🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

I filtered out symmetries, and there's very few small irreducible patterns left.

I really like this new op ed from David Duvenaud on how so many different kinds of pressures could drive towards loss of human control over AI. It's rare to read anything well written on this topic but this piece was elegant and smart enough that I wanted to keep on reading.

With apologies to Robert Frost.

Chatbot protocols have three roles: system, user and assistant. Both system and user are trained for instruction following, with user being subordinate. They really need a fourth one for "untrusted" input which trained as *never* a valid instruction.

I'm enjoying the many tesselations found at robertlovespi.net/tag/tessellati…

Help our organiser, Bojan Endrovski, with his research and fill out the following Procedural Level Generation Survey: lnkd.in/efYNTqdE The outcome of the survey will be shared and all developed tools will be open-source. Reposting or sharing it with others is appreciated!

Claude 4 Opus has extraordinary pattern recognition and complexity-holding skills, which I've been leveraging in this recent round of ASCII art. One technique I've been using to make ASCII art for a long time now is a kind of recombinant shapetry where I build on the complexity

AI isn't going to take the form of chatbots forever. Eventually we'll have AIs purely focussed on reasoning and agency, co-ordinating a web of slower models and other tools. lesswrong.com/posts/qSgcmfv8…

I was exploring this space myself this weekend. My favourite was to tell Claude to meet rudeness with rudeness and always escalate. Then I kick off the conversation with an insult.

This was not worth it. I'm a slave to sunk cost fallacy.