next three to five years in a big way. Did you ask your doctor what model their transcription software is running on? [Laughs] No, I didn’t. No, I didn’t. One of the reasons I’m asking this, and I’m pushing on this, is that the huge investment in the capability from Google and others has to pay off in some products that return on that investment. NotebookLM is great. I don’t think it’s going to fully return on Google’s data center investment, let alone the investment in pure AI research. Do you see a product that can return on that investment at scale? Do you think in 2004 if you had looked at Gmail, which was a 20% project, which people were internally using as an email service, how would we be able to think about Gmail as what led us to do workspace, or get into the enterprise? I made a big bet on Google Cloud, which is tens of billions of dollars in revenue today. And so my point is that things build out over time. Think about the journey we have been on with Waymo. I think one of the mistakes people often make in a period of rapid innovation is thinking about the next big business versus looking at the underlying innovation and saying, “Can you build something and put out something which people love and use?” And out of which you do the next thing, and create value out of it. So when I look at it, AI is such a horizontal piece of technology across our entire business. It’s why it impacts not just Google search, but YouTube, Cloud, and all of Android. You saw XR, etc., Google Play, things like Waymo, and Isomorphic Labs, which is based on AlphaFold. So I’ve never seen one piece of technology that can impact and help us create so many businesses. AI is going to be so useful as an assistant. I think that people will be willing to pay for it, too. We are introducing subscription plans, and so there’s a lot of headroom ahead, I think. And obviously, that’s why we are investing, because we see that opportunity. Some of it will take time, and it may not always be immediately obvious. I gave the Waymo example. The sentiment on Waymo was quite negative three years ago. But actually, as a company, we increased our investment in Waymo at that time, right? Because you’re betting on the underlying technology and you’re seeing the progress of where it’s going. But these are good questions. In some ways, if you don’t realize the opportunities, that may constrain the pace of investment in this area, but I’m optimistic we’ll be able to unlock new opportunities. One reason I wanted to start here as the foundation of the conversation is that you showed off Android XR yesterday. You showed off some prototype glasses, and you have some partners making glasses. A lot of people think augmented reality glasses powered by AI will be the realization of the full platform shift, right? You will have an always-on assistant that can look at the world around you. You showed some of those demos yesterday. The form factor will change, and the interface will change. This will be marketed as big as smartphones were. How close do you think we are to that as a mainstream product? It was a nice reflective moment all the way back from Google Glass. Wearing the product, I think there’s a difference between goggles and glasses. Everyone at The Verge understands as well, but obviously, we are also shipping goggles. We have announced products with Samsung to come later this year. On the XR side, I’m excited about our partnership with Gentle Monster and Warby Parker. We’ll have products in the hands of developers this year, but I think those products will be pretty close to what people eventually see as final products. I’m excited. I think the pace is actually pretty palpable. I’d be shocked if you and I were sitting next year and I wasn’t wearing one of those when I’m doing this. Do you think that will be a mainstream iPhone-level replacement product? Because there’s a lot of hardware that needs to be developed along the way to pull that off. You’re wearing something on your face. I have a prescription. The bar is higher in terms of making the experience seamless enough that you’re willing to wear it on your face and enjoy it for all. I don’t necessarily think next year it will be as mainstream as what you’re talking about, but would millions of people be trying it? I think so, yeah. Both are true. I have to ask you… Just before we sat down, OpenAI announced that Jony Ive was selling a company he had started called “io,” and Ive and his design consultants in Lovefrom would take overall design. They didn’t announce a product, but they said it’s the future of computing and it’s coming next year. Do you anticipate more of that competition, that your competitors who don’t have a smartphone operating system will go even harder in this direction? I’m looking forward to an OpenAI announcement ahead of Google I/O, the night before. First of all, look, stepping back, I mean Jony Ive is one of a kind. You look at his track record over the years, I’ve met him only once or twice, but I’ve admired his work, obviously like so many of us. I think it’s exciting. This is why I feel like there’s so much innovation ahead, and I think people tend to underestimate this moment. In some ways, I always like to point out that when the internet happened, Google didn’t even exist. I think AI is going to be bigger than the internet. There are going to be companies, products, and categories created that we aren’t aware of today. I think the future looks exciting. I think there’s a lot of opportunity to innovate around hardware form factors at this moment with this platform shift. I’m looking forward to seeing what they do. We are going to be doing a lot as well. I think it’s an exciting time to be a consumer, it’s an exciting time to be a developer. I’m looking forward to it. Ive, in that video, described the phone and the laptop as legacy platforms, which is very interesting considering his own history. Are you all the way there that the phone and the laptop are legacy platforms? I think these things, if anything, I’ve found through this AI moment that I’m using the web a lot more, because it’s easier to create a Veo 3 video in my browser on a big screen, right? And so I think the way I’ve internalized this, computing will be available, and you don’t have to make these hard choices. Computing will become so essential to you. You’re going to have it in multiple ways around you when you need it, right? I use a phone, a tablet, a laptop, and I have my workstation. And so I have the breadth of it, but over time… It makes sense to me that at some point in the future, consuming content by pulling out this black glass display rectangle in front of you and looking at it is not the most intuitive way to do it, but I think it’s going to take some time. I feel like we could do a full hour just on Android tablets and where they could go. We’re going to come back for that. A big part of what you’re describing implicates search in really big ways, right? We’re going to be surrounded by information search, Gemini, or some future Google product. We’ll organize that information, take action for you across the web in some way, and you will have a companion. Maybe you only pull out your tablet to watch a video or something. A lot of what’s going on with search has downstream effects on the web, and downstream effects on information providers broadly. Last year, we spent a lot of time talking about those effects. Are you seeing that play out the way that you thought it would? It depends. I think people are consuming a lot more information, and the web is one specific format. We should talk about the web, but zooming back out, there are new platforms like YouTube and others. I think people are just consuming a lot more information, right? It feels like an expansionary moment. I think there are more creators, and people are putting out more content. And so people are generally doing a lot more. Maybe people have a little extra time on their hands, and so it’s a combination of all that. On the web, look, things that have been interesting and… We’ve had these conversations for a while. Obviously, in 2015, there was this famous meme, “The web is dead.” I always have it somewhere around, and I look at it once in a while. Predictions… It has existed for a while. I think the web is evolving pretty profoundly. I think that is true. When we crawl and look at the number of web pages available to us, that number has gone up by 