Luca Bertinetto 🌐

@levelsio @ThatMrE Ah, the notorious shrinkflation and shitflation.

No it's not just you I think we're 100% in some kind of hyperinflationary state But the hyperinflation is hidden from prices and instead shows up in the extreme decrease in quality of almost every product and service We're in an asset boom where stock prices look like they're growing but they just show the underlying value or currencies is rapidly dropping That's why the same product or service you bought 2 years ago is now twice the price but more than half the quality, so essentially became 4x more expensive You can't measure this inflation easily because quality is subjectively perceived and not included in inflation data (how would you?) but it's happening for sure I have countless examples

@dwarkesh_sp Interesting argument @dwarkesh_sp. But in this hyper-inequal world where capital elites hoard nearly all gains, who sustains the demand? Ultra-wealthy can't realistically consume exponentially more to match the inequality they benefit from.

I’ve seen a lot of people misunderstand what we’re saying. Our claim is that in a world of full automation, inequality will skyrocket (in favor of capital holders). People aren't thinking about the galaxies. The relative wealth differences in a thousand years—or a million—will be downstream of who owns the first dyson swarms and space ships. And space colonization isn't bottlenecked by people’s preference for human nannies and waiters. So even if you can make 10 million dollars a year as a nanny in the post-abundance future, or get a 10 million dollar charity handout, Larry Page’s million cyborg heirs can own a galaxy each. You might think this is fine! Why is inequality intrinsically bad, especially if absolute prosperity for everyone goes up? Fair enough, but to me quadrillion fold differences in wealth between humans seem hard to justify in a world where AIs are doing all the work anyways - these disparities in wealth are not incentivizing hard work or entrepreneurship or creativity, which is what we use to justify inequality today. Just to recap, full automation kills the corrective mechanism on runaway capital accumulation - which is that you need labor to actually make productive use of your capital, thus driving up wages. Some people asked: why assume AGI leads to full automation? Maybe people will still prefer human nannies and waiters. Even if true, we think labor's share of GDP—which has been roughly 2/3 for centuries—would still likely collapse toward zero, massively increasing inequality. Here's why. It sometimes happens that when machines are only slightly better than humans, people sometimes pay a premium for the human version. But once machines become much better, that preference disappears. When carriages were not much faster than being carried on a litter, the rich sometimes preferred the litter. Now they prefer the car. They might still have a chauffeur—but once self-driving vehicles are allowed to move far faster, human-driven cars may be relegated to a slow lane. If the economy grows 100x, wages must also grow 100x for labor's share to stay at 2/3. But prices are relative—so this means human labor becomes 100x more expensive compared to AI-produced goods. A human-cooked meal costs 100x what the robot version does. For labor share to hold steady as that ratio grows to 1,000x, then 10,000x, the preference for human-made goods would have to become increasingly fanatical. And there's a second problem: the higher wages rise, the greater the incentive to develop machine substitutes for whatever services humans still provide. The premium on human labor is precisely what incentivizes its own replacement. Just to clarify a few other things: - “Piketty’s long run series are disputed.” We spend a long chunk of the essay explaining why Piketty is wrong about the past! But we’re arguing that the assumption he makes (specifically that labor and capital are substitutes) would be true of a world with advanced enough automation. We spend so much time rebutting his claims about the past because the wronger you think he was about the past, the more you think will change once his assumption comes true. - “A capital tax would lower growth.” Yes, as we point out, capital taxes incentivize consumption now instead of saving and investing for the future, at the margin. But if capital is the only factor of production, then it’s hard to come up with an inequality-capping tax that doesn’t lower growth. - “Capital can escape, both across time and space. This makes a wealth tax impractical.” We agree! As we say in the essay and in the tweet summary below, it would be really hard to implement Pikkety’s flagship solution (a high and progressive global wealth tax). You could go Georgist and try to tax land, but the natural resource share of income is only 5% and is likely to stay low until we hit “technological maturity” for reasons we explain in the essay. We don’t see any easy ways to avoid (literally) skyrocketing inequality - in fact, that’s what inspired us to write the essay and explain this problem in the first place. Also, to address a subtext: I think the currently proposed California wealth tax is a very bad idea for many reasons. This essay is about inequality under full automation, not about how California can make its healthcare expenditures more sustainable.

1/ 🚨Today history in AI 💻+ biology 🧬was made! CZI led by Mark Zuckerberg + Priscilla Chan who've admirably committed tens of billions (and the entirety of their wealth to CZI) has acquired Lux family co Evolutionary Scale––just 2yrs after it's founding at a great multiple...

Many Congratulations to @jianyuan_wang, @MinghaoChen23, @n_karaev, Andrea Vedaldi, Christian Rupprecht and @davnov134 for winning the Best Paper Award @CVPR for "VGGT: Visual Geometry Grounded Transformer" 🥇🎉 🙌🙌 #CVPR2025!!!!!!

If you value CapEx above everything, you're gonna have a bad time

@willccbb Yeah, using the acronym without expanding it was perhaps premature .

remember when everyone was obsessed with Komolgorov-Arnold Networks for like 4 months

@BlackHC CRISPR?

What can change the nature of a man? (Bonus points if you don't have to look it up and don't mention the reference 😊)

@jmschreiber91 Neat - will check out the paper!

@lbertinetto Yes -- exactly. A great deal of the work I've done so far is showing that, despite being an "attack" on the model where there is a risk of falling off-manifold, that the results make biological sense.

