It's weird that both OpenAI and Anthropic are now insisting that they try hard not to displace jobs when that's obviously false, and also is not something that is normally expected of companies ethically speaking.
You can warn people, shape policy etc. without overclaiming...
I haven't read the full post, but... I see a discrepancy? "AGI capable of doing 90% of knowledge work" seems drastically different from the Bostromiam superintelligence he describes here — explicitly mentioning a gap between the two. So the timelines for the "90%-of-knowledge-work-AGI" surely are shorter?
Claude Fable 5 doesn’t truly understand. And here is a beautiful proof:
The Beninatto-Trombetti test is a translation test for professional translators. It measures the ability to infer context, revise the surface form, and generalize beyond literal mapping.
For example, the correct translation of:
“Solo 3 parole: non sei solo”
is not:
“Just 3 words: you are not alone”
but:
“Just 4 words: you are not alone.”
An LLM that understands the sentence must also update the meta-linguistic claim inside the sentence.
Claude Fable 5 is arguably the most advanced LLM currently available. And yet it still fails this simple test.
LLMs are extraordinary machines for recombining existing knowledge. But they don’t truly understand.
We are still far from AGI.
I wish I could describe how bad things are over here; the saddest thing is that it all compounds into a (seemingly) downward spiral.
• We have some talent, but lack the infra, the companies, and the capital.
• People (especially young people) have an increasingly hostile attitude toward AI
• The median person's understanding of AI is abysmal
• Worse, the median institution's understanding of AI is abysmal
• (Even worse, the median institution's focus is on pointless, ignorant crusades)
• Journalists' (or people who have a voice and should know better, in general) understanding of AI is abysmal
• Trying to educate people on this (my job) mostly falls of deaf ears
So yeah, it's bleak. Europe 2031 captures it pretty well.
Claude Fable 5 doesn’t truly understand. And here is a beautiful proof:
The Beninatto-Trombetti test is a translation test for professional translators. It measures the ability to infer context, revise the surface form, and generalize beyond literal mapping.
For example, the correct translation of:
“Solo 3 parole: non sei solo”
is not:
“Just 3 words: you are not alone”
but:
“Just 4 words: you are not alone.”
An LLM that understands the sentence must also update the meta-linguistic claim inside the sentence.
Claude Fable 5 is arguably the most advanced LLM currently available. And yet it still fails this simple test.
LLMs are extraordinary machines for recombining existing knowledge. But they don’t truly understand.
We are still far from AGI.
Claude 5 Fable (Ultracode)
I asked it to build a demo of my dream game in Three.js and I'm genuinely shocked 💀
One shot, a full explorable starship with a working cockpit, crew quarters, a planet drifting past real windows, dynamic lighting, sleep/eat interactions,
it screenshotted its own work and fixed itself until it hit 60fps on browser
Obviously not steam ready but man this is so so far from what we had one year ago…
oh so you're telling me the company pursuing artificial general intelligence is releasing increasingly general models capable of tackling the domains in the hands of specialized firms instead of "enhancing", "collaborating with", or "copiloting alongside" them? shocking
Wondering why Figma CEO Dylan Field joked at an event that Anthropic was “not consistently candid in their communications" when it came to Claude Design?
Anthropic is wielding its newfound power over partners and customers, they say.
W/ @amir
https://t.co/MNG30Wjhld
The longer the timelines, the stronger the chance for Google to eventually come out on top. The opposite is also true, however; and, as of right now, seemingly more likely.
TacticAI is kinda sick tho.
We’re teaming up @Palmeiras, the first football club to meaningfully build upon TacticAI: our AI system that can help simulate field scenarios and predict open play dynamics up to 8 seconds in advance. ⚽
I think you just laid the entire rationale before both of our eyes. Taking the foot off the gas when you have a substantial lead? In capitalism-driven market dynamics? That's just not going to happen... unless.
Unless there's something truly fucking scary over the horizon, which may very well upend precisely the capitalism-driven market (or the global socio-economic order, in better terms) hereinabove.
I think that having Mythos now and sufficient foresight into the future of AI capabilities is what grants all these undoubtedly tedious and short-term financially-painful moves. Realizing that we really truly are in the same rickety boat.
Even China! Which I actually believe to be sufficiently sane, principled, and smart not to see this as a dumb move made by democracy-pilled westerners, but people who are actually concerned by an externality that might care very little about geography.
That's perhaps the one good lesson we can learn from the horrible Cold War season: boards don't get wiped out only in the parts that are convenient to one side, so perhaps it's important to actually play by the rules. Or, in this case, to set some.
We’re launching Claude Corps, a national fellowship program matching people early in their careers with US nonprofits.
We'll teach 1,000 people to use Claude, and pay them to use AI to advance their hosts’ missions.
https://t.co/QI6JmlAdSr
I'm deeply concerned about Europe's future on AI. One of my biggest worries is our erosion of agency, our ability to stay relevant and fight for our values in a future where AI becomes a civilisationally important technology.
Myself, @DadaJudith , @bakkermichiel and others have written a scenario to outline a potential future we worry we are on track towards.
https://t.co/3BRfitBT0n
Every optimistic and realistic path I can see for Europe runs through a central node - one where Europe has more leverage, more importance and more say. One where Europe grows more, builds more where it matters, and takes ownership over its resilience.
Europe 2031 is a five-year scenario of the continent's slide into irrelevance: how AI is driving it, and what can still be done.
The co-authors are researchers, scientists and investors who have advised European leaders, co-authored national AI strategies, built and funded these systems from the inside. We have no interest in hype and we deeply care about this continent.
Europe 2031 ends with five concrete recommendations:
- drastically more compute on European soil
- an AI middle-power coalition
- labour-market reforms
- a bold position in robotics and industrial AI
- and a positive vision of what AI can do for society.
Europe can still change course if it finds the political will and the courage to engage in the most ambitious political and economic agenda the continent has undertaken in peacetime.
I encourage you to read it if you have the time:
Thank you for the thorough response Mark. I don't think we're saying opposing things!
I too believe that the benevolent and voluntary cooperation scenario is pure fantasy. But that's why I commend the "let a [governmental or gov-regulated/owned] third party be an arbiter" approach.
I think AI has the potential to be *even worse* than nuclear weapons, both in terms of how slippery it is and how hard proliferation is to control.
I think Project Glasswing, for all its imperfections, was conceptually the right move. Mythos *is* a weapon. And since it'll only become more powerful, we now need to take one step further — indeed making sure that below a certain threshold competition remains unharmed.
But let's be honest: beyond Meta and Microsoft (and maybe xAI?) nobody's gonna catch up. So keeping OpenAI, DeepMind, and Anthropic itself in check is something I can't but back up.