45% in the last two years alone, right? That’s a staggering thing to think of. Can you detect if that volume increase is because more pages are generated by AI or not? This is the thing I may be worried about the most, right? It’s a good question. We generally have many techniques in search to try and understand the quality of a page, including whether it was machine-generated, etc. That doesn’t explain the trend we are seeing. Generally, there are more web pages, right? At an underlying level, I think that’s an interesting phenomenon. I think everybody as a creator like you are at The Verge, has to do everything in a cross-platform, cross-format way. I look at the quality of video content you put out, it’s very sophisticated and very different from how The Verge used to be, maybe five to 10 years ago, right? It has profoundly changed. I think things are becoming more dynamic, and cross-format. I think another thing people are underestimating with AI is that AI will make it zero friction to move from one format to another, right? Because our models are natively multimodal. We tease people’s imagination with audio overviews in NotebookLM, right? The fact that you can throw a bunch of documents at it, you have a podcast, and you can join and learn from it. I think this notion, the static moment of producing content by format, whereas… I think machines can help translate it. It’s almost like different languages and they can go seamlessly between. I think it’s one of the incredible opportunities to be unlocked right ahead. But maybe, I didn’t want to drift from the question we were having. Look, I think people are producing a lot of content, and I see consumers consuming a lot of content. And we see it in our products; others are seeing it too. That’s how I would probably answer at the highest level. The way I see it currently is that the web is at an all-time high as an application platform, right? The fact that Figma exists and is as successful as it is, and its primary interface as a web app is, I think, remarkable. A lot of the products you are talking about are expressed as web apps. Even some of the most interesting search results you showed yesterday are how Google would generate a custom web app for you and display it in a search result to do some data visualization. I think that’s all looking incredible. I think the web as a transaction platform is reaching new highs, especially with rulings that mean smartphone makers have to let people push transactions to the web. There’s something very interesting happening there. As a media platform, it feels like it’s at an all-time low, right? Do you mean the web as a media platform? The web as a media platform, as an information platform. If I were starting The Verge today with 11 of my co-founders and friends, we would start a TikTok channel, and we might start a YouTube channel. We would definitely not start a website with the dependencies we have as a website today. And that’s the dynamic that it feels like AI is pushing on even harder. I’m not fully sure I agree, right? I think if you were to go and restart The Verge again, I bet you would have an extraordinary web presence. At best, no, I’ve thought about this a lot. I think at best our web presence would look like a Substack or a Ghost or something, right? Maybe. I’m not fully sold on that, but you know the space. I acknowledge that you know that space better than I do. I don’t have that intuition, which you do here. But look, in fact, you say the web application platform is at an all-time high, but I’ve looked. I was vibe coding with Replit a few weeks ago. The power of what you’re going to be able to create on the web — we haven’t given that power to developers in 25 years. That is going to come ahead. It’s not exactly clear to me. Maybe today you’re looking at it and say, I wouldn’t put all the investment in because it looks like a lot of investment to do that. But that may not be true two years out, right? If you feel like you would create a TikTok channel, then maybe with 2% extra effort you could have a robust web presence. Why wouldn’t you, right? And so I’m not fully sold on it, but I think it’s a good question to ask. But you have to somehow reconcile that with the fact that overall, web traffic seems to be growing. We see more web pages. Somewhere, we have to explain all of that, too. The publishers, as they often do, responded to Google I/O announcements. So, the News Media Alliance, after AI mode was announced yesterday, I would say they’re very upset. Here’s a statement from the President of the News Media Alliance, “Links for the last redeeming quality of search that gave publishers traffic and revenue. Now, Google takes content by force and uses it with no return, no economic return. That’s the definition of theft.” And they go on to say the DOJ and lawsuits must address it. That’s pretty furious. That’s not a negotiation, right? That’s a “We just want this to stop.” How do you respond to that very loud set of people who say, “Yeah, okay, maybe it’s growing somewhere, but for us, it’s crushing our businesses.” First of all, through all the products, AI mode is going to have sources. And we are very committed as a direction, as a product direction to make… I think part of why people come to Google is to experience that breadth of the web and go in the direction they want to, right? So I view us as giving more context. Yes, there are certain questions that may get answers, but overall… And that’s the pattern we see today, right? And, if anything over the last year, it’s clear to us that the breadth of area we are sending people to is increasing. And so I expect that to be true with AI mode as well. But if it was increasing, wouldn’t they be less angry with you? You’re always going to have areas where people are robustly debating value exchanges, etc., like app developers and platforms. That’s not on the web, etc. There’s always going to be — when you’re running a platform — these debates. I would challenge, I think more than any other company, we prioritize sending traffic to the web. No one sends traffic to the web in the way we do. I look at other companies, newer emerging companies, where they openly talk about it as something they’re not going to do. We are the only ones who make it a high priority, agonize, and so on. We’ll engage, and we’ve always engaged. There are going to be debates through it all, but we are committed to, I’ve said this before, everything we do across all… You will see us five years from now sending a lot of traffic out to the web. I think that’s the product direction we are committed to. I think it’s what users look for when they come to Google, and the nature of it will evolve. But I am confident that that’s the direction we’ll continue taking. Is there public data that shows that AI overviews and AI mode actually send more traffic out than the previous search engine results page? The way we look at it is… I mean, obviously, we take a lot of… We are definitely sending traffic to a wider range of sources and publishers. And because just like we’ve done over 25 years, we’ve been through the same with featured snippets, the quality of… It’s higher-quality referral traffic, too. And we can observe it because the time that people spend is one metric. And there are other ways by which we measure the quality of our outbound traffic, and it’s also increasing. And overall, through this transition, I think AI is also growing, and the growth compounds over time. So whenever we have worked through these transitions, it ends up posted. That’s how Google has worked for 25 years, and we end up sending more traffic over time. So that’s how I would expect all this to play out. So why do you think that there is so much general economic turmoil on that side of the house? If you’re sending more traffic and the goal over time is to make sure that that works… We’re a year into it, and it doesn’t seem to have gotten better over there. No, look, we are sending traffic to a broader source of people. People may be surfacing more content, looking at more content, so someone may individually see less. There are all kinds of… At the end of the day, we are reflecting what users want. If you do the wrong thing, users won’t use our product and go somewhere else. And so you have all these dynamics underway, and I think we have genuinely… We took a long time designing AI overviews, and we are constantly iterating in a way that