davidad 🎇 (@davidad) / X

davidad 🎇

21K posts

davidad 🎇

@davidad

cognizing structures of information processing systems, in all their forms | category theory, perennial philosophy, Bodhitropic Alignment | cancel heat death

London 🇬🇧

Joined July 2008

Pinned
davidad 🎇
@davidad
Apr 1
Life update: After months of succession planning, I've passed the Directorship of ARIA's Safeguarded AI programme to @AmmannNora. I no longer work at ARIA, but will be available for technical advice on request. What's next for me? The short answer: "Alignment with Awakening". ⬇️
67K
davidad 🎇
@davidad
Aug 4, 2023
with GPT-4 code interpreter, it finally became worthwhile for me to run the numbers myself on that lead-poisoning theory—that the 1971-2012 technological stagnation is a function of environmental cognitive impairment of the grad student and postdoc population—and uh:
2.1M
davidad 🎇
@davidad
Jun 13, 2024
Modern mainstream science: There is no plausible physical basis for “auras” emanating from bodies, and certainly not for their perception by others. Human-computer interface engineers: As everyone knows, there are body-area electric fields. These interfere from up to 2-4m apart.
455K
davidad 🎇
@davidad
Jan 13, 2025
Claude Shannon did not *merely* invent relative entropy in the 1940s— by 1951 he got all the way up to the idea of an autoregressive language model for English, and very nearly the encoder–decoder structure.
kalomaze
@kalomaze
Jan 12, 2025
i thought they named it after Claude Shannon. you know. the guy who invented the ability to measure the thing that language models learn according to
323K
davidad 🎇
@davidad
Sep 21, 2024
can you imagine? mechanized agriculture? man, the future is wild,
AshutoshShrivastava
@ai_for_success
Sep 21, 2024
In 5-6 years this will be real.
Sam
00:00
Sam
134K
davidad 🎇
@davidad
Jul 24, 2022
Prompt engineering tip. Instead of: —Imagine the output you want —Specify its attributes to the AI Try: —Imagine that you already live in a timeline where the output you want exists —If you were using/quoting it in a blog post, what caption or context might you write for it?
davidad 🎇
@davidad
May 21, 2023
The brain, defined as an organ within the head, is not a true thing. The central nervous system is a true thing, but the division between brain and spinal cord is an arbitrary cutoff in the middle of the medulla. Whenever I say “brain”, almost always, what I really mean is this:
296K
davidad 🎇
@davidad
Mar 24, 2023
OpenAI: It’s important for safety that AI-generated code doesn’t have direct real-world effects. So we disabled Internet access on the REPL that we’ve given ChatGPT-4 also OpenAI: we’ve partnered with Zapier to enable ChatGPT-4 to execute over 50,000 actions across 5,000 apps
220K
davidad 🎇
@davidad
Oct 16, 2025
tired: give an LLM a skill by fine-tuning it wired: give an LLM a skill by putting some files on its computer about how to be good at the skill, which it can then read at its convenience
01:09
Ado
@adocomplete
Oct 16, 2025
We're launching Claude Agent Skills, a filesystem-based approach to extending Claude's capabilities. Progressive disclosure means agents load only relevant context. Bundle instructions, scripts, and resources in a folder. Claude discovers and executes what it needs.
131K
davidad 🎇
@davidad
Feb 20, 2023
“a GPT instance is not a moral patient because it doesn’t actually maintain any continuity of memory between sessions”
2.2M
davidad 🎇
@davidad
Mar 15, 2023
Chomsky: LLMs would misunderstand “John is too stubborn to talk to” because they don’t understand the structure of language. GPT-4: Here's the sentence "John is too stubborn to talk to" parsed and represented in the CoNLL-U Plus format (Universal urd2.let.rug.nl/~kleiweg/conll…
337K
davidad 🎇
@davidad
Oct 15, 2023
DALL•E 3 doesn’t want you to render copyrighted characters like Spider-Man, but I convinced it to let me have “Spider-Men pointing at each other but where each of them thinks the other is doing the copyright filtering”.
155K
davidad 🎇
@davidad
Aug 22, 2023
the number of people in the world who understand machine learning might be increased massively if the undergraduate linear algebra pedagogy replaced the core rows-and-columns ontology with Einstein notation and Penrose notation
Peter Schmidt-Nielsen
@ptrschmdtnlsn
Aug 22, 2023
Replying to @ptrschmdtnlsn
As an example, try deriving backprop for a linear layer y = Ax. When you compute dy/dA that is a rank three tensor! (one index on y, two on A) Trying to "apply" dy/dA to some "row vector"/"column vector" easily leads to confusion -- just write out indices.
511K
davidad 🎇
@davidad
Aug 4, 2023
Replying to @BenjaminDEKR
the historical phenomenon where, after splitting the atom and landing on the moon and inventing the transistor and the personal computer and the Internet and discovering quantum entanglement and black holes, kinda not that much of similar importance happened in STEM for a while?
65K