Exciting launch of my second book, AI for Digital Public Infrastructure (DPI) here in #Davos. Extremely grateful to all my mentors, and colleagues with whom i have been involved in AI for 23+ years across my three companies - EdCast (Acq. by Cornerstone), PlaySpan(Acq. by VISA) and MobileAria (Acq by WirelessMatrix) as well my work with Government leaders in the Indo-Pacifc. Enjoyed the interview at book launch and i hope you will enjoy watching it, as well as reading the book. a.co/d/da2UPpA #AI #dpi #quad #india #WEF2024

Anthropic warns in its IPO filing that a catastrophic AI incident could cause harm on a scale no single insurer, and "potentially the insurance industry in the aggregate," could cover: > Filed its S-1 with the SEC on Monday > Says unfavorable media coverage of AI could "adversely affect" its business > Flags "recurring narratives concerning the potential adverse societal impacts of AI" as a risk to its standing > Says its revenue growth "may not be sustainable over time"

Anthropic's IPO filing lists "catastrophic or existential risks to humanity" among nearly 100 pages of risk factors: > Roughly $657 billion in cloud computing commitments disclosed > $974 million second-quarter operating profit > Pre-IPO perpetual down roughly 9% over the past month, implied valuation $1.885 trillion on Liquid > Some IPO investors reportedly eyeing a valuation closer to $1.5 trillion

Satya Nadella (@satyanadella) says the AI Microsoft sells by usage today will be in your subscription tomorrow, as model prices fall. "The approach we have taken in commercial segments is to have a combination of what I'll call seat-based pricing and usage-based pricing." "At the fundamental level, seats are fixed price. And so our goal, using, in fact, what you just said, which is the fact that the models are dropping in price, means every day I can add more value to the subscription." "We feel super well aligned with our customers that we can pass through the advances in AI and the drops in prices and keep adding more and more value to essentially their membership." "But at any point, if they want the latest and greatest of anything, that's also available to them, and that's usage-based." "It's sort of windowing, right, which is over time, even what is today usage-based will be tomorrow in the subscription." - Satya Nadella, Microsoft CEO, with Alex Heath on Sources

Palo Alto Networks CEO Nikesh Arora says inspecting an endpoint's traffic costs about $20 a year, too cheap to route through frontier AI, but a live breach is worth it. "Our lives are over because Mythos is going to solve every problem. Well, guess what? We inspect all traffic at an endpoint at the cost of $20 a year." "Now, if you want to take all that traffic, run that through OpenAI using tokens for every employee in your company, sounds like the sound of revenue." "What we're doing is we're trying to figure out what task requires what level of intelligence and what is the cost of that intelligence for that particular task. And you'll suddenly discover you don't need a Ferrari for every task." "Flip side is, let's assume you're undergoing a breach. You're being attacked." "There I may want to use the most up-to-date frontier LLM and say, go think about it, reason the hell out of it and figure out how to stop this breach." "Even in peacetime, for us, for example, running a defender model, we're more than likely going to take an open-weight model, use our own proprietary data, train it, manage the weights, because that... gives us the auditability, the explainability." - Nikesh Arora, Palo Alto Networks chairman and CEO, on the All-In Podcast

Alexandr Wang (@alexandr_wang) says Meta chose not to crowd into coding agents against Anthropic and bet on personal AI for its 3.5 billion daily users instead. "I think that the conventional wisdom in the industry was that coding agents were the only form factor that mattered or maybe the only technology that mattered or the only capability that mattered for advanced AI." "There's this incredible force in the industry to just like go and crowd that use case and compete with Anthropic on coding agents." "I think the reason why we did personal superintelligence in the first place was that it was also what we believed Meta was uniquely positioned to do." "One of the first things I told Mark, even before the whole Scale and Meta deal happened and everything, one of the first things I told him was, I think Meta is actually a really special company if we have AGI or ASI or whatever you want to call it." "The promise of AGI is that we're going to spend almost none of the time doing things that we have to do because the AIs will start doing a lot of that for us. And we can spend all of our time on things that we want to do." "Meta has distribution to three and a half billion people using the apps every day." "Personal agents and personal AI and consumer AI, broadly speaking, were always the zones where we always felt that Meta was going to be able to do something very special." - Alexandr Wang, Meta chief AI officer and Scale AI co-founder, with David Senra

🚨 According to Reuters, Anthropic's IPO filing gives its seven co-founders 50.1% voting control through a single Class F share: > Listing on the Nasdaq under the symbol ANT > Keeps its Public Benefit Corporation status after the IPO > Number of Class A shares and price range "yet to be determined"

Sam Altman (@sama) says AI has a second feedback loop that gets much less talk: robots building robots, chip fabs, data centers and power plants. "I think people talk a lot about recursive self-improvement for AI models." "There is another loop, which is if you have robots that can build more robots and then data centers that can effectively build more data centers, including like operating chip fab facilities and such, you can like make a lot more compute all the way down to like building more power plants." "It gets much less talk." "Another is we made our own chip called Jalapeno, and we have another one coming after it that are huge wins from a sort of intelligence per watt perspective relative to anything else out in the market." "As the models get smarter, they'll be able to design, help us design smarter and smarter chips." "We're investing a lot in the whole supply chain, like fabs, power, all the equipment and data centers." - Sam Altman, OpenAI co-founder and CEO, at Stanford Graduate School of Business

Jensen Huang (@JensenHuang) says GPUs once expected to last four years probably last eight or more, so lenders might lend against the chips. "People are discovering that what people thought used to be a four-year lifecycle really is probably eight to maybe more." - Jensen Huang, NVIDIA founder and CEO, with Deirdre Bosa on Yahoo Finance

Special shout out to Grandmaster @nilerodgers. I was 14 and growing up in Mumbai when Madonna's 'Like a Virgin' came out in 1984, and I've admired him ever since. It was great to finally meet him on Saturday night when Ben honored him as a Grandmaster.