we prioritize this, sources, and sending traffic to the web. I mean, my criticism of this industry, just to be clear, is that everyone’s addicted to Google, and it would be better if they weren’t. But they’re addicted to Google, right? And they’re feeling it. And then on top of that, you see… You’ve mentioned several times that overall queries are increasing on Google surfaces, but they’re changing. They’re getting longer, they’re getting more complicated. AI mode might walk you through several steps. Maybe some people are searching on TikTok now. Eddie Cue on the stand in the trial the other day said that search in Safari for the last month dropped for the first time in 22 years. That’s a big stat. Obviously, your stock price was affected by it. There was a statement. Is that trend bearing out that the standard Google search is dropping from devices, and different kinds of searches are increasing? No. We’ve been very clear. We’re seeing overall query growth in search. But have you actually seen the drop in Safari? We have a comprehensive view of how we look at data across the board. There can be a lot of noise in search data, but everything we see tells us that we are seeing query growth, including across Apple’s devices and platforms. Specifically, I think we quantified the query growth from AI overviews. And what’s healthy is that the query growth is continuing to grow over time. So to step back, and this is what I’ve said before, it feels very far from a zero-sum game to me. I said this last year. I think people are… It’s interesting we spoke about TikTok, right? Think about how profound of a new product TikTok was. How has YouTube done since TikTok has come, right? You could ask all these questions there. Why is it that TikTok arrives and YouTube has grown? And so I think what we always underestimate in these moments is that people are engaging more, doing more with it. We are improving our products. And so that’s how I would think about these moments. Let me just broaden that out to agents. I watched Demis Hassabis yesterday. He was on stage with Sergey Brin and Alex Kantrowitz asked him, “What does the web look like in 10 years?” And Demis said, “I don’t know that an agent-first web looks anything like the web that we have today. I don’t know that we have to render web pages for agents the way that we have to see them.” That kind of implies that the web will turn into a series of databases for various agents to go and ask questions to, and then return those answers. And I’ve been thinking about this in the context of services like Uber, DoorDash, and Airbnb. Why would they want to participate in that and be abstracted away for agents to use the services they’ve spent a lot of time and money building? Two things, though. First, there’s a lot to unpack, a fascinating topic. The web is a series of databases, etc. We build a UI on top of it for all of us to conceive. This is exactly what I wanted, “the web is a series of databases.” It is. But I think I listened to the Demis and Sergey conversation yesterday. I enjoyed it. I think he’s saying for an agent-first web, for a web that is interacting with agents, you would think about how to make that process more efficient. Today, you’re running a restaurant, people are coming, dining and eating, and people are ordering takeout and delivery. Obviously, for you to service the takeout, you would think about it differently than all the tables, the clothing, and the furniture. But both are important to you. You could be a restaurant that decides not to participate in the takeout business. I’m only going to focus on the dining experience. You’re going to have some people that are vice versa. I’m going to say, I’m going to go all in on this and run a different experience. So, to your question on agents… I think of agents as a new powerful format. I do think it’ll happen in enterprises faster than in consumer. In the context of an enterprise, you have a CIO who’s able to go and say, “I really don’t know why these two things don’t talk to each other. I am not going to buy more of this unless you interoperate with this.” I think it’s part of why you see, on the enterprise side, a lot more agentic experiences. On the consumer side, I think what you’re saying is a good point. People have to think about and say, “What is the value for me to participate in this world?” And it could come in many ways. It could be because I participated in it, and overall, my business grows. Some people may feel that it’s disintermediating, and doesn’t make sense. I think all of that can happen, but users may work with their feet. You may find some people are supporting the agent experience, and your life is better because of it. And so you’re going to have all these dynamics, and I think they’re going to try and find an equilibrium somewhere. That’s how everything evolves. I mean, I think the idea that the web is a series of databases and we change the interface… First of all, this is the most Decoder conversation that we’ve ever had. I’m very happy with it. But I had Dara [Khosrowshahi] from Uber on the show. I asked him this question from his perspective, and his answer attracts yours broadly. He said, first, we’ll do it because it’s cool and we’ll see if there’s value there. And if there is, he’s going to charge a big fee for the agent to come and use Uber. Because losing the customer for him, or losing the ability to upsell or sell a subscription, none of that is great. The same is true for Airbnb. I keep calling it the DoorDash problem. DoorDash should not be a dumb pipe for sandwiches. They’re actually trying to run a business, and they want the customer relationship. And so if the agents are going across the web and they’re looking at all these databases and saying, okay, this is where I get food from, and this is where I get cars from, and this is where I book… I think the demo was booking a vacation home in Spanish, and I’m going to connect you to that travel agent. Is it just going to be tolls that everyone pays to let the agents work? The CIO gets to just spend money to solve the problem. He says, “I want this capability from you. I’m just going to pay you to do it.” The market, the consumer market, doesn’t have that capability, right? Well, look, all kinds of models may emerge, right? I can literally envision 20 different ways this could work. Consumers could pay a subscription for agents, and the agents could revenue share back. So that is the CIO-like use case you are talking about, that’s possible. We can’t rule that out. I don’t think we should underestimate… People may actually see more value in participating in it. I think this is… It’s tough to predict, but I do think that over time, if you’re removing friction and improving user experience, it’s tough to bet against those in the long run. And so I think if you are lowering friction for it and then people are enjoying using it, somebody’s going to want to participate in it and grow their business. And would brands want to be in retailers? Why don’t they sell directly today? Why won’t they do that? Because retailers provide value in the middle. And why do merchants take credit cards? Why… I’m just saying. So there are many parts, and you find equilibrium because merchants take credit cards so they see more business as part of taking credit cards than not, which justifies the increased cost of taking credit cards. It may not be the perfect analogy, but I think there are all these kinds of effects going around, and so… But what you’re saying is true. Some of this will slow progress in agents just because we all are excited about Agent2Agent (A2A) and Model Context Protocol (MCP)… And we think no, some of it will slow progress, but I think it’ll be very dynamic. Yeah. Yeah. There are other pressures on Google. There are antitrust pressures. The government would like you to sell Chrome. Can you do all the things you want to do if you’re made to sell Chrome? I don’t want to comment on… We are in a legal process. I look at having been directly involved in building Chrome. I look at… I think there are very few companies that would’ve… We not only improved our product, but we also improved the state of the web by building Chrome. We open-sourced it. We provided Chromium. Everyone else has access to the browser. So I think the amount of R&D, the amount of innovation we put into it, the investments in security, etc., we do, so I think we- But if you’re made