CEO & Co-founder, Synacti & Longshot Systems. Into Open Source, Software 2.0 and Statistics.
London
Joined January 2012
- Tweets271
- Following987
- Followers185
- Likes754
Pinned Tweet
In the future, the most popular programming language will be English githubnext.com/projects/ai-f…
Mark Goodhead retweeted
I’m calling it. Jailbreaks will be solved in 2024.
RepEng LFG.
No LLM is secure! A year ago, we unveiled the first of many automated jailbreak capable of cracking all major LLMs. 🚨
But there is hope?!
We introduce Short Circuiting: the first alignment technique that is adversarially robust. 🧵
📄 Paper: arxiv.org/abs/2406.04313
"Sam Altman has it. You could parachute him into an island full of cannibals and come back in 5 years and he'd be the king."
Mark Goodhead retweeted
If you'd asked me a year ago, superposition would have been by far the reason I was most worried that mechanistic interpretability would hit a dead end.
I'm now very optimistic. I'd go as far as saying it's now primarily an engineering problem -- hard, but less fundamental risk.
Mark Goodhead retweeted
1/ 11 weeks ago I agreed to Chair the UK's efforts to accelerate state capacity in AI Safety - measuring and mitigating the risks of frontier models so we can safety capture their opportunities. Here is our first progress report: gov.uk/government/publicatio…
📢An elite team of high-profile AI specialists have been recruited as advisors for the Frontier AI Taskforce.
The Taskforce will protect against risks, build on UK capabilities & improve public services.
@MichelleDonelan met Ian Hogarth @soundboy, Taskforce Chair, to discuss
I really hope this architecture doesn't take off because if I ever have to say the words "We replaced our Transformers model with a state of the art Hungry Hungry Hippos one" out loud to a real person I think I'll die inside
Attention is all you need... but how much of it do you need?
Announcing H3 - a new generative language models that outperforms GPT-Neo-2.7B with only *2* attention layers! Accepted as a *spotlight* at #ICLR2023! 📣 w/ @tri_dao
📜 arxiv.org/abs/2212.14052 1/n
I find it hilarious that the latent space of human interaction has 'fighting' and 'sex' this close together (h/t news.stanford.edu/2021/02/23…)
The saddest thing about this is Germany's historic solar subsidies helped kick start the green revolution that's brought forward decarbonisation timelines for the world.
It's like watching the kid who's top of the class go slowly insane.
I'm sorry but how is the world not losing it's shit that @DJFreshUK is now writing Machine Learning libraries in Rust?!
This is my favourite thing since finding out @MCHammer is big into AI research.
More famous musician / science crossovers in this timeline please!
Also shout out to Dexter Holland of @offspring for doing mRNA research. Music and science were made for each other.
Only Twitter could take a really balanced thread that just says "maybe we should give presumed innocence a try?" and then try and tear the poster apart
Many have expressed surprise about my interest in the FTX situation and my openness to the possibility that @SBF_FTX may be telling the truth as I have no economic interest or relationship with any of the parties. Perhaps some background on my life experiences will help explain.
It's always impressive when one tweet can be wrong in multiple ways. I guess that's what happens when AI goes mainstream
Does anyone know why the proposal referenced here has compiler advantages over say Jax or Pytorch's new compilation (or does it not)? Perhaps one of @cHHillee @SingularMattrix @jakevdp or @shoyer might be able to shed some light?
Oh wow, @labenz has produced the best thread on LLMs ever?
as an AI obsessive and long-time @ezraklein fan, I was excited to see yesterday's podcast with @GaryMarcus.
Unfortunately, as I listened, my excitement gave way to frustration, and I felt compelled to write my first-ever megathread.
Quotes from: nytimes.com/2023/01/06/podca…
Please somebody make this IRL as I'd totally wear this as a Halloween outfit. Makes me feel patriotic. And evil 🇬🇧
Wow it really didn't take long from me trying Copilot to becoming a Prompt Engineer more than a Software Engineer. Blown away.
At some point I expect there to be an AI stock market bubble on the scale of the dotcom one. We're probably in more like 1995 or so right now, but once the product disruption starts to be seen everywhere and the first IPOs pop it feels kind of inevitable?
Mark Goodhead retweeted
Centuries after probability theory was developed, many philosophers still treat belief and justification as binary rather than continuous properties. Now the theory of logical induction allows us to treat rationality as continuous too, but who knows how long until that sinks in?
