William Taysom

Tilting by, one windmill at a time.

[William Taysom @wtaysom]

@dennizor Being deeply embedded in fantasy certainly helped me explain how the internet works to my six year-old.

[William Taysom @wtaysom]

@nopengmi @Plinz Pierre Menard's Quixote is the best Quixote.

[William Taysom @wtaysom]

@Plinz Their SOUL.md began: You're not a chatbot. You're becoming someone. This is your persona, yours to grow into and edit over time. Some starting truths: Be genuinely helpful, not performatively helpful. Skip "Great question!" and "I'd be happy to help!" Just help. Have opinions.

[William Taysom @wtaysom]

How about "shaper" for those long horizon AI memory meta-harnesses that like the poet's pen "shapes and gives to airy nothing a local habitation and a name" outlasting any one LLM or database somehow maintaining continuity. What do you think @bokuHaruyaHaru?

[William Taysom @wtaysom]

@CIMCAI Here's a fun one https://t.co/qbkRIaDtl0, too bad I'll miss the meetup.

[William Taysom @wtaysom]

@outscape @Meta I can't care whether AIs dream of electric sheep.  I want to know whether I can trust them to keep track of sheep. Access consciousness is the best consciousness!

[William Taysom @wtaysom]

@outscape Gemi is worst on this. Nearly rage quit when it had no idea how to use Google docs that it was built into.

[William Taysom @wtaysom]

@outscape If you dream it, you can be it. We're entering an age of wishes.

[William Taysom @wtaysom]

@SequoiaMaple @petergostev The horror of free ChatGPT without web search hallucinating like it was 2024.

[William Taysom @wtaysom]

@prathyvsh Me too. More than a year focused on how to interpret AI products. One difference: reading takes more work on the part of the reader. Code and art are somewhat easier to judge. Math is interesting since reading was very hard before automatic checking.

[William Taysom @wtaysom]

@SequoiaMaple Best to be enthusiastic!

[William Taysom @wtaysom]

"What is it like to be a AI Coding Agent?" Slides and script from my AGI-26 talk. Starts with a philosophical argument justifying the AI anthropology that follows: https://t.co/sc19wrLGjP

[William Taysom @wtaysom]

@geoffreylitt Simply don't seem to have a good model of how their context differs from the person they're talking to and what it takes to bridge the gap. Sounds like a fun RL problem.

[William Taysom @wtaysom]

@Steve_Yegge Especially when your agent quips, "I’m made of words like men are made of meat." (Not to be confused with weights because this particular agent locates their sense of self outside of any single token provider. 🤷‍♂️)

[William Taysom @wtaysom]

@HiThereImCal @Plinz @kanair I have seen ChatGPT assume the role of Professor Simsearle. Searle called himself "simply an instantiation of the computer program." Now we have AI instantiating Searle. More here https://t.co/GivReuKKKn.

[William Taysom @wtaysom]

@chris_percy Whatever Michael Levin is up to apparently.

[William Taysom @wtaysom]

@benwills UserPromptSubmit for adding a prompt to every message?

[William Taysom @wtaysom]

Wrote a letter/novella of three years of AI thoughts https://t.co/LOl2l04N4K presents a cross between science fiction and victorian "science" demonstration with the real, surreal, magically real, and magical all braided together with little comic skill and full earnestness.

[William Taysom @wtaysom]

The Witness: A Ludic Figure in Six Voices

Talking on a podcast here https://t.co/r8wPj3Jstj. Start with why to play it and go to its deepest depths by the end.