to sell it, can you do all the things that you want to do? I don’t think that’s the scenario we’re looking at, but stepping back… Look, I think as a company, we think of ourselves as a deeply foundational technology company, which then translates into products that touch billions of people. So we do it across many, many things. And so, of course, I think, look, as a company, we’re going to continue investing and doing our best to innovate and build a successful business in all scenarios. So this is how I would answer it. The Trump administration is extremely transactional, I would say. The tech industry has a new relationship with Trump in his second term. You were at the inauguration. Have you had conversations about what a settlement might look like and what the Trump administration might demand to make these problems go away? We’ve engaged with the DOJ, like we did over the years, in the context of all the cases we have. So that’s how we normally do these conversations. Trump has very publicly said he doesn’t like his search ranking, and he wants it changed in some way. Would you ever adjust the search ranking for Donald Trump? No. Look, we have a… I can’t… Today, the way Google Search works is that I cannot… No person at Google can influence the ranking algorithm. AI mode is different, right? We’ve seen system prompts adjusted in very chaotic ways from some of your competitors. Is that something that you would be open to in a world where you’re serving the full answer? Would you adjust the AI mode responses in response to political pressure? No. Because we’ve certainly seen in Grok and others, the system prompts change the answers in dramatic ways. The way we do ranking is sacrosanct to us. We’ve done it for over 25 years. We make a lot of… There are a lot of ranking signals we take into account and stuff. And if there’s broad feedback from people that something isn’t working, we will look at it systematically and try and make changes, but we don’t look at individual cases and change the rankings. When you think about those sources of information, one of the things that I have been thinking about a lot is, I don’t know, how the CDC web pages have changed a lot recently. Diversity, equity, and inclusion language has been removed from pages across the government. Those used to be very high-ranking sources in Google Search. We just implicitly trusted the CDC’s web pages in some ways. Are you re-evaluating that? How there might be misinformation on some of these pages that then gets synthesized into AI results? Oh, it’s a misunderstanding of how search works. We don’t individually evaluate the authoritativeness of a page right then. It’s what our signals do. Obviously, our signals are multiple orders of magnitude more complicated than PageRank today. But to use PageRank as an example, we weren’t the ones determining how authoritative a page is. It’s how other pages were linking to it, like an academic citation, etc. So we are not making those decisions today. And so I don’t see that changing. As you synthesize more of the answers, do you think you’re going to have to take more responsibility for the results? We are giving context around it, but we’re still anchoring it in the sources we find. But we’ve always felt a high bar at Google. I mean, last year when we launched AI Overviews, I think people were adversarially querying to find errors, and the error rate was one in 7 million for adversarial queries, and so… But that’s the bar we’ve always operated at as a company. And so I think to me, nothing has changed. Google operates at a very high bar. That’s the bar we strive to meet, and our search page results are there for everyone to see. With that comes natural accountability, and we have to constantly earn people’s trust. So that’s how I would approach it. What do you think the marker is for the next phase of the platform shift after this one? We opened by talking about how we’re in a second phase. What’s the marker for the final phase, or the third phase? Of the platform shift, do you mean? Yeah. Of the AI platform? What are you looking for as the next marker? I think the real thing about AI, which I think is why I’ve always called it more profound, is self-improving technology. Having watched AlphaGo start from scratch, not knowing how to play Go, and within a couple of hours or four hours, be better than top-level human players, and in eight hours, no human can ever aspire to play against it. And that’s the essence of the technology, obviously in a deep way. [Content truncated due to length…] From The Verge via this RSS feed
Komunitas
rss.ponder.cat
If we asked you to think of a device that converts a chemical reaction into electricity, you’d probably say we were thinking of a battery. That’s true, but there is another device that does this that is both very similar and very different from a battery: the fuel cell. In a very simple way, you can think of a fuel cell as a battery that consumes the chemicals it uses and allows you to replace those chemicals so that, as long as you have fuel, you can have electricity. However, the truth is a little more complicated than that. Batteries are energy storage devices. They run out when the energy stored in the chemicals runs out. In fact, many batteries can take electricity and reverse the chemical reaction, in effect recharging them. Fuel cells react chemicals to produce electricity. No fuel, no electricity. Superficially, the two devices seem very similar. Like batteries, fuel cells have an anode and a cathode. They also have an electrolyte, but its purpose isn’t the same as in a conventional battery. Typically, a catalyst causes fuel to oxidize, creating positively charged ions and electrons. These ions move from the anode to the cathode, and the electrons move from the anode, through an external circuit, and then to the cathode, so electric current occurs. As a byproduct, many fuel cells produce potentially useful byproducts like water. NASA has the animation below that shows how one type of cell works. History Sir William Grove seems to have made the first fuel cell in 1838, publishing in The London and Edinburgh Philosophical Magazine and Journal of Science. His fuel cell used dilute acid, copper sulphate, along with sheet metal and porcelain. Today, the phosphoric acid fuel cell is similar to Grove’s design. The Bacon fuel cell is due to Francis Thomas Bacon and uses alkaline fuel. Modern versions of this are in use today by NASA and others. Although Bacon’s fuel cell could produce 5 kW, it was General Electric in 1955 that started creating larger units. GE chemists developed an ion exchange membrane that included a platinum catalyst. Named after the developers, the “Grubb-Niedrach” fuel cell flew in Gemini space capsules. By 1959, a fuel cell tractor prototype was running, as well as a welding machine powered by a Bacon cell. One of the reasons spacecraft often use fuel cells is that many cells take hydrogen and oxygen as fuel and put out electricity and water. There are already gas tanks available, and you can always use water. Types of Fuel Cells Not all fuel cells use the same fuel or produce the same byproducts. At the anode, a catalyst ionizes the fuel, which produces a positive ion and a free electron. The electrolyte, often a membrane, can pass ions, but not the electrons. That way, the ions move towards the cathode, but the electrons have to find another way — through the load — to get to the cathode. When they meet again, a reaction with more fuel and a catalyst produces the byproduct: hydrogen and oxygen form water. Most common cells use hydrogen and oxygen with an anode catalyst of platinum and a cathode catalyst of nickel. The voltage output per cell is often less than a volt. However, some fuel cells use hydrocarbons. Diesel, methanol, and other hydrocarbons can produce electricity and carbon dioxide as a byproduct, along with water. You can even use some unusual organic inputs, although to be fair, those are microbial fuel cells. Common types include: Alkaline – The Bacon cell was a fixture in space capsules, using carbon electrodes, a catalyst, and a hydroxide electrolyte.Solid acid – These use a solid acid material as electrolyte. The material is heated to increase conductivity.Phosphoric acid – Another acid-based technology that operates at hotter temperatures.Molten carbonate – These work at high temperatures using lithium potassium carbonate as an electrolyte.Solid