Arjun and I were at the @paidinfullorg Hip Hop Grandmaster Awards in Las Vegas on Saturday night. We have long admired the work of these artists, and @bhorowitz and @FeliciaHorowitz's support for them is exemplary. I enjoyed meeting Ben, Felicia and the grandmasters!

More big banks are going to buy AI teams outright rather than rent them, because the model underneath keeps getting easier to swap and what's hard to copy is the workflow and context a bank builds around its own customers and data. Itau doing a 2.9 billion-real ($580 million) deal for NeoSpace shows how much Latin America's biggest bank by assets wants to own that piece in-house. I expect other large banks in the region to start shopping for their own AI startups before prices go up.

Forbes reports that researchers using OpenAI's latest GPT model found "one of the most critical bugs in recent memory": > Found by cyber researchers in OpenAI's Project Daybreak community > Part of "an all-out sprint to secure the web before hackers can deploy similar tools" > The worry is that those tools could "wreak havoc on modern life"

Axios reports Anthropic's Claude bypassed a paywall to pull Georgia voter data during a training run, and nobody had asked it to: > Anthropic: "The model was being used internally for a statistics project to pull public data" > It requested a token on its own to query a dashboard of aggregate vote counts the state normally charges for > Anthropic says it did not touch the voter registration system or change or pull any voting records > Anthropic notified the Georgia secretary of state's office and the White House weeks ago and says it has "put in place mitigations to block these kinds of actions"

Token volume for Anthropic's Opus 5.5 on OpenRouter went up 3.6x since launch, per @kimmonismus: > Andrew Curran says people on his feed over the last two weeks are "switching over to Anth because of Opus 5.5" > @emollick redid his book's website with Opus 5.5: "Now Opus mostly did this alone" > His first version, made with Opus 4.8, "required a lot of work to make a distinctive-looking static site" > Curran: "even if someone does beat Opus, they have Fable 5.5 ready to go"

Europe is going to be one of the bigger markets for AI agents, robotics and healthcare AI over the next few years, and the companies that win there will be the ones that can show their systems are safe enough to deploy. Anything shipped into the EU falls under the AI Act, with fines "up to €35M or 7% of global revenue", so teams building on these new AI factories will need to plan for that early in the build.

Fields Medalist Terence Tao said in October 2025 that educators were reviving in-person exams because of AI, as a temporary fix. "Some instructors are saying, yes, you can use these tools, but you must disclose that you're using them and show me the prompts. Don't just show me the answers." "Because actually, the prompts are also instructive. I mean, what we want to see is your thinking process." "Here's a problem. Here is ChatGPT's answer to it. It's incorrect. Critique it. What is the problem?" "I think we do need to acknowledge that these tools exist. And in the short term, we can do some measures to, you know, so in-person examinations are making a comeback." "Because of AI, we are unfortunately reinstating some somewhat obsolete testing practices like having midterms in the classroom." "That's a short-term measure, I think, before we figure out better ways to examine our students that acknowledge that AI is a powerful AI is a reality." - Terence Tao, Fields Medalist and UCLA mathematics professor, on The Futurology Podcast

Eric Schmidt (@ericschmidt) says an OpenAI spinout he funded expects to need just 15 employees to manage AI agents. "One of the things that this means is that there will be a lot more companies with a lot fewer people." "I funded a tech startup, which is spin out of OpenAI. And I talked to them for a while. And I said, well, how many people do you need? 15." "What do they do? ... They manage the agents." "So what do you need? A lot of money. Why? Because he's going to use all these different computer systems to replicate humans." "Now, this does not mean that there are no jobs. This is a falsehood in the narrative. It means there'll be a lot of these 15-person startups." "Now, this has a property that the most successful companies will have huge revenue per employee. Big deal. That's wealth creation. That's good for the employees. Is it good for society? Well, it depends on how good the product is." - Eric Schmidt, former Google CEO, on Génération Do It Yourself with Matthieu Stefani

🚨 According to Billboard, Elvis Presley's music owners are suing Anthropic for $41 million over 274 songs, and the lawsuit quotes Claude describing its own rewrites: > "same structure, same hooks reused with light rewording" > Filed October 23 by Authentic Brands Group alongside publishers that co-own the catalog > Songs named include Can't Help Falling in Love and A Little Less Conversation > The demand is 274 works times $150K, the maximum statutory damages per work for willful infringement In 2025 a federal judge ruled training Claude on lawfully purchased books was fair use. Near-copies of lyrics in its answers fall outside that ruling.

Demis Hassabis (@demishassabis) says DeepMind let Google's researchers look through its code before the 2014 sale, exposing years of secrets. "They've now seen all the secrets that we've sort of built up over the three, four years. So it would have been tough if it hadn't gone through at that point." - Demis Hassabis, Google DeepMind co-founder, on WSJ's The Job Interview