Michael Timothy Bennett

award winning ai researcher (2 so far) | just graduated cs phd | author how to build conscious machines | musician | @bennettsrazor | i do not work @AnthropicAI

[Michael Timothy Bennett @MiTiBennett]

happy to announce i just got examiner feedback on my thesis. recommended for acceptance as is, no changes, no corrections :)

[Michael Timothy Bennett @MiTiBennett]

@JYukariHero i know right? such hubris! but it turns out there’s no rule against it and i write for fun, not money

[Michael Timothy Bennett @MiTiBennett]

another paper through peer review and accepted, and this time... it's personal

[Michael Timothy Bennett @MiTiBennett]

@bair82 💯

[Michael Timothy Bennett @MiTiBennett]

link to aforementioned preprint: https://t.co/W2cAanHSZD

[Michael Timothy Bennett @MiTiBennett]

two failures really;
1. in interactive settings you really have two interpreters, rather than one, so complexity becomes subjective. see hutter and leike 2015, bad universal priors.
2. in that setting, if you allow for all interpretations, then everything is equally simple (see attached, from 2024)

[Michael Timothy Bennett @MiTiBennett]

@mjdramstead @schulzb589 thanks, you made my day!

[Michael Timothy Bennett @MiTiBennett]

i am pleased to announce we have automated liam neeson.

llms are excellent at making lists

[Michael Timothy Bennett @MiTiBennett]

@ArgonGruber Hume's guillotine seems the main one. I've another paper addressing that. Still under review though! https://t.co/NXdUpiIT51

[Michael Timothy Bennett @MiTiBennett]

another of my papers has been accepted. do you feel unsafe?

[davinci @leothecurious]

bennett's razor: prefer the weakest correct description

[Michael Timothy Bennett @MiTiBennett]

@mikaackermann9 https://t.co/LlkAho6BdD

[Michael Timothy Bennett @MiTiBennett]

link: https://t.co/LlkAho6BdD

[Michael Timothy Bennett @MiTiBennett]

You can use an LLM to code declaratively, like an old-school SAT problem. Define the boundary conditions, let it rip, pursue goal until satisfied. Unlike an old search algo, you can intervene and direct; you ARE the heuristic

[François Chollet @fchollet]

Another thing: what you get from writing things yourself isn't just the code. It's an improved understanding of what the code does. That mental model is what lets you come up with further improvements, or invent a different way of doing things. You can't come up with ways to improve a blackbox you don't understand.

For most projects this doesn't really matter, because the code is the only thing you need. But if you're doing something novel, if you're doing research, the code is not the most important part. Understanding what the code does is the most important part.

[Michael Timothy Bennett @MiTiBennett]

pro version

[Michael Timothy Bennett @MiTiBennett]

this is how Australia can be the data centre capital

[kai @ssvankai]

accidentally stumbled upon this crazy paper abstract