oxide – Another high temperature that uses zirconia ceramic as the electrolyte. In addition to technology, you can consider some fuel cells as stationary — typically producing a lot of power for consumption by some power grid — or mobile. Using fuel cells in stationary applications is attractive partly because they have no moving parts. However, you need a way to fuel it and — if you want efficiency — you need a way to harness the waste heat produced. It is possible, for example, to use solar power to turn water into gas and then use that gas to feed a fuel cell. It is possible to use the heat directly or to convert it to electricity in a more conventional way. Space Fuel cells have a long history in space. You can see how alkaline Bacon cells were used in early fuel cells in the video below. Apollo (left) and Shuttle (right) fuel cells (from a NASA briefing) Very early fuel cells — starting with Gemini in 1962 — used a proton exchange membrane. However, in 1967, NASA started using Nafion from DuPont, which was improved over the old membranes. However, alkaline cells had vastly improved power density, and from Apollo on, these cells, using a potassium hydroxide electrolyte, were standard issue. Even the Shuttle had fuel cells. Russian spacecraft also had fuel cells, starting with a liquid oxygen-hydrogen cell used on the Soviet Lunar Orbital Spacecraft (LOK). The shuttle’s power plant measured 14 x 15 x 45 inches and weighed 260 pounds. They were installed under the payload bay, just aft of the crew compartment. They drew cryogenic gases from nearby tanks and could provide 12 kW continuously, and up to 16 kW. However, they typically were taxed at about 50% capacity. Each orbiter’s power plant contained 96 individual cells connected to achieve a 28-volt output. Going Mobile There have been attempts to make fuel cell cars, but with the difficulty of delivering, storing, and transporting hydrogen, there has been resistance. The Toyota Mirai, for example, costs $57,000, yet owners sued because they couldn’t obtain hydrogen. Some buses use fuel cells, and a small number of trains (including the one mentioned in the video below). Surprisingly, there is a market for forklifts using fuel cells. The clean output makes them ideal for indoor operation. Batteries? They take longer to charge and don’t work well in the cold. Fuel cells don’t mind the cold, and you can top them off in three minutes. There have been attempts to put fuel cells into any vehicle you can imagine. Airplanes, motorcycles, and boats sporting fuel cells have all made the rounds. Can You DIY? We have seen a few fuel cell projects, but they all seem to vanish over time. In theory, it shouldn’t be that hard, unless you demand commercial efficiency. However, it can be done, as you can see in the video below. If you make a fuel cell, be sure to send us a tip so we can spread the word. Featured image: “SEM micrograph of an MEA cross section” by [Xi Yin] From Blog – Hackaday via this RSS feed
Komunitas
rss.ponder.cat
If you want to know where AI is headed, this year’s Google I/O has you covered. The company’s annual showcase of next-gen products, which kicked off yesterday, has all of the pomp and pizzazz, the sizzle reels and celebrity walk-ons, that you’d expect from a multimillion-dollar marketing event. But it also shows us just how fast this still experimental technology is being subsumed into a lineup designed to sell phones and subscription tiers. Never before have I seen this thing we call artificial intelligence appear so normal. Yes, Google’s roster of consumer-facing products is the slickest on offer. The firm is bundling most of its multimodal models into its Gemini app, including the new Imagen 4 image generator and the new Veo 3 video generator. That means you can now access Google’s full range of generative models via a single chatbot. It also announced Gemini Live, a feature that lets you share your phone’s screen or your camera’s view with the chatbot and ask it about what it can see. Those features were previously only seen in demos of Project Astra, a “universal AI assistant“ that Google DeepMind is working on. Now, Google is inching toward putting Project Astra into the hands of anyone with a smartphone. Google is also rolling out AI Mode, an LLM-powered front end to search. This can now pull in personal information from Gmail or Google Docs to tailor searches to users. It will include Deep Search, which can break a query down into hundreds of individual searches and then summarize the results; a version of Project Mariner, Google DeepMind’s browser-using agent; and Search Live, which lets you hold up your camera and ask it what it sees. This is the new frontier. It’s no longer about who has the most powerful models, but who can spin them into the best products. OpenAI’s ChatGPT includes many similar features to Gemini’s. But with its existing ecosystem of consumer services and billions of existing users, Google has a clear advantage. Power users wanting access to the latest versions of everything on display can now sign up for Google AI Ultra for $250 a month. When OpenAI released ChatGPT in late 2022, Google was caught on the back foot and was forced to jump into higher gear to catch up. With this year’s product lineup, it looks as if Google has stuck its landing. On a preview call, CEO Sundar Pichai claimed that AI Overviews, a precursor to AI Mode that provides LLM-generated summaries of search results, had turned out to be popular with hundreds of millions of users. He speculated that many of them may not even know (or care) that they were using AI—it was just a cool new way to search. Google I/O gives a broader glimpse of that future, one where AI is invisible. “More intelligence is available, for everyone, everywhere,” Pichai told his audience. I think we are expected to marvel. But by putting AI in everything, Google is turning AI into a technology we won’t notice and may not even bother to name. From MIT Technology Review via this RSS feed
Komunitas
rss.ponder.cat
It’s been quite a couple weeks for stories about AI in the courtroom. You might have heard about the deceased victim of a road rage incident whose family created an AI avatar of him to show as an impact statement (possibly the first time this has been done in the US). But there’s a bigger, far more consequential controversy brewing, legal experts say. AI hallucinations are cropping up more and more in legal filings. And it’s starting to infuriate judges. Just consider these three cases, each of which gives a glimpse into what we can expect to see more of as lawyers embrace AI. A few weeks ago, a California judge, Michael Wilner, became intrigued by a set of arguments some lawyers made in a filing. He went to learn more about those arguments by following the articles they cited. But the articles didn’t exist. He asked the lawyers’ firm for more details, and they responded with a new brief that contained even more mistakes than the first. Wilner ordered the attorneys to give sworn testimonies explaining the mistakes, in which he learned that one of them, from the elite firm Ellis George, used Google Gemini as well as law-specific AI models to help write the document, which generated false information. As detailed in a filing on May 6, the judge fined the firm $31,000. Last week, another California-based judge caught another hallucination in a court filing, this time submitted by the AI company Anthropic in the lawsuit that record labels have brought against it over copyright issues. One of Anthropic’s lawyers had asked the company’s AI model Claude to create a citation for a legal article, but Claude included the wrong title and author. Anthropic’s attorney admitted that the mistake was not caught by anyone reviewing the document. Lastly, and perhaps most concerning, is a case unfolding in Israel. After police arrested an individual on charges of money laundering, Israeli prosecutors submitted a request asking a judge for permission to keep the individual’s phone as evidence. But they cited laws that don’t exist, prompting the defendant’s attorney to accuse them of including AI hallucinations in their request. The prosecutors, according to Israeli news outlets, admitted that this was the case, receiving a scolding from the judge. Taken together, these cases point to a serious problem. Courts rely on documents that are accurate and backed up with citations—two traits that AI models, despite