Quto

X is an attention-optimised platform: you will be exposed to propaganda, cherry-picked anecdotes, miscaptioned media and engagement bait.

[Quto @Qutossar]

@paradite_ It is the point of training, but not achieved completely. Just like with humans.

[Quto @Qutossar]

@ianbremmer Maybe they should vote for people who seem to have moral compass.

[Quto @Qutossar]

@haider1 Considering the main problem of transformers is generalization, which leads to the sharp jagged peaks, I'm sure they are very helpful.
But I doubt they will develop ASI on their own or even be a huge help with this kind of innovation. But I may be wrong.

[Quto @Qutossar]

@ischinger @jaegerthomas2 Europa ist relevant, weil es die Ukraine unterstützt. Wenn wer nicht mit uns reden will, dann ist das deren Problem.
Wir sollten weniger darüber nachdenken, wie wir erreichen, dass jemand mit uns redet, und mehr darüber wie wir noch relevanter werden können. Rest ergibt sich.

[Quto @Qutossar]

@JulianRoepcke He may be right. The correct way to think about this, however, is not to surrender against nuclear powers, but to accept, that the price of freedom can be nuclear retaliation.
In fact, if we don't accept this, we icrease the effect of a nuclear strike and thus incentivise it.

[Quto @Qutossar]

@matvelloso It does understand that particular question really well, actually.

[Quto @Qutossar]

@gbgreecearis @rohanpaul_ai @ylecun @grok: How does one open a door ?

[Quto @Qutossar]

@rohanpaul_ai @ylecun It is true that models cannot jump or run or evade.
But they can write great text. And anything that can be expressed in text can be learned by them. Did great CEOs or scientists learn their job by anything other than spoken/written text?
Text is not everything. But it is a lot!

[Quto @Qutossar]

@deanwball Show them how AGI is done properly! 😉

[Quto @Qutossar]

@synthwavedd Don't see where the 'bending' is in this headline

[Quto @Qutossar]

@ChrissGPT I do trust that Elon has good chance to get something difficult going - much higher chance than most other people. Let's say about 30%.
But I never believe any of his timelines.

[Quto @Qutossar]

@_simonsmith All the saved compute now goes to model improvement .. Dario pointed this out himself, in a different context, some time ago.

[Quto @Qutossar]

@rahul It is not about Fable It is about MAGA doesn't like Anthropic.
It's been years now and people think MAGA means nothing changes.

[Quto @Qutossar]

@robinebers He is willing to get dirty to ensure the best result for humanity that he can. I respect that a lot.

[Quto @Qutossar]

@McFaul You think the US society will elect such presidents in the future ?

[Quto @Qutossar]

@ianbremmer @SpencerHakimian well, you didn't give them 300 so they stop attacking your ships...

[Quto @Qutossar]

@andre_quentin agreed.
They are also terrible at *applied* theory of mind.
And they constantly want to write code to test whether the respnse from an agent does comply with some static demand. Usually through checking keywords ...

[Quto @Qutossar]

@jurgen_nauditt @bundeskanzler @POTUS Mal ernsthaft ... it's really not all that bad.
It's just a T-Shirt. People call this diplomacy. It costs Germany nothing.

[Quto @Qutossar]

@robertwiblin Believe it or not: Building a model at 90% top capability from scratch is actually not all that hard, given the money.
This isn't a race.

[Quto @Qutossar]

@petergyang People just don't believe MAGA incompetence. It is not just Iran, guys!
No administration can withstand such an executive for years. Eventually incompetence and corruption do trickle down ...