Knowledge catalogue
arXiv:2508.10016v3 Announce Type: replace Abstract: Building interactive omni-modal assistants often relies on end-to-end multimodal alignment to fuse heterogeneous modalities, which incurs substantia
arXiv:2605.06990v1 Announce Type: new Abstract: Multimodal self-supervised learning (MSSL) has emerged as a key paradigm for pretraining geospatial foundation models. However, existing geospatial MSSL
arXiv:2410.21438v3 Announce Type: replace Abstract: By pretraining on trillions of tokens, an LLM gains the capability of text generation. However, to enhance its utility and reduce potential harm, SF
arXiv:2408.15339v4 Announce Type: replace-cross Abstract: RL alignment methods, including RLHF and DPO, are primarily based on pairwise preference data. Although scalar or score-based feedback has bee
arXiv:2605.07356v1 Announce Type: new Abstract: Semantic segmentation of large-scale 3D point clouds is crucial for applications such as autonomous driving and urban digital twins. However, the sparse
arXiv:2508.02129v2 Announce Type: replace Abstract: Dynamic urban scene modeling is a rapidly evolving area with broad applications. While current approaches leveraging neural radiance fields or Gauss
arXiv:2602.10693v3 Announce Type: replace-cross Abstract: Off-policy updates are inevitable in reinforcement learning (RL) for large language models (LLMs) due to rollout staleness from asynchronous t
arXiv:2605.05848v2 Announce Type: replace-cross Abstract: Video large multimodal models increasingly face a scalability bottleneck: long videos produce excessively long visual-token sequences, which s
arXiv:2605.06094v2 Announce Type: replace-cross Abstract: Training VideoLLMs for complex reasoning remains challenging due to sparse sequence level rewards and the lack of fine grained credit assignme
what i have been saying for 6 years. maybe now you will believe it? 📁 Fei-Fei Li, former Google Chief Scientist, says the industry is dangerously fixated on language models. Most of the real economy i
arXiv:2605.06868v1 Announce Type: new Abstract: Fixed-budget nonconvex optimization can fail not because local descent is unstable, but because it is too stable: after reaching a nearby stationary poi
Which has better odds? Generative AI earning $1.6 trillion/year or the roulette wheel landing on zero? One estimate of how much annual revenue AI needs to “make sense”: 1.6 trillion. That’s four times
arXiv:2605.05558v2 Announce Type: replace Abstract: A natural intuition about the economics of AI agents is that, because agents can be replicated at very low marginal cost, agent labor may be supplie
arXiv:2605.08075v1 Announce Type: new Abstract: Decoding imagined speech from non-invasive brain recordings is challenging because imagined datasets are scarce and difficult to align temporally across
32,000 views for my tweet; 2.2 million for the nonsense it critiques. so typical. lies and distortions beating debunking by a factor of roughly 100. Wanna get a million views? Make stuff up. Take a ti
a tweet to keep in mind when Sam testifies this coming week How can anybody take seriously your claim that “Working towards prosperity for everyone, empowering all people, and advancing science and te
FAR more scary than the misunderstood METR time horizon graph everyone here seems to be (wrongly) panicking about. The #AI circular funding bubble is already twice the size of the outstanding debt in
Happy to put money against superintelligence in 2029. Why Superintelligence is Alien Intelligence (A Brief Outline of the Future) We are no longer in the realm of normal technological progress. If the
hey @elonmusk, if you really care about making X a source of truth as you used to claim, take note: @GaryMarcus lies are optimized for engagement, truth is optimized for accuracy the algorithm doesnt
Many are valiantly fighting against the age of bullshit. Bullshit is winning, I'm sorry to report. 😠 32,000 views for my tweet; 2.2 million for the nonsense it critiques. so typical. lies and distorti
Oy. According to a new paper in The Lancet, the rate of made-up citations in biomedical papers has increased by more than 12x since 2023. https://www.thelancet.com/journals/lancet/article/PIIS0140-673
Remarkable numbers here. Goldman Sachs: 'The consensus of analysts is now for the mega-cap US hyperscalers to spend $755 billion on capex in 2026, representing growth of +83% vs. 2025. This capex is e
> shipped the agent > opened the dashboard > latency: fine > error rate: fine > users: unhappy > checked the responses > technically correct > wrong tool called 3 steps earlier > no trace to follow >
Terence Tao - 'AI tools are like taking a helicopter to drop you off at the site. You miss all the benefits of the journey itself. You just get right to the destination, which actually was only just a
The post asks about user engagement metrics for a specific mode within Codex, requesting data on what percentage of users utilize it and how adoption has changed over the past month. The author sugges
Wow! Wonder how many people on this site hyped this paper – and how many of them will walk back their hype now that the paper has been retracted. A year-old nature paper that advocated the use of Chat
This post likely discusses a cognitive or computational principle where one should primarily focus on horizontal (broad, general, or foundational) aspects of a problem or system, then use inference or
but as noted here, not zero threat, either: The detailed investigation of a prior hantavirus human-human transmission event has evidence of airborne transmission of Andes strain: Patient 1 --> Patient
Gary Marcus discusses how symbolic tools and verification methods may be undervalued in AI development, suggesting that if harnesses and verification systems are performing the primary computational w
David Shepardson / Reuters: GM agrees to pay 12.75M to resolve a California investigation into claims that it illegally sold the location and driving data of OnStar subscribers to brokers — GM (GM.N)