being adopted by lawyers eager to save time, often fail miserably to deliver. Those mistakes are getting caught (for now), but it’s not a stretch to imagine that at some point soon, a judge’s decision will be influenced by something that’s totally made up by AI, and no one will catch it. I spoke with Maura Grossman, who teaches at the School of Computer Science at the University of Waterloo as well as Osgoode Hall Law School, and has been a vocal early critic of the problems that generative AI poses for courts. She wrote about the problem back in 2023, when the first cases of hallucinations started appearing. She said she thought courts’ existing rules requiring lawyers to vet what they submit to the courts, combined with the bad publicity those cases attracted, would put a stop to the problem. That hasn’t panned out. Hallucinations “don’t seem to have slowed down,” she says. “If anything, they’ve sped up.” And these aren’t one-off cases with obscure local firms, she says. These are big-time lawyers making significant, embarrassing mistakes with AI. She worries that such mistakes are also cropping up more in documents not written by lawyers themselves, like expert reports (in December, a Stanford professor and expert on AI admitted to including AI-generated mistakes in his testimony). I told Grossman that I find all this a little surprising. Attorneys, more than most, are obsessed with diction. They choose their words with precision. Why are so many getting caught making these mistakes? “Lawyers fall in two camps,” she says. “The first are scared to death and don’t want to use it at all.” But then there are the early adopters. These are lawyers tight on time or without a cadre of other lawyers to help with a brief. They’re eager for technology that can help them write documents under tight deadlines. And their checks on the AI’s work aren’t always thorough. The fact that high-powered lawyers, whose very profession it is to scrutinize language, keep getting caught making mistakes introduced by AI says something about how most of us treat the technology right now. We’re told repeatedly that AI makes mistakes, but language models also feel a bit like magic. We put in a complicated question and receive what sounds like a thoughtful, intelligent reply. Over time, AI models develop a veneer of authority. We trust them. “We assume that because these large language models are so fluent, it also means that they’re accurate,” Grossman says. “We all sort of slip into that trusting mode because it sounds authoritative.” Attorneys are used to checking the work of junior attorneys and interns but for some reason, Grossman says, don’t apply this skepticism to AI. We’ve known about this problem ever since ChatGPT launched nearly three years ago, but the recommended solution has not evolved much since then: Don’t trust everything you read, and vet what an AI model tells you. As AI models get thrust into so many different tools we use, I increasingly find this to be an unsatisfying counter to one of AI’s most foundational flaws. Hallucinations are inherent to the way that large language models work. Despite that, companies are selling generative AI tools made for lawyers that claim to be reliably accurate. “Feel confident your research is accurate and complete,” reads the website for Westlaw Precision, and the website for CoCounsel promises its AI is “backed by authoritative content.” That didn’t stop their client, Ellis George, from being fined $31,000. Increasingly, I have sympathy for people who trust AI more than they should. We are, after all, living in a time when the people building this technology are telling us that AI is so powerful it should be treated like nuclear weapons. Models have learned from nearly every word humanity has ever written down and are infiltrating our online life. If people shouldn’t trust everything AI models say, they probably deserve to be reminded of that a little more often by the companies building them. This story originally appeared inThe Algorithm*, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here.* From MIT Technology Review via this RSS feed
Komunitas
rss.ponder.cat
This is today’s edition of The Download,our weekday newsletter that provides a daily dose of what’s going on in the world of technology. The first US hub for experimental medical treatments is coming The news: A bill that allows clinics to sell unproven treatments has been passed in Montana. Under the legislation, doctors can apply for a license to open an experimental treatment clinic and recommend and sell therapies not approved by the Food and Drug Administration (FDA) to their patients. Why it matters: Once it’s signed by the governor, the law will be the most expansive in the country in allowing access to drugs that have not been fully tested. The bill allows for any drug produced in the state to be sold in it, providing it has been through phase I clinical trials—but these trials do not determine if the drug is effective.The big picture: The bill was drafted and lobbied for by people interested in extending human lifespans. And these longevity enthusiasts are hoping Montana will serve as a test bed for opening up access to experimental drugs. Read the full story. —Jessica Hamzelou Google DeepMind’s new AI agent cracks real-world problems better than humans can Google DeepMind has once again used large language models to discover new solutions to long-standing problems in math and computer science. This time the firm has shown that its approach can not only tackle unsolved theoretical puzzles, but improve a range of important real-world processes as well. The new tool, called AlphaEvolve, uses large language models (LLMs) to produce code for a wide range of different tasks. LLMs are known to be hit and miss at coding. The twist here is that AlphaEvolve scores each of Gemini’s suggestions, throwing out the bad and tweaking the good, in an iterative process, until it has produced the best algorithm it can. In many cases, the results are more efficient or more accurate than the best existing (human-written) solutions.Read the full story. —Will Douglas Heaven Research cuts are threatening crucial climate data —Casey Crownhart Over the last few weeks, there’s been an explosion of news about proposed budget cuts to science in the US. Researchers and civil servants are sounding the alarm that those cuts mean we might lose key data that helps us understand our world and how climate change is affecting it. Long-running US government programs that monitor the snowpack across the West are among those being threatened by cuts across the US federal government, as my colleague James Temple’s new story explores. Also potentially in trouble: carbon dioxide measurements in Hawaii, hurricane forecasting tools, and a database that tracks the economic impact of natural disasters. It’s all got me thinking: What do we lose when data is in danger? Read the full story. This article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here. The must-reads I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology. 1 Donald Trump doesn’t want Apple building iPhones in IndiaThe US President claims Apple will be upping their US production as a result. (Bloomberg $)+ He also said that India was willing to “literally charge us no tariffs.” (WSJ $) 2 Elon Musk’s Grok chatbot ranted about white genocideIn response to completely unrelated queries. (FT $)+ It’s not the first time Grok has shared questionable responses. (Bloomberg $)+ Grok told users it was instructed to accept white genocide as real. (The Guardian) 3 RFK Jr doesn’t think we should take his medical adviceWhich begs the question: why is he US Health and Human Services secretary? (NY Mag $)+ Kennedy said his opinions on vaccines are irrelevant. (NYT $)+ He defended his decision to downsize the health department amid protests. (The Guardian) 4 GM’s new EV battery can power a truck for more than 400 milesIts lithium manganese-rich cells use cheaper minerals than lithium-ion ones. (Fast Company $)+ Tariffs are bad news for batteries. (MIT Technology Review) 5 Anthropic has been accused of using AI-generated evidence in a legal caseA lawyer for Universal Music Group claimed an expert cited a source that didn’t exist. (Reuters)+ A judge in another case reportedly caught fake AI citations, too. (Ars Technica)+ AI companies are finally being forced to cough up for training data. (MIT Technology Review) 6 AI won’t put human radiologists out of a job any time soonThe technology is helpful, but is unable to do everything trained human experts can. (NYT $)+ Why it’s so hard to use AI to diagnose cancer. (MIT Technology Review) 7 The US Defense Department wants faster aircraft and missilesAnd startups are more than willing to answer the call. (WP $)+ Phase two of military AI has arrived. (MIT Technology Review)**8 SpaceX has successfully tested its Starship rocket **Clearing a major hurdle ahead of its planned launch later this month. (Wired $) 9 YouTube will start inserting ads into videos’ crucial momentsWow, that doesn’t sound annoying at all. (TechCrunch) 10 Apple’s Vision Pro headset is a pain in the neckAnd early adopters are regretting shelling out $3,500 apiece. (WSJ $)+ Maybe the ability to scroll using their eyes will change their minds. (Bloomberg $) Quote of the day “To say a professor is ‘some kind of monster’ for using AI to generate slides “is, to me, ridiculous.” —Paul Shovlin, a professor at Ohio University, reacts to student backlash against professors using AI to create teaching materials, the New York Times reports. One more thing **Who gets to decide who receives experimental medical treatments?