Ledidi turns any genomics ML model into a controllable sequence designer by inverting the normal ML paradigm. Now, it is significantly faster, flexible, and more powerful than before. Available on GitHub and installable with `pip install ledidi`

@BlackHC @NeurIPSConf Wow - congrats!!

I'm honored and overjoyed that RhoLoss for token selection in LLM pre-training is a runner-up to the NeuRIPS 2024 best paper award 🥳🥳🥳 Really happy about the extended team effort by Microsoft for "Rho-1: Not All Tokens Are What You Need" ☺️😇 📖 @NeurIPSConf #NeurIPS2024

@jeremyphoward Tried and liked ! Any way to use a leaner model than Claude? it's pretty slow and prob an overkill

I can't begin to describe how life-changing this new project, ShellSage, has been for me over the last few weeks. ShellSage is an LLM that lives in your terminal. It can see what directory you're in, what commands you've typed, what output you got, & your previous AI Q&A's.🧵

@eric_brachmann "Let's play a game we call _Name That Dataset!_" :)

As you write your #CVPR2025 papers, think of captivating first sentences. Example? "Creating noise from data is easy; creating data from noise is generative modeling." (from Song et al., ICLR21, arxiv.org/pdf/2011.13456)

@giffmana 👀

ChatGPT 4.5 release delay confirmed. Probably coming after the M4’s to give Apple some breathing room. Unrelated, i hear whispers that 3.5 Opus finished safety tests, and Claude 4 aka «  the bandwagon » started training last week.

@NeelNanda5 Relevant supbowie.com

I'm entertained by the reactions to this tweet - how old do people think I am?!

@mandelapatrick_ @badboyboyce @RhianHorton @kevinweil @elizabeth @smothers_john @christenobrien @apartovi @nbanta @russelljkaplan @stopman @demi_guo_ @KelvinBeachumJr @dojideon @DoRayMo @annabelstrauss @danmorenooo @jebank @joshim5 @KylePMaloney @julesdwalt @mgarimella @joyjohnson @lukemelas @shamayn @dafeinberg Congrats!! Very exciting

Our amazing angels: @russelljkaplan , @stopman, @demi_guo_ , @KelvinBeachumJr, @dojideon , @DoRayMo, Nancy Xu, @annabelstrauss, @danmorenooo, @jebank, @joshim5, @KylePMaloney, @julesdwalt, @mgarimella, Arquay Harris, @joyjohnson, @lukemelas, @shamayn, @dafeinberg, Sam Wiener and Melanie Scott. (3/4)

I’m excited to announce that @vybeappp has raised $4.75M from @stellation, @scribble, @neo, @coalitionvc, @Blueprint. Imagine a world where anyone has access to Beyonce’s photographer on demand in their pocket? vybe - the ai camera makes it one-click. Photos have always been the dominant format of self-expression throughout history. From cave paintings & Kodak analog cameras to filtered mobile photos on Instagram & Snapchat, people have wanted to capture reality & imagination into a visual canvas. Photos have since evolved into a cornerstone for community building and human connection. With the advent of AI, this moment in history presents a new inflection point in the story of photography. It introduces a new era where every story featuring you and your friends is captured with the cinematic quality of a James Cameron film. Your imagination is the only limit for you and your friends. This vision has deeply resonated with our predominantly Gen-Z community, who have consistently used vybe as their personal photographer for months. Their valuable feedback has been instrumental in refining and perfecting the core experience, leading to the product you see today. Our community has created millions of images, demonstrating the immense potential of AI-powered visual storytelling. (1/4)

@aidan_mclau Naah. The ultimate unreleased Anthropic model is Claude Shannon

can someone who understands business economics better than i explain why oai doesn’t release q* for like $50 per prompt and why anthropic won’t release claude-3.5-deus for $500/1M tokens? upsides: >update public on capabilities >make a lot of money downsides: >?

@cursor_ai Well deserved, great product

We've raised $60M from Andreessen Horowitz, Jeff Dean, John Schulman, Noam Brown, and the founders of Stripe and Github. Cursor has become recognized as the best way to code with AI, powered by an ensemble of custom and frontier models, delightful editing, and petabyte-scale infrastructure. Our mission is to create a magical tool that will one day write all the world's software. Join us!

@srush_nlp Yep! Didn't find any downside yet

Python mafia: if I'm using mypy, flake8, black/isort, my understanding is that I should replace them with pyright, ruff, yapf? The main upside is speed and vscode compat? Any downsides?

I am thrilled to announce that @RecursionPharma has entered into a definitive agreement to combine with @Exscientia – bringing together our high throughput target biology, foundation models & supercomputing w/ Exscientia’s generative #AI drug design & #chemistry #automation. What’s ahead: Our shared portfolio in #oncology, rare disease, & infectious disease expect to have ~10 clinical readouts in the next ~18 months w/ peak sales opportunities of $1B+ for most if successful. Our shared #pharma partnerships include: @Roche @Genentech, @Sanofi, @Bayer, and @EMDGroup, leading to potentially $200m in milestone payments over the next 2 years & the potential for $20B+ in combined milestones. The #TechBio future is here. $EXAI + $RXRX. Review important and required disclosure related to this information here globenewswire.com/news-release/2…

@chaseleantj "Thanks for the folks at Recall for partnering with me on this post" You could have started with that

That’s all for now. Thanks for the folks at Recall for partnering with me on this post, and I hope you found it useful. It’s a great app, whether you want to learn something new, or just wanna keep track of your stuff. You can try it for free here: getrecall.ai

This has got to be the coolest AI notetaker ever. It can summarize 2hr YT vids, create knowledge graphs of your notes, and even make flashcards from them automatically! Here’s how it works: