Worth reading
I said "auto-regressive LLMs, in and of themselves, will not lead human-level AI"
That statement is still totally true.
First, the reasoning abilities of current AI systems are based non-auto-regressive search (which is what I have always advocated for). But AFAICT, they do it in token space, which is limited and inefficient. I have claimed that human-like reasoning must be a search in continuous representation space. It looks like the industry is moving towards that.
Second, the self-improvement methods, as currently practiced, only work for domains where the quality of outputs can be scored without human intervention, such as mathematics, code, and scenarios that can be simulated accurately. Not anything else. Humans and animals learn new skills way more efficiently than current RL methods.
Third, the multimodal capabilities of current AI assistants generally use separately-trained encoders (that are not LLMs). This is also what I've been advocating. Except that I think the best way to do this is with JEPA trained with self-supervised learning. The research community is clearly moving towards that (3000 papers on JEPA in just 4 years).
Fourth, if LLMs were a path to human-level AI, we would have domestic robots and Level-4 or Level-5 self-driving cars for consumers by now. And we don't. We certainly don't have cars that can learn to drive in 20 hours or practice like any teenager. We're still missing something pretty huge to claim human-level intelligence (let alone superhuman).
Sure, we now have computer systems that are impressive, very useful, and whose performance is superhuman in an increasing number of domains (coding being one of them).
But that's true of the entire history of progress in computer technology.
Lastly, there is a basic confusion about what intelligence actually is.
It is not the mere accumulation and regurgitation of existing declarative knowledge (which is essentially what LLMs do).
As Jean Piaget famously said, "intelligence is not what you know, it is what you do when you don't know."
It is your ability to solve new problem without any prior training, to act in previously-unknown scenarios, and to adapt very quickly to new situations with minimal training.
We're still far from that.
19
Fuck around, find out.
Companies like Nestlé and Auchan chose to stay in russia and keep funding Putin's genocidal war machine invading Europe.
Yesterday he signed a decree confiscating their assets.
But hey at least they got Tucker Carlson his loaf of bread. Hope it was worth it @AUCHAN_France
5
"IF" (with a big pinch of salt) it's true... actually no news in Italy about the 147GB of government data stolen.
@REPUBBLlCA_lT @sole24ore @Corriere @reportrai3
‼️ BREAKING: We're in contact with the Revolut hacker. According to them, they didn't only take Revolut data, they've also compromised multiple Italian law enforcement departments.
They say the operation targeting Revolut ran for six months, and that they used Italian law enforcement systems to send data requests to Revolut.
They claim to hold 147 GB taken from the Italian side, including internal docs, calendars and personal material, among it the chat logs of a federal officer arguing with his wife.
32
AMD aquires Taalas.
What Taalas does? Well... destroys OpenAi, Anthropic and many other business model and investments on datacenter they already did.
Not to mention NVidia.
1
45
Andrea Consorti 🇮🇹 🇺🇦 retweeted
mes tweets un peu techniques sont pas super evidents à comprendre pour tout le monde alors je vais tenter de vulgariser au mieux pour expliquer toute la comedie qui se joue autour de l’IA en ce moment
sachez déjà que nous assistons probablement à la + grande manipulation tech de notre epoque car quand je vois les dirigeants des plus grands labos de la silicon valley reclamer soudainement une pause et demander des exemptions aux lois antitrust au nom de la securitée globale je pense que nous devons regarder exactement la où ils veulent detourner notre attention
mettez vous en tête que ce spectacle pseudo ethique masque une panique financière tres concrete face aux miliards engloutis dans des supercalculateurs dont les rendements sessoufflent combinee à la terreur dexposer ces pertes dans leurs prochaines formulaires d’IPO (entrée en bourse)
puis je trouve que la methode brille par son cynisme sils estiment sincerement que leurs modeles non publies representent un danger ce duopole (Anthropic /openAI) possede deja toute la latitude pour freiner ses propres deploiements sans demander la permission a quiconque au lieu de cela ils nomment leurs propres organismes dévaluation soi disant independants comme METR en y placant leurs anciens employes et investisseurs & réclament une exemption antitrust officielle pour s’entendre sur le rythme de l’innovation en cherchent à substituer un tampon bureaucratique à leur responsabilite civile réelle
comprenez que leur demarche vise avant tout a imposer leurs propres controleurs à des concurrents qui ne sont meme pas à la frontiere technologique…
perso je pense que le calcul geopolitique qui se cache derriere ce recit s’avère tout aussi bancal car prétendre limiter le progrès aux seules nations alignées sous couvert de coordination entre democraties pour maintenir une avance artificielle relève de l’illusion pure
dites vous que pendant que la silicon valley redige des chartes morales pour preserver ses rentes, verrouiller le marche americain et criminaliser la concurence exterieure, les laboratoires chinois optimisent leurs algorithmes tirent parti dune energie abordable et distribuent la puissance de calcul a une fraction du cout occidental, les batisseurs a lautre bout du monde avancent et construisent pendant que les geants americains debattent dun cartel sur mesure
de même je pense que l’erreur fondamentale de ces acteurs reside dans la confusion entre la fin de leur monopole et la fin de la civilisation
sachez que de l’invention de l’imprimerie au deploiement de la cryptographie moderne les pouvoirs centraux ont toujours qualifie la democratisation des outils de la pensée de menace pour lordre public
le problème pour eux cnest que l’intelligence artificielle depasse la notion d’actif prive enfermé dans des coffres forts d’entreprise, elle simpose comme une propriete fondamentale de la matière et du calcul & vouloir rationner le code par decret administratif revient à tenter de breveter les lois de la physique
je pense que l’avenir appartiendra inévitablement aux protocoles ouverts au calcul souverain et à l’inférence locale car la veritable resilience d’une societe emerge directement dun systeme immunitaire décentralisé ou des millions dutilisateurs inspectent, eprouvent et ameliorent la technologie en temps reel loin de la stérilité des laboratoires ultra centralisés
les institutions qui tentent deriger des murs autour du savoir finissent par isoler leur propre fragilite pendant que le reste du monde batit le siecle sur des fondations libres
127
672
2,640
184,102
Worth reading.
Could rogue agent swarms take over the entire internet in the next six months? garymarcus.substack.com/p/co…
1
13
Andrea Consorti 🇮🇹 🇺🇦 retweeted
Tell the truth: You've run out of places to get money from and your model is still operating at a loss. What you're trying to do is just an OPEC-style cartel. Even after creating artificial scarcity around hardware and even electricity, you keep operating in the red. Every time you think you're going to dominate the market, a Chinese model is released doing more for less, or even running on someone's home CPU. So you need the magic card of regulation, which obviously favors your business. Maybe that works with the boomers still in power, and some hysterics from other generations, but it's forced.
2
12
1,060
Tell the investors you need to slow down.
Tell China they need to act on AI regulation.
Tell useful idiots how AI could kill humanity.
The AI race stops when stakeholders see no returns next year.
You can't slow down the bubble popping. U 'r fucked
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
35
We had the Murphy theory brutally replaced by the FAFO theory.
Revolut did not get hacked.
It got an email that looked like the state. Real domain. Valid credentials. They sent the files.
Passports. Selfies. Home addresses. IBANs. Full statements. Every Bitcoin move they logged.
That is the KYC tax. You give the bank a copy of your life so they can stay legal. Then one fake request walks out with it.
The app is fine. Your identity is not.
11
Andrea Consorti 🇮🇹 🇺🇦 retweeted
if you are a researcher, engineer or deeptech founder, never trust a closed model with your unpublished alpha
prompting proprietary AI with confidential breakthroughs turns your lab into an uncredited training hub for a cloud monopoly
keep your sensitive research off their servers
34
128
798
39,808
Worth reading
go fuck yourself @sama
claiming that you solved navier stokes because 10000 agents ran in circles for 88hours on a multimillion dollar gpu cluster to formalize in lean a blowup case under controlled external forcing is pure scientific vulgarity the clay mathematics institute millennium prize does not ask m whether you can artificially force a singularity in a fluid by injecting an ad hoc smooth external forcing term f(x,t) to twist the vortex until it breaks the real problem questions the fundamental stability and global smooth existence for 3dimensional incompressible euler & navier stokes equations under natural conservation laws and viscous dissipation alone using a mathematical loophole on forced equations to parade a century old victory is a major conceptual scam Altman
technically & epistemologically what you present as an agi breakthrough is nothing more than bruteforce combinatorial autoformalization the ai did not understand fluid mechanics it simply navigated a continuous search space previously mapped out and constrained by the monumental work of human mathematicians like tristan buckmaster/ levent alpöge / diego córdoba or tarek elgindi coordinating 10000 agents to check the logical consistency of a 100 page proof via lean is a software engineering feat and computational parallelization triumph not an intrinsic scientific discovery it is the victory of the compute bulldozer over abstract human intuition repackaged for the public as a higher mathematical consciousness
to this theoretical imposture you add a disgusting ethical and industrial cynicism taking advantage of private codex sessions and informal preprints from academic researchers to siphon their research leads and then trying to redact or erase the contribution of levent alpöge under the pretext that he works at rival anthropic is intellectual serfdom openai behaves like a feudal lord of silicon appropriating the cognitive subsistence of independent scholars threatening their careers behind closed doors if they protest and turning community academic labor into a privatized pressrelease
this entire staged event serves a desperate financial agenda in a pre ipo panic facing the slowdown of scaling laws and growing investor skepticism over the profitability of foundational models openai needs to manufacture an artificial sputnik moment claiming to solve a millennium prize without immediately submitting the proof to traditional peer review means using the prestige of fundamental mathematics as cheap marketing fuel to inflate a delusional valuation!!!
real science is not a clout chase on social media or a compute spike spent to rob the clay mathematics institute it is a quest for elegance physical truth and universal rigor to decode reality true artificial intelligence will not emerge from hostile corporate takeover of academic work hidden behind computational bruteforce but from architectures capable of generating new conceptual paradigms by masquerading constrained formalization as the collapse of physics greatest mysteries you did not solve navier stokes you only proved how far silicon valley will go to prostitute scientific integrity for capitalist spectacle
6
When this F bubble will explode?
>math professor spends a year on one of the hardest unsolved problems in math
>"Navier-Stokes"
>his drafts for the whole project went through codex sessions
>openai had his logs in codex
>suddenly rumor leaks AI solved it
>he emails openai to ask what's going on
>openai calls within days
>"funny story, our model also solved it"
>using his exact approach
>asks openai: did you train on my sessions?
>no answer
8