Gary Marcus publicly challenged Joseph Jacks with a $10,000 bet regarding a prediction Jacks made about 2030, requesting direct message contact to formalize the wager. The challenge was posted on X (f
Gary Marcus shares a humorous projection suggesting Anthropic could potentially reach $2 trillion in annual revenue if certain growth assumptions materialize. The post likely comments on AI industry r
🌷 If X had been around during the Dutch tulip craze🌷 1. Prices would have gone even higher before the crash 2. Tweets saying “there’s no bubble” would have gone viral. 3. Prices still would have crash
Is there a version of the bull case that spells out where the multitrillion dollar revenue will come from that doesn’t ultimately revolve around “trust me bro”? (And does it factor in the inevitable p
Questions people will be writing about in the coming years: • Was it a Ponzi scheme? • Why didn’t more people worry about the circular financing? • Why were investors so blasé about the lack of profit
Sorry, @peterwildeford, but this is wrong. Please don’t play along. The measurement “wall” you mention is hit ONLY if you don’t insist on reliability. If you demanded 95% accuracy on the task, the sys
A 2018 study by D.K. Roy and Soroush Vosoughi published in Science found that false information spreads faster on social media platforms than accurate information. The research demonstrates a quantifi
This post likely references the famous opening lyric from The Rolling Stones' 'Satisfaction' (1965) as a commentary on contemporary issues, possibly relating to AI confusion, information overload, or
Gary Marcus discusses the 'too big to fail' concept in the context of AI systems, likely examining how large AI companies or models may pose systemic risks that make them difficult to regulate or shut
'Trump’s most lethal policy will almost surely be his 71% cut in humanitarian aid from 2024 to 2025....The aid cuts cost more than 750,000 lives worldwide in their first year' and 'will cost 9.4M live
Wanna get a million views? Make stuff up. Take a tiny tiny bit of truth and distort it wildly. Consider the tweet below, 1.4M views. Take the chess thing. The paper that is linked doesn’t actually say
'50 percent of Americans told the Pew Research Center last year they were more concerned than excited about what’s to come from A.I. Only 10 percent said they were more excited. That is a yawning gap
Asking for HTML explanations of things is pretty neat, I tried it just now with the obfuscated Python POC for the new http://copy.fail Linux vulnerability: https://simonwillison.net/2026/May/8/unreaso
“Combined free cash flow across Microsoft, Alphabet, Amazon, Meta, and Oracle is projected to FALL more than -70%, to ~100 billion, by the end of …. AI capital expenditure is consuming nearly every do
Alex Lekander / CyberInsider: EU warns that VPNs are being used to bypass online age-verification systems, calling their use “a loophole in the legislation that needs closing” — The European Parliamen
Gary Marcus critiques the AI industry's financial model, noting that companies with high burn rates continue to see rising market valuations despite unprofitable operations. The post sarcastically que
Happy to see that The White House has finally realized that the Andreessen-Sacks zero regulation approach to AI was going to blow up in the their face. Also glad to see them finally consider the kind
Holy AI “fact checking” disaster We are in deep, deep trouble. A reader wrote in to me this week saying that they wouldn't read my Trump corruption story because ChatGPT 'fact-checked the piece' and i
I warned you early that covid-19 could get really serious. Hantavirus is not as easily transmitted; not likely to be the same level of threat. Nearly 40 cruise passengers exposed to a deadly virus may
i wish i could fast forward five years to see how badly this all turned out. it’s painful to watch. “Combined free cash flow across Microsoft, Alphabet, Amazon, Meta, and Oracle is projected to FALL m
Jan Leike expresses gratitude for collaborators he has worked with on AI alignment research throughout his career. The post, shared on X (Twitter), reflects on the privilege of working with talented i
It’s honestly painful to root for Elon in the trial, given the kind of person he has become. Elon Musk’s DOGE “blatantly used” race, gender and other protected characteristics to execute the largest m
A judge ruled that the Department of Government Efficiency (DOGE) violated regulations by using ChatGPT to process grant applications in an unauthorized manner, determining the practice was both incom
Morgan Stanley sees capital spending by the five largest AI hyperscalers topping 800 billion this year and 1.1 trillion next year. At 3.3% of gross domestic product, next year’s figure would exceed pr
My view on Mythos was and is somewhere in between. It’s real, it’s a wakeup call, and it’s not quite what some of the media coverage suggested. • Mythos is not going to allow an 8 year old to accident
Next week at The Trial could be wild. Here’s why: Sam Altman will be on the stand. He is one of the world’s most convincing (though not always truthful) talkers (as I saw up close when we testified si
Professions with guilds or membership associations are going to get different AI policy reactions than those without The Bar & the AMA will ensure that human doctors or lawyers are legally required fo
Remember that MIT study that showed that the ROI for generative AI wasn’t really there for most businesses? Or any of the six or seven studies from other teams that followed, showing basically the sam
Sadly I am forced to agree. @karaswisher shilled for her friend Sam (who interviewed her for her book tour!), and ran me down for seeking (and stating) the truth. And never apologized, even after my t
This appears to be a post by cognitive scientist and AI researcher Gary Marcus on X (formerly Twitter) expressing skepticism or disagreement about a claim or development, likely related to AI, machine