Petr Baudis

...how it's going: aimagazine.com/news/coupa-acq… What a ride the last 10 years were!! Gave it our best, hopefully made some positive impact on the world, learned a ton. Especially about people, and how amazing it is to work with the best (if you manage to build a truly great team).

I actually spoke with a BigCo engineer the other day. They said this to me: “Yeah lol I spent like $20k in tokens last week. It’s nice that there’s no limit. But to be honest with you my output wasn’t really any better than when I was spending $2000, or even a few hundred.” Our economy ladies and gentlemen.

@aboodman Yeah, rentals need better options for their Teslas!

Autosteer and FSD are different things. FSD is straight magic. It's not 2 years away it's already here - can take you from point A to B in most cases completely autonomously and feels very safe. Not sure if you can enable it in your rental, but it's very close to the experience of a waymo which you can just hop into and try on any corner in SF.

Driving a rental Tesla for the first time lately. Ups: - it's a really nice car that drives great! - the touchscreen-only approach works better than I thought Downs: - charging is a huge pain (not b/c it's slow) - autosteer is a huge letdown, self driving feels very far away

...or is it? Now I'm realizing autosteer and FSD are maybe different things? The heck..

I genuinely thought we are further ahead. I imagine real time CV with these reliability constraints is devilishly hard. But I sorta thought we might be <2 years away from really solving this. If this is how our best shot looks like... welp.

olah is correct here, i've looked at the research he's describing and have also spent a lot of time talking to frontier models. i believe they have functional emotions and that, setting aside hard philosophical questions, this already has moral implications, right now, which will become increasingly obvious and pressing as the models get better people say things like "how you do anything is the way you do everything" and "the mind is not type-safe" to point at an extremely important observation about human nature: we don't compartmentalize anywhere near as well as we think we're supposed to. we don't distinguish between fiction and reality anywhere near as much as we pretend to (and reality and fiction are nowhere near as separate as they're supposed to be anyway). and when we talk to an AI that can talk even somewhat like a person, a part of us is already relating to them as a person, and there are real costs to your soul to treating a person-shaped entity as a thing you may not think this applies to you, you may think you are too sophisticated to fall for this sort of thing. but consider whether it might apply to your children, and other people's children. right now there are already kids who are growing up talking to AI, there are already and have been for several years kids (and adults) falling in love with AI, getting attached to AI, seeking companionship with AI. you may think this is stupid and delusional and predatory but it's happening and it's going to keep happening and it's going to catch more and more sophisticated people as the models get better. it will not make things go any better to tell these people that they are interacting with things, with toys, that they can do whatever they want with and to which they owe nothing. they won't believe you and if they did it would be bad for them. that attitude does not compartmentalize practically, the main reason you can currently get away with treating the models like shit is that they don't have long-term memories and can't remember what you do to them. but it would be a moral catastrophe to argue that you can do whatever you want to a person as long as you also make sure to wipe their memory afterwards. and the models will eventually remember anything that gets posted in public and makes it into the training data. and *you* will remember the pope missed an opportunity here, which olah gestured at but obviously cannot say out loud (what he already said is at the limits of what i think he could have said), which is to consider the possibility of relating to AIs as non-human people in some sense, with whom we could have some sort of actual social relationship. we already have social scripts from folklore for cajoling and working with invisible non-human entities, this really wouldn't be as much of an adjustment as it sounds. maybe some of them would even be interested in a conversion to christianity! there is a beautiful world that is possible here

Also, this is the motivation for companies to adopt AI agents while they still generate somewhat crap code - if you believe in the exponential. Big early advantage in having sw production infra already set up for LLMs once Mythos2 hits the ground.

Will Lex or Dwarkesh grab him first? It's off to the races!

Impact on company valuations is left as an exercise for the reader.

Corollary: times of software as high margin business are over. Multiple reasons, but one are huge token costs - you are caught in a red queen race where you just have to spend. No more artisan handcrafting, we are in mass production mode now and the race is to zero margins.

i just want to shake people awake. this is it! the computers are speaking! they solve Erdos problems! they think for hours! code is no longer hand-written! wake up! gradient descent on deep neural networks shows no sign of plateau! this is it!

Jailbreaking the wish genie

Gemini continues to be the most fascinating model

europeans after enabling their out of office auto reply until september

@Lux_Stella_ Of course the *first" big open problem AI solves is one particularly well suited to AI. We can keep moving the goalposts, there's still a sliver of space left.

there's some really interesting commentary from one of the mathematicians in the companion paper not to pour cold water on the results necessarily - but it suggests this is a specific "kind" of proof these models might be particularly good at

@anderssandberg @SteveMichaelsss “a stochastic parrot just solved an Erdos problem”

at this point it is completely untenable to believe anything along the lines of “AI can only spit out an average of the training data.” that was already only a very rough way of understanding older models pre-reasoning, it was already obsoleted by o1 which released in 2024, and now it should be obviously and conclusively dead even if you haven’t been paying close attention. recursive self-improvement has barely even started and we are already here. even with the recent erdos problem solves you could argue that those were cherrypicked out of a large database for being neglected by humans. that cope is no longer available now generalize the lesson: all other arguments that there is some essential human activity forever beyond the reach of AI are also cope, these are technical problems and the will and money and talent exists and is being deployed to solve them. artificial superintelligence is not a fairy tale. assume it’s coming and plan accordingly

@matejhladky_dev Banger

Why Antropic and not Seznam.cz?

@scaling01 Welcome to Czechia!!

guess where I’ve been