cs702

According to page 6, year to date in 2026, operations generated $127B in cash, and capital spending on property and equipment was $132B, so year to date, the business consumed $5B.

According to page 8, in the last quarter, operations generated $42B and capital spending was $51B, so over the past three months, the business consumed $9B.

Those figures don't include any capital spending made through off-balance sheet arrangements. No idea what those might be, or for how much Alphabet could be on the hook.

I'm curious to see if all that money invested in infrastructure will pay off. I hope it pays off, for everyone's sake, but hope is not certainty.

Those holdings are carried in the books at valuations that arguably could be justified in the past but are no longer realistic.

Private equity firms are basically sitting on them, unable to sell them or list them at current valuations.

It's not that surprising, given that these models generate each new token by sampling from a context-determined probability distribution.

If you run the same prompt repeatedly many times, it becomes increasingly probable that a tiny percent of the generated sequences will go off in, ahem, unusual directions.

Lots of theories seeking to explain this reality, but that's all they are: theories. Maybe some will be proven right. Maybe all will be proven wrong.

As of right now, the economic reality lags far behind the hype. But down the road... I wouldn't place a bet either way.

> In the years ahead, this may cost American companies tens if not hundreds of billions of lost revenue.

The question is, if European governments and businesses conclude they cannot trust non-European software and hardware, what can they do about it?

On the software side, it seems to me the only realistic alternative they have is to build upon existing open-source projects.

On the hardware side, their options seem more limited to me, but I'm not an expert on the subject.

Modern best practices for a medical team (standard roles, responsibilities, consequences, processes, etc.) have evolved through trial and error, since the advent of civilization, to prevent costly human error.

In hindsight, it's not too surprising that the same best practices can be applied by teams of AI agents to minimize costly AI error.

It's still incredible. We sure live in interesting times!

Thank you to the author for writing and to the poster for sharing it on HN :-)

It seems the AI wrote code in Lean that proves there are solutions to Navier-Stokes that can blow up, but...

the AI's explanation of the code, in natural language, does not correspond to the Lean proof!

That is... so rich in irony.

What I do know is that those cute one-syllable names are meant to make you feel comfortable with AI agents who are deep in your business all the time.

---

[a] https://www.artsy.net/article/artsy-editorial-life-death-mic...

They keep talking about all the very bad things their AI could do, but instead of assuming responsibility and promising to find solutions, they're saying catastrophe can only be avoided if the government regulates AI development, which would make life much harder for smaller labs and open-source competitors. No one should be surprised by this.

On the flip side, it's understandable that none of the labs want to take responsibility when their AI software does something naughty. The potential liabilities are basically infinite.

The root of the problem is that no one wants to take responsibility when AI software does something naughty.

I'd say it's the influence is kinda narrow, focused on issues that are important to members and lurkers but not necessarily mainstream.

Educated people make societies more civilized, and they in turn invest to educate more people.

Unless there's a catastrophe. In that case all bets are off.

It was also an economic disaster. Surely that contributed.

Is it possible to have a civilization that works well, with institutions that work well, if the population is poorly educated?

Sigh.

What I don't know, and was asking about, is whether the cost of providing that service is also falling rapidly, including the cost of training.

What about the cost? Inference looks like a viable business model for those operators that have access to SOTA model and serving infrastructure, but the investment required to have it is enormous, and appears to be never-ending, because if an operator stops investing aggressively, its model and serving infrastructure quickly become non-competitive, and customers will quickly leave for alternatives.

You may be right. I'm not so sure. Inference looks like a viable business model for those operators that have SOTA infrastructure in place, but the investment required to have it is enormous, and appears to be never-ending, because if an operator stops investing aggressively, its infrastructure quickly becomes non-competitive, and customers will quickly leave for alternatives. SOTA infrastructure is a moving target.

Free cash flow (operating profit less investment), actual cash coming in, is deeply in the red.

EBITDA can be a sensible measure of profitability when there isn't much need for additional investment. That doesn't seem to be the case with these operators. They need to invest aggressively to avoid losing customers to competitors. All of these operators have made multi-year commitments to invest more in infrastructure. In addition, they have guaranteed quite a bit of debt to fund it.

Maybe it all will work out fine (and I sure hope it does!), but I didn't see any hard data from the OP, or from you, supporting that view.

The only aspect that is poorly analyzed by the OP is business model viability. All players are investing insane amounts of money in infrastructure with the expectation that their future profits will justify all that investment. The winner or winners in the AGI race, they believe, will find the proverbial "pot of gold at the end of the rainbow."

The OP glosses over questions of business model viability with a brief qualitative discussion and very little hard data. For example, to earn an annual return > 10% on every trillion dollars of capital sunk into infrastructure, the owners of that infrastructure must earn free cash flow (operating profit less investment) in excess of $100 billion per year in perpetuity. Is that feasible? Why? How?

The OP does not really consider such questions.

While there are some necessary simplifications, I couldn't find a single mistake on a first pass.

Anyone who wants to understand current monetary systems should read it.

Substituting "stones" for gold is a brilliant stroke of the pen. It will surely be helpful to any readers who are hung up on gold.

Thank you for writing this and sharing it on HN.

But the submission title calls it a "backprop alternative." It is not, at least not yet.

I think my initial assessment is correct: interesting research, but not an alternative, at least not yet.

I didn't see ImageNet. TinyImageNet is something else.

How does it do on CIFAR-10, or even better, ImageNet?

Interesting research, not sure it's a backprop alternative.

===

EDIT: accuracy on MNIST is not ~90%. It's ~85%.