tom

on a task to demystify cognition. father, builder & ml engineer. into maths, cogsci & movement

[tom @tombukic]

Don’t ban Chinese open-weight models. Tariff the hosted ones and send that money back to Americans as AI credits to further support the industry. https://t.co/Bw0Z5iXrWv

[tom @tombukic]

@Anthony_Bonato Mathematics is art of exploring and communicating shared patterns of human minds. It's a navigation map of mentally comprehensible, as well as the process of building that map.

[tom @tombukic]

@max_paperclips @Aut4rk Fair.
Good luck in this stage; we all know you'll do it well 😎

[tom @tombukic]

@Anthony_Bonato Mathematics is art of exploring and communicating shared patterns of human minds. It's a navigation map of mentally comprehensible, as well as the process of building that map.

[tom @tombukic]

@Aut4rk @max_paperclips Tbh it would be nice to see a kind of explanation for founding engineer leaving.

I guess there might be some heavy hearts involved, but they're building in the open and all very cool guys - I'd expect that level of transparency even with such stuff

[tom @tombukic]

@max_paperclips @Aut4rk Parting ways with Nous? ☹️

[tom @tombukic]

@DanielSamanez3 OP is very confused https://t.co/FsDsQLqYAG

[tom @tombukic]

Very wrong . #1 is irrelevant.

I.e. humans are conscious but noons treats household as a sentient being.

Hard problem is about STATES and interaction of mental (nonphysical) and brain (physical) states which by the definition (physical vs non physical) shouldn't be interacting.

[tom @tombukic]

Very wrong . #1 is irrelevant.

I.e. humans are conscious but noons treats household as a sentient being.

Hard problem is about STATES and interaction of mental (nonphysical) and brain (physical) states which by the definition (physical vs non physical) shouldn't be interacting.

[tom @tombukic]

@thsottiaux So reset in 3 hrs?

[tom @tombukic]

@mdfnlshts Daj im pristup svim github repima

[tom @tombukic]

@SydSteyerhart @inductionheads You could have written this without "conscious" and I'd not along. But why tf did you need to add "conscious" into the story. Where it enters, how, why? What do we need it for running excellent self-improving agents?

[tom @tombukic]

I can say it with 100% confidence that "pain vector" has nothing to do with pains being felt by real beings.

They're of a different kind.
One is calculated change to the mathematical function, text-generation simulator, that biases it to generate text consistent with text humans generate when in pain.

Another is internal view from the being that has evolved across all levels of existence (from chemical, across biological, to cognitive) - like Matryoshka doll - to react in the way which will preserve its own physical existence across all those levels.

It is absurd to postulate text generation simulator is anything but text generation simulator. It has very nice properties and correlations showing we have been able to mathematically approximate various, but it is absolutely absurd comparing abstract matrix multiplication with a consequence of layered biological existence of a real being in a physical world.

One had to be systematically retarded into abstract thinking for their whole childhood to even consider equating between life and a text generator based on such surface similarities.

[Alec @AlecTPhD]

Yep! Claude Science has a python repl with a host bridge that gives it access to harness primitives so it can do lots of things programmatically. MCP tool calling, skill creation, agent profile creation, LLM calling, delegation, and more :) The python repl is also perfect for doing computational science since bulky datasets stay in memory as agents iterate on their analysis!

[tom @tombukic]

While I agree polluting pretraining data is a real problem for future models, as well as polluting discourse is for fellow humans, I am unsure "once machines get sentience" is either coherent picture of the future or important checkmark for anything even if possible.

Where and how is the particular LLM deployed is important our "find out" phase of polluting training data. Neither has anything to do with eventual model's or system sentience.

[tom @tombukic]

@DJLougen @MiaAI_lab @grok TLDR on the whole issue?

[tom @tombukic]

Bro suffers from multidimensional messiah complex.

[David Z. Morris @davidzmorris]

This is, unironically, delightful. They are their own worst enemies.

[Cosmin Negruseri @cosminnegruseri]

Great piece of writing, I love mostly the positive perspective at the end:

"The age of lone climbers is closing, and the age of mapping the whole continent is opening. As for me, I can't wait to see those vast mountain ranges from my supersonic airplane."

And I also love Christian sharing about his background related to math olympiads - feels related/familiar with a lot of eastern european educations.

[tom @tombukic]

The only thing I find myself jealous about Mac is a battery life. I can't hold my laptop on my lap how warm it gets if it is not ventilated enough; and if I work unplugged I can get 1hr of battery life from it.

I have no idea why would any self-respecting dev be eager to work in closed ecosystem which is continuously phoning home. Soo creepy.

[Andrew Torba @BasedTorba]

The EA cultists are so utterly repulsive to any normal person. I mean for one just from a physiognomy standpoint lol, but then they post the demonic insanity that floats around in the empty space between their ears and it makes it even worse.

[tom @tombukic]

You're just not thinking abstractly enough to notice this is very unaligned behavior 🤣

I am extremely frustrated when seemingly smart people abstract away all relevant details and reach ultracrazy conclusions. I mean, why bother yourself with such minutae like there was never a real human in the scene?!