next gen computer-use agents @salesforce | ex @convergence_ai_ @UnitaryAI @CompSciOxford @imperialcollege | interested in steering ai to do more good than bad
London, England
Joined August 2018
- Tweets220
- Following528
- Followers618
- Likes1.2K
Pinned Tweet
Excited to release code and models for detoxify 🙊, a simple python library built @UnitaryAI that aims to detect hate speech and toxic comments online! Includes training scripts built with @PyTorchLightnin ⚡ and transformers checkpoints @huggingface 🤗!
github.com/unitaryai/detoxif…
this is also why i've chosen to work on computer use over the last 2 years, it's a rare combo of many genuinely interesting research problems and immediate real world impact
a recent personal highlight was showing my parents how to use @codex computer use and watch them set up automations in their website backend despite having no real coding background ☺️
just a small example of how it can empower more people without software backgrounds to build things for themselves
New blog post: What's the Point of Computer Use Agents?
There's been a lot of hype about CUAs recently. But where do CUAs shine? When should we use CUAs over API- or text-based agents?
And what are the remaining research problems?
jykoh.com/blog/whats-the-poi…
watching @3blue1brown's beautiful new video on why compression is intelligence youtube.com/watch?v=GlYgs6v2… reminded me of @jack_w_rae's great lecture on compression for agi 3 years ago youtube.com/watch?v=dO4TPJke…
one nuance that often gets missed in this framing, though it's become much clearer in the last 1-2 years with the renewed focus on rl, is that compression is intelligence *given the data*. the missing bit is getting the right data through search. pretraining llms works great for compressing the outputs of human reasoning into a prior, something akin to intuition. this prior then makes it much easier to learn the process of reasoning by drastically narrowing down the search space for rl during post-training
Laura Hanu retweeted
Watch me control my computer with just my voice. This is the future of operating systems.
No hands.
GPT-Realtime 2.0 is very, very underrated.
Demo:
say what you will about ai slop but it’s so cool that you can just generate a 4 page comic about the history of epistemology
these are with @OpenAI images 2.0
🤖 Made with AI
this was a balanced, fun and surprisingly touching doc, @DanielRoher did well capturing the ai zeitgeist and its defining promise/peril tension in a very human way, kudos to @Fremond_ @ry_paddy for getting their hands on it & organising the watch party
Laura Hanu retweeted
claude cowork is making me think maybe we’ll look back and it’ll be obvious that humans were never meant to spend their lives working behind a screen. we’ll see it as inevitable that computers do everything for us on computers and the future of work is cooler than we can imagine
pretty good high level intro to the current state of llms youtube.com/watch?v=xCRvOUyk…
main takeaways from the @dwarkeshpodcast @karpathy interview:
*RL limitations*: hard to get past reward sparsity problem in RL when it comes to real world tasks, one promising direction could be more sample efficient learning by reflecting on mistakes
*reduced model size by getting rid of redundant memorisation*: soon the cognitive engine might look like a 1 billion parameter model that just knows how to think without the need to memorise lots of data - it should be able to recognise what it doesn't know and look it up
*gradual loss of control and understanding*: even if we still have humans delegating tasks to autonomous entities, it will get increasingly harder to fully control them, let alone understand what they're doing; similar thoughts in this great albeit on the pessimistic side paper arxiv.org/abs/2501.16946
*role of AI education post-AGI*: pre-AGI AI education is useful, post-AGI education is fun, could be seen as a way to train mentally just how you do physically
pantheon is now up there with some of my favourite sci fi of all time like asimov’s the end of eternity, the last question or story of your life with some elements of snow crash or dark the tv show but upping the scale even more, dare i say something akin to what @DavidDeutschOxf was describing in the beginning of infinity
a nice straight forward summary on some of the grpo limitations beyond just not being great for multi-turn e.g.
* if you have multiple reward signals -> the model won't know which one it is being rewarded for since they're usually all collapsed into one
* only the scalar reward is used for policy update when a more detailed textual feedback could be used (what gepa kinda does with their reflective prompt evolution)
nice watch for a reading group on verifiers, like a jigsaw falling into place🤭
Video summary for
"Prover-Verifier Games improve legibility of LLM outputs"
youtu.be/EMDa4urzz-M
1/2
Why does RL work so well to learn preferred completions in llm post training and why can't we just use supervised finetuning?
This paper has an interesting information-theoretic explanation:
arxiv.org/pdf/2503.01067
TLDR:
The literature shows that the 2 stage RL approach (1. learn a good reward model that can score llm completions similar to how a human would 2. use that to learn good policies with RL) used today in the sota models outperforms using only supervised finetuning. The authors posit that this happens when the verification of an output is simpler than generating it. This is because, in the RL case, the search space is constrained to the subset of policies that are optimal for the learnt reward model. When they reduce the verification-generation gap empirically the difference in results diminishes too.
It's been interesting to see RL having a comeback lately. Guess in retrospect it's not surprising that RLHF is not enough since it naturally hits a ceiling limited by the quality of human feedback. What's particularly exciting though is that this shift is happening just as AI is becoming more agentic and able to interact with the digital world. Combine this with some grounded reward signals (e.g. profit, likes, citations) and we should see AI not just replicate what humans do but create new knowledge through their own experiences. The next few years should be a fun and wild ride 🎢
For another interesting paper on the topic: incompleteideas.net/papers/T…
or this podcast on it:
youtu.be/zzXyPGEtseI?si=5VL-…
note to self, don’t start addictive puzzles late at night, but also level 4 lets go?🕵️♀️ fairly consistent strategy so far 👀
We challenge you to break our new jailbreaking defense!
There are 8 levels. Can you find a single jailbreak to beat them all?
claude.ai/constitutional-cla…
great discussion! we need more public discourse between top ai leaders and economists 🤝
would’ve liked to see them engage with hinton’s concern that the benefits ai brings will deepen the economic divide in a capitalist system further creating fertile ground for fascism 🌶️
Watch our 2024 Nobel Prize laureates talk about their research and careers in a unique roundtable discussion, 'Nobel Minds', moderated by BBC's Zeinab Badawi.
youtu.be/1tELlYbO_U8
Laura Hanu retweeted
Introducing Cambrian-1, a fully open project from our group at NYU. The world doesn't need another MLLM to rival GPT-4V. Cambrian is unique as a vision-centric exploration & here's why I think it's time to shift focus from scaling LLMs to enhancing visual representations.🧵[1/n]
The 2nd part of the Intro to multi-node machine learning is out! 🥳 This all about how to use Slurm to scale up your ML applications! 🚀
unitary.ai/articles/intro-to…
If you haven't seen it, part 1 dives into how to set up a HPC cluster!
unitary.ai/articles/intro-to…
