AI from @Cambridge_Uni - 20+ years coding experience - published AI research - deep tech startup veteran
Cambridge, England
Joined September 2007
- Tweets7.2K
- Following902
- Followers563
- Likes5.6K
Pinned Tweet
If human lifetimes are extended, maybe it doesn't matter. We have AI to come up with new ideas, so necrosis of thought doesn't matter that we don't die. This is neither good nor bad, it just is
There are three things that regularly make me thankful to have lived:
- to have loved
- to have been a father
- to have heard the music of J S Bach
OpenAI: We want to engage with mathematicians to push the frontier together
Mathematicians: don't just release loads of results
OpenAI: fuck it, YOLO!
Psychedelics seem to have done more to prepare me for the singularity than my formal education and experience in symbolic AI and ML
Neural nets are not deterministic:
The order of parallel oeprations can affect the result in binary floating point representation:
a + (b + c) != (a + b) + c
(think rounding errors)
And all those matmuls used in NNs use a LOT of sums.
This is particularly an issue where the output of one decision can effect future state, like in:
- LLM inference
- training
- RL
You can end up with quite incredible phase changes in behaviour
As a computer scientist, I've always found threesomes much more complex
The what now?!
arxiv.org/abs/2610.06783v1
The only entity that we can prove a concious is oneself.
You might as well all be stochastic parrots to me.
Tom Nicholson retweeted
new prime minister's first speech at the un. the big ai announcement: "uk ready to lead global push on ai standards"
lead on the rules. not on building it, not on the compute, not on the companies. on the standards...
the us and china are building the thing. we're volunteering to be the hall monitor. same energy as telling civil servants not to say thank you to gemini
once again we're deadset on writing the rules instead of building the thing. it's embarrassing.
I can't believe I hit the opus max (20x)5h limits..I was only running 15 concurrent agents for hours on end
Tom Nicholson retweeted
the amount of people thinking this is fake is hilarious. it's real. it's on ai.gov.uk. i found it because i'm weird and browse government ai pages
but it shows exactly what's wrong. we aren't taking ai seriously. the us and china are pouring billions into building it and we're writing guidance on environmentally friendly prompts
there is no reason we can't compete. none. it's complacency, being a stickler for rules, and the constant gaslighting that we're still superior
we're not. we have to earn it back by building, not by lecturing civil servants on how to not say thank you to a chatbot
the uk government's official ai guidance tells civil servants to use gemini flash instead of pro, keep prompts short, and not say thank you to it, for the environment
i do wonder at which point we forked the road. instead of building and using it to automate the work, we write guidance on how to use less of it
we have the attitude of a high school prefect. caring more about the rules than actually doing the work