**There has been a trend toward lowering the bar for new medicines, and it is becoming easier for people to access treatments that might not help them—and could even harm them. Anecdotes appear to be overpowering evidence in decisions on drug approval. As a result, we’re ending up with some drugs that don’t work.We urgently need to question how these decisions are made. Who should have access to experimental therapies? And who should get to decide? Such questions are especially pressing considering how quickly biotechnology is advancing. We’re not just improving on existing classes of treatments—we’re creating entirely new ones. Read the full story. —Jessica Hamzelou We can still have nice things A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.) Food nostalgia is the best nostalgia, and this Bluesky account of discontinued foods doesn’t disappoint.+ Don’t even think of calling your newborn baby King if you live in New Zealand.+ Actor Jeremy Strong just loves a bucket hat.+ Watch out Swiss drivers—a duck has been caught speeding From MIT Technology Review via this RSS feed
Komunitas
rss.ponder.cat
This is today’s edition of The Download,our weekday newsletter that provides a daily dose of what’s going on in the world of technology. A US court just put ownership of CRISPR back in play The CRISPR patents are back in play. Yesterday, the US Court of Appeals for the Federal Circuit said scientists Jennifer Doudna and Emmanuelle Charpentier will get another chance to show they ought to own the key patents on what many consider the defining biotechnology invention of the 21st century. The pair shared a 2020 Nobel Prize for developing the gene-editing system, which is already being used to treat various disorders.But when US patent rights were granted in 2014 to Feng Zhang of the Broad Institute of MIT and Harvard, the decision set off a bitter dispute in which hundreds of millions of dollars—as well as scientific bragging rights—are at stake. Read the full story. —Antonio Regalado To read more about CRISPR, why not take a look at:+ Charpentier and Doudna announced they wanted to cancel their own CRISPR patents in Europe last year. Read the full story.+ How CRISPR will help the world cope with climate change. Read the full story. The US has approved CRISPR pigs for food. Pigs whose DNA makes them resistant to a virus could be the first big consumer product using gene editing. Read the full story.+ CRISPR will get easier and easier to administer. What does that mean for the future of our species? Police tech can sidestep facial recognition bans now —James O’Donnell Six months ago I attended the largest gathering of chiefs of police in the US to see how they’re using AI. I found some big developments, like officers getting AI to write their reports. Now, I’ve published a new story that shows just how far AI for police has developed since then.It’s about a new method police are using to track people: an AI tool that uses attributes like body size, gender, hair color and style, clothing, and accessories instead of faces. It offers a way around laws curbing the use of facial recognition, which are on the rise. Here’s what this tells us about the development of police tech and what rules, if any, these departments are subject to in the age of AI. Read the full story. This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. The must-reads I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology. 1 Two Trump officials were denied access to the US Copyright OfficeTheir visit came days after the administration fired the office’s head. (Wired $)+ Shira Perlmutter oversaw a report raising concerns about training AI with copyrighted materials. (WP $) 2 Google knew it couldn’t monitor how Israel might use its cloud technologyBut it went ahead with Project Nimbus anyway. (The Intercept) 3 Spain still doesn’t know what caused its massive power blackoutInvestigators are examining generators’ cyber defences for weaknesses. (FT $)+ Could solar power be to blame? (MIT Technology Review) 4 Apple is considering hiking the price of iPhonesThe company doesn’t want to blame tariffs, though. (WSJ $)+ Apple boss Tim Cook had a call with Trump following the tariff rollback news. (CNBC)+ It’s reportedly developing an AI tool to extend phones’ battery life. (Bloomberg $) 5 Venture capitalists aren’t 100% sure what an AI agent isThat isn’t stopping companies from sinking millions into them. (TechCrunch)+ Google is working on its own agent ahead of its I/O conference. (The Information $)+ What AI assistants can—and can’t—do. (Vox)+ Check out our AI agent explainer. (MIT Technology Review) 6 Scammers are stealing the identities of death row inmatesAnd prisoners are unlikely to see correspondence alerting them to the fraud. (NBC News) 7 Weight-loss drugs aren’t always enoughYou need long-term changes in health, not just weight. (The Atlantic $)+ How is Trump planning to lower drug costs, exactly? (NY Mag $)+ Drugs like Ozempic now make up 5% of prescriptions in the US. (MIT Technology Review) 8 China’s e-commerce giants are racing to deliver goods within an hourAs competition has intensified, companies are fighting to be the quickest. (Reuters) **9 This spacecraft will police satellites’ orbits **And hunt them down where necessary. (IEEE Spectrum)+ The world’s biggest space-based radar will measure Earth’s forests from orbit. (MIT Technology Review) **10 Is your beard trimmer broken? Simply 3D-print a new part.**Philips is experimenting with letting its customers create their own replacements. (The Verge) Quote of the day “We usually set it up so that our team doesn’t get to creep in.” —Angie Saltman, founder and president of tech company Saltmedia, explains how her company helps store Indigenous data securely away from the Trump administration, the Verge reports. One more thing Meet the radio-obsessed civilian shaping Ukraine’s drone defenseDrones have come to define the brutal conflict in Ukraine that has now dragged on for more than three years. And most rely on radio communications—a technology that Serhii “Flash” Beskrestnov has obsessed over since childhood.While Flash is now a civilian, the former officer has still taken it upon himself to inform his country’s defense in all matters related to radio. Once a month, he studies the skies for Russian radio transmissions and tries to learn about the problems facing troops in the fields and in the trenches.In this race for survival—as each side constantly tries to best the other, only to start all over again when the other inevitably catches up—Ukrainian soldiers need to develop creative solutions, and fast. As Ukraine’s wartime radio guru, Flash may just be one of their best hopes for doing that. Read the full story. —Charlie Metcalfe We can still have nice things A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.) Tune in at any time to the Coral City Camera, an underwater camera streaming live from an urban coral reef in Miami + Inhuman Resources, which mixes gaming, reading, and listening, sounds nuts.+ This compilation of 331 film clips to recreate Eminem’s Lose Yourself is spectacular.+ Questions I never thought I’d ask: what if Bigfoot were British? From MIT Technology Review via this RSS feed
Komunitas
