IMO the early indicators of a pop here will be…
- a substantial drop in the cost of GPU compute on markets like hyperbolic.ai
- wait times for gas-fired turbines going from years to months
- commercial HVAC companies forecasting a decrease in datacenter revenues

That’s extra concerning here because capital investment in GPUs depreciates particularly quickly & consistently. If revenues don’t soon explode upwards, investors (not just confined to big tech) may be left holding datacenters that they aren’t financially viable.

A great example is AWS letting Blue Owl finance/own 80% of their new $27bn datacenter in Louisiana. This kind of debt-fuelled investing is indicative of late-ish stage bubbles.

As far as I understand, this “image encoding” is also something like a much richer text embedding. Curious as to whether this could open up some new avenues for semantic analysis of more complex/abstract textual data.

Some really cool potential use cases here like compressing session contexts or improving the feasibility of using LLMs for retrieval across large quantities of unstructured textual data.

Like it’s one thing to disagree with a specific statement he made about the Artsakhci genocide or whatever, it’s another to imply that it’s perfectly fine for border control to intimidate him due to his political beliefs because he isn’t leftist enough for you.

Christ, you probably have more in common with him politically than about 80% of Americans? I’ll never understand the compulsion to continually exclude people from your sphere of acceptable opinions until it’s just you. Focus on the views you share - no one will agree with you on everything.

Either way, a super cool concept that arguably opens up one possible path towards AI that surpasses human intelligence.

Without having looked into it in much detail I wonder if it might work great for short, leetcode-style questions without showing much promise for longer problems that require more complex reasoning (closer to real-world SWE tasks).

What do you suggest he does? Genuinely - what outcome here would make you happy? I actually can’t fathom the level of cynicism required to look at someone voluntarily giving away >$100bn to the most vulnerable people on the planet and still call him an asshole.

en.m.wikipedia.org/wiki/Buffett...

You are the most smart and enlightened 🙇‍♂️. We all applauded your refusal to acknowledge the existence of nuance. Truly an inspiration to us all.

Many would see this as a regression rather than progress, which is fair. But I think given the political environment over there at the moment and how bad things could get, something a bit steadier would be good for us all.

It might require a (Bill) Clinton level communicator with Obama charisma that can convincingly distance themselves from both major parties, but I think there’s enough contempt for both parties and apprehension towards radical ideas that it’s possible.

Optimistically, I feel like it’s possible this could be achieved through a repackaging of basic neoliberal ideas with 2018 dem levels of social justice.

Since not enough time has passed for the dems to rehabilitate their image it’s more possible than ever (granted that isn’t saying much) that something new could emerge and become a force in US politics.

The democrats opened the door to anti-establishment populism by being chronically bad at communication/messaging and allowing themselves to be cast as elitist, just for those riding the populist wave to make such glaringly bad decisions that it seems they might lose support quicker than expected.