ainch

It's becoming impossible to keep up - in the last week or so we've had: Gemini 3 Deep Think, Gemini 3.1 Pro, Claude Sonnet 4.6, GPT-5.3-Codex Spark, GLM-5, Minimax-2.5, Step 3.5 Flash, Qwen 3.5 and Grok 4.20.

and I'm sure others I've missed...

He's always said ARC is a necessary but not sufficient condition for testing intelligence afaik

The fact it's still Jan 2025 is weird to me. Have they not have a successful pretrain in over a year?

Knowing a couple people who work at Anthropic or in their particular flavour of AI Safety, I think you would be surprised how sincere they are about existential AI risk. Many safety researchers funnel into the company, and the Amodei's are linked to Effective Altruism, which also exhibits a strong (and as far as I can tell, sincere) concern about existential AI risk. I personally disagree with their risk analysis, but I don't doubt that these people are serious.

Is it proven that they serve the models at cost? Amodei has said that Anthropic's models make back their training cost - the reason they're so deep in the red is because they're investing substantially more in subsequent runs, and R&D dwarfs inference cost[1]. If the tech plateaus I would expect to see a lot of that R&D spend move into just powering inference.

[1] https://epoch.ai/data-insights/openai-compute-spend

I think it'll be 3.1 by the time it's labelled GA - they said after 3.0 launch that they figured out new RL methods for Flash that the Pro model hasn't benefitted from.

I find Gemini's web page much snappier to use than ChatGPT - I've largely swapped to it for most things except more agentic tasks.

The Treasury is the finance ministry in charge of all public income and spending.

It's a week-by-week injection - you can always stop taking it if you're unhappy with the effects.

Moltbot is an attempt to do that. Would you hire it as a personal secretary and entrust all your personal data to it?

As a user that doesn't pay for the subscription, I use their Google Drive integration quite often. I believe they support Dropbox and OneDrive too, no?

It's a good paper by Hooker but that specific comparison is shoddy. Llama and Aya were both trained by significantly more competent labs on different datasets to Falcon and BLOOM. The takeaway there is "it doesn't matter if you have loads of parameters if you don't know what you're doing."

If we compare apples-to-apples, eg. across Claude models, the larger Opus still happily outperforms it's smaller counterparts.

I'm not going to bat for GPTZero, but I think it's clearly possible to identify some AI-written prose. Scroll through LinkedIn or Twitter replies and there are clear giveaways in tone, phrasing and repeated structures (it's not just X it's Y).

Not to say that you could ever feasibly detect all AI-generated text, but if it's possible for people to develop a sense for the tropes of LLM content then there's no reason you couldn't detect it algorithmically.

OpenAI are testing ads in the free tier of ChatGPT, but they state that the actual LLM responses won't include advertising/product placement [0].

[0]: https://openai.com/index/our-approach-to-advertising-and-exp...

I'm not sure Anthropic will become ad-supported - the vast bulk of their revenue is b2b. OpenAI have an enormous non-paying consumer userbase who are draining them of cash, so in their case ads make a lot more sense.

As LLMs are productionised/commodified they're incorporating changes which are enthusiast-unfriendly. Small dense models are great for enthusiasts running inference locally, but for parallel batched inference MoE models are much more efficient.

Perhaps they need more advertising around the correct spelling of his name.

Gemini 3.0's cutoff is January. I think you can get away with it if the model has good search/tool use capability.

Which other group is that?

Of course people don't say it, but there are many cases where reported algorithmic improvements are attributable to poor baseline tuning or shoddy statistical treatment. Tao is exhibiting a lot more epistemic humility than most researchers who probably have stronger incentives to market their work and publish.

Inferior in what sense? Genie 3 is addressing a fundamentally different problem to a physics sim or procgen: building a good-enough (and broad-enough) model of the real world to train agents that act in the real world. Sims are insufficient for that purpose, hence the "sim2real" gap that has stymied robotics development for years.

That's a good correction, thanks.

The reports are definitely bland, but I find them very helpful for discovering sources. For example, if I'm trying to ask an academic question like "has X been done before," sending something to scour the internet and find me examples to dig into is really helpful - especially since LLMs have some base knowledge which can help with finding the right search terms. It's not doing all the thinking, but those kind of broad overviews are quite helpful, especially since they can just run in the background.

Generally you train each expert simultaneously. The benefit of MoEs is that you get cheap inference because you only use the active expert parameters, which constitute a small fraction of the total parameter count. For example Deepseek R1 (which is especially sparse) only uses 1/18th of the total parameters per-query.

That's an interesting idea, it sounds similar to the principles behind low precision models like BitNet (where each weight is +-1 or 0).

That said, I know Deepseek use fp32 for their gradient updates even though they use fp8 for inference. And a recent paper shows that RL+LLM training is shakier at bf16 than fp16, which would both imply that numerical precision in gradients still matters.

Cutting people at FAIR is a real shame though - great models like DINO and SAM have had massive positive impact - hopefully that work doesn't slow in favour of LLM-only development at MSL.

I think you're overstating the distinction between ML and generation - plenty of ML methods involve generative models. Even basic linear regression with a squared loss can also be framed as a generative model derived by assuming Gaussian noise. Probabilistic PCA, HMMs, GMMs etc... generation has been a core part of ML for over 20 years.

There is a lot of overlap between AI and Neuroscience, especially among older researchers. For example Karpathy's PhD supervisor, Fei-Fei Li, researched vision in cat brains before working on computer vision, Demis Hassabis did his PhD in Computational Neuroscience, Geoff Hinton studied Psychology etc... There's even the Reinforcement Learning and Decision Making conference (RLDM - very cool!), which pairs Reinforcement Learning with neuro research and brings together people from both disciplines.

I suspect the average AI researcher knows much more about the brain than typical CS students, even if they may not have sufficient background to conduct research.

Their recent paper suggests the active user base is continuing to grow consistently with consistent/growing usage based on how long they've been using the app.

https://cdn.openai.com/pdf/a253471f-8260-40c6-a2cc-aa93fe9f1...

Gemini 2.5-Pro was great when it released, but o3 and GPT-5 both eclipsed it for me—the tool use/search improvements open up so many use cases that Gemini fails at.