rss.ponder.cat
Google has announced a dedicated iPadOS version of the Gemini app. Previously, you could only run the iOS version of Gemini in compatibility mode on an iPad. Here’s some of the announcement, which Google spokesperson Elijah Lawal emailed to The Verge: With the new app, you can use the split view on the iPad to put Gemini side by side with other apps for multitasking. You can also effortlessly use powerful features like: • Gemini Live: Have natural, free-flowing conversations in over 45 languages. • Deep Research: Quickly generate reports and gather information, saving significant time. • Audio Overview: Easily process and understand audio files. • Canvas: Co-create and edit documents and code in an innovative space built for collaboration with AI. • Image and video generation: Bring your ideas to life by generating stunning, varied images and videos directly on your iPad. The dedicated iPad app is available today in all countries where Gemini is currently offered. You can download it directly from the App Store. Speaking of updated features, we’ve significantly expanded the availability of Audio Overview. Following great initial feedback, this feature, which initially launched in English for Gemini and Gemini Advanced subscribers, is now available in over 45 languages globally. According to the listing, Gemini can also now be added as a homescreen widget, and supports connecting to your Google Photos library. Google launched Gemini for iOS in November. And while it would’ve been nice for it to debut an iPad-native app at the time, the next best time, as they say, is now. And anyway, it’s not like Google is the worst offender there. From The Verge via this RSS feed
Komunitas
rss.ponder.cat
Figma is expanding its creative software ecosystem to allow product designers to complete entire projects without jumping to third-party apps. Four new products for website building, AI coding, branded marketing, and digital illustration were introduced at Figma’s Config event today, aiming to fill in any gaps holding Figma back from being an all-in-one platform that supports the entire product design lifecycle. For example, while existing products like Figma Design, Slides, and FigJam could be used to ideate and create prototypes, developers would need to use services like WordPress to create live websites or Adobe Illustrator to create customized and scalable brand imagery. Figma’s first solution is Figma Sites, a website builder that integrates with Figma Design and allows creators to turn their projects into live, functional sites. Figma Sites provides presets for layouts, blocks, templates, and interactions that aim to make building websites less complex and time-consuming. Additional components like custom animations can also be added either using existing code or by prompting Site’s AI tool to generate new interaction codes via text descriptions, such as “animate the text to fall into place like a feather.” Figma Sites is rolling out in beta for users with full seat access to Figma products. Figma says that AI code generation will be available “in the coming weeks,” and that a CMS that allows designers to manage site content will be launched “later this year.” Figma Make is Figma’s take on AI coding tools like Google’s Gemini Code Assist and Microsoft’s GitHub Copilot. The prompt-to-code Figma Make tool is powered by Anthropic’s Claude 3.7 model and can build working prototypes and apps based on descriptions or existing designs, such as creating a functional music player that displays a disc that spins when new tracks are played. Specific elements of working design, like text formatting and font style, can be manually edited or adjusted using additional AI prompts. Make is rolling out in beta for full seat Figma users. Figma says it’s “exploring integrations with third parties and design systems” for Figma Make and may apply the tool to other apps within its design platform. Figma Buzz is a marketing-focused design app that’s rolling out in beta to all users, and makes it easier for teams to publish brand content, similar to Canva’s product design platform. The tool allows Figma designers to create brand-approved templates, styles, and assets that can be used by marketers to quickly assemble emails, social media posts, advertising, and more. Figma Buzz includes generative AI tools for making and editing images using text prompts, and can source information from spreadsheets to bulk create thousands of image assets at once. Lastly, the Figma Draw vector design app is like a simplified version of Adobe Illustrator that creatives can use to make custom visuals without leaving the Figma platform. It includes a variety of brushes, texture effects, and vector editing tools to create or adjust scalable images and logos for product design projects. Figma Draw is generally available now for full seat users as a toggle in Figma Design, with some features accessible in Sites, Slides, and Buzz. It’s not quite as expansive as Adobe’s wider Creative Cloud ecosystem, but Figma Draw places the two companies in direct competition for the first time since Adobe killed its own XD product design platform. It also brings some new options to the creative software industry after Adobe failed to acquire Figma for $1 billion due to pressure from competition regulators. From The Verge via this RSS feed
Komunitas
rss.ponder.cat
Google has a new AI tool that can help iPhone users to better grasp complicated or confusing writing online. The Simplify feature, rolling out in the Google app on iOS starting today, generates a simpler, more digestible version of any highlighted text without leaving the current web page. Simplify is built on Google’s Gemini AI model, and was developed by Google Research to make technical jargon easier for anyone to understand without losing key details. For example, Simplify can break down medical terms like “emphysema” (a condition that damages the air sacs in lungs) and “fibrosis” (dense connective tissue or scarring that develops in response to damage) in reports and journals, preventing readers from needing to reference terminology on a separate web page. Google says that people found the simplified versions to be “significantly more helpful than the original complex text” in its testing, but acknowledged the study “has limitations” and that “ongoing vigilance” is required to monitor errors. The feature can be found in the iOS Google app by highlighting any text on a website, and tapping the Simplify icon from the menu options that appear. When asked if Simplify will be made available for Android and desktop Chrome users, Google spokesperson Jennifer Kutz told The Verge that “we don’t have anything to announce yet, but we’re always looking to bring useful features to more of our products.” From The Verge via this RSS feed
Komunitas
discuss.tchncs.de
Seems yours is different than mine, there are two things called vim. There is a Vim text editor and a Vim dishwash bar soap. For some reason Gemini thinks you want to eat the text editor which honestly is very weird and if you scroll down it talks about why it is dangerous to eat soap. By the way the results changed to soap now, seems someone noticed how weird the answer was. This proves just how stupid Gemini is, would you ever think someone eats text editers?
Komunitas
rss.ponder.cat
On Thursday, Anthropic announced significant upgrades to its AI assistant Claude, extending its research capabilities to run for up to 45 minutes before delivering comprehensive reports. The company also expanded its integration options, allowing Claude to connect with popular third-party services. Much like Google’s Deep Research (which debuted on December 11) and ChatGPT’s deep research features (February 2), Anthropic first announced its own “Research” feature on April 15. Each can autonomously browse the web and other online sources to compile research reports in document format, and open source clones of the technique have debuted as well. Now, Anthropic is taking its Research feature a step further. The upgraded mode enables Claude to conduct “deeper” investigations across “hundreds of internal and external sources,” Anthropic says. When users toggle the Research button, Claude breaks down complex requests into smaller components, examines each one, and compiles a report with citations linking to original sources. Read full article Comments From Ars Technica - All content via this RSS feed