mondrian

Europe is a different beast. They build and distribute rails, they don’t compete on frontier capabilities. When the civic benefits of AI become clear, EU is in a position to mandate their distribution. US develops capabilities that remain stuck in heterogeneous corporate silos without interop. Payments is a good point of comparisons between the two approaches.

Switzerland still has parties and a senate but the people can veto any law and the people can amend the constitution. This arguably is a sensible balance as it’s probably not tenable for the people to legislate full-time. Representatives can legislate and the people will override them when they disagree.

Van Gogh, Cezanne, and Picasso were Europeans, and Europe is where art speculation and the "Art Market" was born, and where Art History became equated with Art Market History. Americans showed up with a 20 year lag and kept lagging, with e.g. Pollock imitating Picasso in the 1940s. By then, globally-viral-Picasso was selling paintings for hundreds of thousands of inflation adjusted dollars.

This argument goes too far to the other end. Never name a group, cuz by golly you’re broad stroking. Only specific character assassination is allowed or keep quiet? Like a Cancel Culture Wittgenstein.

Sought

It sounds ‘the singularity is near’ not so much because AI is reaching escape velocity forward, but because we’re systematically pushing humans back beyond the threshold where computers are legible?

Key word: perform, i.e. execute. Scale indeed comes by performing more things per unit time, things we understand, on execution engines we understand.

Undecidability is a problem of answer-extraction from a process, it doesn’t preclude the process from executing deterministically. The universe could well be the live execution of a deterministic, even basic algorithm, with all kinds of questions about its execution being undecidable.

Guy made an account just to post that gem.

One of the key points of that paper is that the body of written works is biased and those biases will be amplified by compressing that body into LLMs, with outlier data, with low coincidence rate, being suppressed as unreliable - data that is structurally dissimilar to the bulk is noise. Similarly to how PageRank suppressed nodes with low number of edges, probably contributing significantly to the homogenous corporate mall-internet we enjoy today.

Structurally speaking, being left and pro-open borders is incoherent. The left is about collective power controlling labor supply and using it to tamp down corporate overreach. Open borders and H1B that allow corporations to side-step local worker bargaining power are squarely antithetical to being Left.

I think 'liberal progressives' that are mainly concerned with identity representation, not working class power, are rhetorically bundled as 'far-left'. To people who buy that, Bernie Sanders (a Leftist) and Kamala Harris (a Third Way liberal progressive) are both far-left. It's incoherent but rhetorically potent.

A better chart: https://ichef.bbci.co.uk/news/1536/cpsprodpb/f3fb/live/1ca3d...

The index of your 20 line file is <20TB number>

Of course native people didn’t call themselves Indian prior to settlers assigning them that name, with legal consequences. But native people usually refer to their tribe identity, like Diné or Lakota, rather than the generic term Indian.

The irony in your example is the modern Pentagon is largely a collection of private companies.

Let's get back to filling the front page with Web3, DeFi, NFTs. Oh the good ol' days.

Sounds like this: https://solidproject.org/about

But yea privacy is a silly thing to propose to a surveillance industry.

Black Mirror episode: https://en.wikipedia.org/wiki/Common_People_(Black_Mirror)

I was off by a zero, thanks! He sold about 100M of NVDA on 16B net worth which would be 6.25k on 1M.

So this 'portfolio' is <200M on a net worth of 16B? This is like someone worth 1M selling $600 worth of stock.

The parent comment is probably just talking about temporarily timing the bubble pop.

How will the AI wealth get distributed to people at large?

Bernie Sanders talks about a "robot tax" that is roughly what you're talking about. https://www.businessinsider.com/bernie-sanders-robot-tax-ai-...

Comparing it to GDP doesn’t seem to make sense. Maybe to government revenue.

Good point. In the scenario I described, I'm assuming Stripe will launch their own stablecoin. I tend to think all major tech companies are incentivized to launch stablecoins and give you discounts and perks when you transact using their stablecoin in their own ecosystem. The more of their stablecoin they issue out, the more money they make on interest.

Yeah and this is codified in the GENIUS act which passed recently. It enables tech companies to act like banks in certain dimensions, without being regulated like banks.

Yeah. Stablecoins create demand for Treasuries which drives the price of Treasuries up and interest rate down. So this pressure lowers debt servicing cost for the US government, and Stripe is the holder of those Treasuries and gets paid interest.

This would also serve to counter the drop in global Treasury demand due to recent tariff stuff where presumably our traditional debt holders are losing appetite for US debt...

It also creates a kind of strange situation where stablecoins are basically spendable "Treasury tokens". So you give 1 USD to Uncle Sam (via a middle man like Stripe), get back 1 stablecoin. Then you go and spend the stablecoin, and Uncle Sam goes and spends the USD. It's like a weird double spend situation. Prior to stablecoins, you buy a treasury bill with USD, you hold this unspendable treasury bill while Uncle Sam gets USD to spend.

The GENIUS act enables tech companies to become reserve holders -- buy US Treasuries with customers' money. Stripe offers a "transactional ecosystem" to the customer in stablecoins, the customer gives USD to Stripe in exchange for stablecoins, Stripe buys short-term Treasuries and makes a shitload of money on interest.

Part of the very high level play is the US Govt seeks to diversify away from depending on nation states for borrowing, and to promote tech companies to the status of reserve holders.

This doesn't add much to the consumer however. I think in fact we are looking at a "fragmented currency" future where you hold like 36 different stablecoins in your wallet because certain platforms accept certain stablecoins. The GENIUS act doesn't offer strict guarantees for getting out of a stablecoin into USD, so I predict dark patterns and "incentives" to make it hard to get out of a stablecoin.

Trying to make AIs more factually-accurate with more training is probably hopeless. Current events and encyclopedic knowledge will be provided by tools. The LLM's core job is to choose the right tools for the job and synthesize their outputs.

Also in the vicinity: https://www.anthropic.com/research/tracing-thoughts-language...

There's also distillation, where you can drastically improve a small model by training it on chains of thoughts of larger models. You can't achieve the same performance by training on original human texts. This suggests that those chains of thoughts reliably contain "densely packed reasoning", meaning the LLM probably has developed internal clusters of "reasoning circuitry", loosely speaking.