It’s time for the next chapter.

AI Seoul Summit | AI 서울 정상회의
🇬🇧 🇰🇷

Find out more:
gov.uk/government/new…

I'm hiring a technical specialist to be my right hand person in running the programme. Looking for expertise in neuroengineering/systems-neuro/bioengineering, across academia and/or industry.

If you are interested, or know someone who might be, please checkout the link below👇

Essentially, trying to reset information without spending energy is like trying to balance a ball on a hill, as in the figure below!
6/n

Our Nature Reviews Physics perspective article on neural operators and their ability to accelerate simulations and design is now out. rdcu.be/dD8BI nature

1. Neural operators learn mappings between functions, e.g. spatiotemporal processes and partial differential equations.…

OpenAI and Anthropic also have London offices. And a big chunk of Google DeepMind is there.
On the AI Safety side, there's also UK AISI, the Alignment team at Google DeepMind, Apollo Research and LISA.

PiSSA: Principal Singular Values and Singular Vectors Adaptation of Large Language Models

Significantly improved finetuned perf by simply changing the initialization of LoRA's AB matrix from Gaussian/zero to principal components of W

repo: github.com/GraphPKU/PiSSA
abs:…

I am once again asking frontend and dataviz engineers to apply to Epoch.

🚨Excited to share that our programme thesis is live!🚨
We lay out a vision for ***precision neurotech*** capable of tackling some of the most complex brain disorders at the circuit level. We believe this is possible by leveraging advances in bio- & hardware-engineering🧠🧬💻

I'm very grateful for Abram Demski for having written two important-yet-underdiscussed AI safety claims.

1) LLMs R&D abilities by default will favor programming-intensive tasks (& hence capabilities) over safety.

2) LLMs R&D abilities should be steered towards non-LLM AI…

davidad 🎇 Audiomulch had matrix routing as modules embedded within a topological graph. Worked incredibly well for re-routing patches as part of a live performance.

Still one of my favorite HCIs 25 years on.

Theoreticians who are at the frontier of understanding why algorithms work are sometimes disappointed when their theoretical framework recovers something that is “close but not quite”.

But this is the best outcome: then practitioners can benefit from the theory by matching it.