Hacker News: red75prime

New comment by red75prime in "I think Anthropic and OpenAI have found product-market fit"

red75prime — Wed, 27 May 2026 21:25:20 +0000

And around 400k H-2A workers. Humanoid robots... Who works on them I wonder.

New comment by red75prime in "I think Anthropic and OpenAI have found product-market fit"

red75prime — Wed, 27 May 2026 21:12:00 +0000

> 200m knowledge workers in the world, 30m developers

Your scope is too narrow. The companies target more than white-collar jobs. And $1t is around 0.5% of the world economy.

New comment by red75prime in "Where does next-token prediction leave us?"

red75prime — Wed, 27 May 2026 18:25:19 +0000

Nah. It was welfare for Roman citizens: cura annonae.

New comment by red75prime in "Where does next-token prediction leave us?"

red75prime — Wed, 27 May 2026 07:41:25 +0000

Correct me if I wrong, up to 40% of Roman citizens were on "welfare". This, at least, shows that it's not impossible to have a functioning society with a large percentage of people who only contribute political activity.

New comment by red75prime in "Waymo pauses Atlanta service as its robotaxis keep driving into floods"

red75prime — Thu, 21 May 2026 20:40:13 +0000

VLAs (vision-language-action models), which are offspring of LLMs, and their versions that are more suitable to edge devices are being used in self-driving to add common sense to path planners ("don't drive through a police standoff", things like that).

New comment by red75prime in "What Do Gödel's Incompleteness Theorems Mean?"

red75prime — Thu, 21 May 2026 15:50:51 +0000

Another way around the Rice's theorem is the Curry-Howard correspondence. A constructive proof of existence of a program that has a property can be transformed into a program that has this property. Yet another way is to have a programming language where syntactic correctness implies a range of semantic properties.

New comment by red75prime in "An OpenAI model has disproved a central conjecture in discrete geometry"

red75prime — Thu, 21 May 2026 10:50:56 +0000

StockFish amplifies weak "intuition" (heuristic and simple neural estimators) through extensive Alpha-Beta search made possible by the low(ish) branching factor of chess. It doesn't work for Go already (KataGo uses larger neural networks to guide search more efficiently). I doubt either will work for math where branching factor is even bigger and the success criteria (is the result interesting and so on) are not strictly defined.

New comment by red75prime in "An OpenAI model has disproved a central conjecture in discrete geometry"

red75prime — Thu, 21 May 2026 10:29:29 +0000

"The Nature and Meaning of Numbers" (1888) by Dedekind. He proved that any set of mathematical objects conforming to the second order Peano axioms is isomorphic to the natural numbers. Peano axioms basically formalize the notion of well-behaved (that is practically useful) counting numbers.

If you were to dig deeper, you'd get to the murky philosophical depths of foundations of mathematics, but I prefer to not go there. Practically, if you want to reliably count something, you end up with the natural numbers (or, maybe, their subset that ultrafinitists are trying to formalize).

New comment by red75prime in "An OpenAI model has disproved a central conjecture in discrete geometry"

red75prime — Wed, 20 May 2026 22:35:21 +0000

It has been tried. Lenat's Automated Mathematician, for example. The problem is that the system succumbs to combinatorial explosion, not knowing which directions are interesting/promising/productive. LLMs seem to pick up some kind of intuition from the data they are fed. The generated data might not have the needed "human touch" or whatever it is.

New comment by red75prime in "An OpenAI model has disproved a central conjecture in discrete geometry"

red75prime — Wed, 20 May 2026 22:18:13 +0000

On the other hand, it is proven that if you need to count things, the only thing you can discover/invent is the natural numbers.

New comment by red75prime in "Tesla Solar Roof is on life support as it pivot to panels"

red75prime — Sun, 17 May 2026 23:19:03 +0000

The claim was that FSD disengages before collision without user input (presumably to fool someone into thinking that it's the driver's fault). It's not clear who they are trying to fool though. The public will not buy it. NHTSA requires reporting the crash as ADAS-related if ADAS was active at any point during 5 seconds before the crash. The court would just classify this tactic as criminal negligence in the design (if they can detect imminent crash, why they disengage FSD instead of initiating emergency braking?).

All-in-all, the trick that can't fool anyone and that doesn't make any sense. If this claim was true, the only explanation would be that Tesla is evil for the sake of evil and to the detriment to itself. Evil and dumb.

On the other hand, pressing the brake is a common way of disengaging ADAS. Tesla is no better and no worse in this regard than other ADAS manufacturers.

New comment by red75prime in "Tesla Solar Roof is on life support as it pivot to panels"

red75prime — Sun, 17 May 2026 21:31:11 +0000

I caught him a few times not writing significant details that go against his narrative.

"FSD disengages just before the collision." The other video angle shows that the driver presses the brake, which disengages FSD. "Tesla consistently hides information from the court." There are two different cases separated by years. The police got all the information they needed in the first case. "FSD is 10x worse than the average driver." The uncertainty of the number due to insufficient statistics makes the comparison moot.

New comment by red75prime in "Accelerando (2005)"

red75prime — Sun, 17 May 2026 09:08:44 +0000

At the time I devoured the work in one sitting with great delight. Today, the libertarianesque naivete detracts from the story a bit.

New comment by red75prime in "The sigmoids won't save you"

red75prime — Sun, 17 May 2026 08:56:49 +0000

LLMs capture a snapshot of the "collective brain" functionality as represented by literate people. I don't speculate how our brains get to this point.

New comment by red75prime in "Accelerando (2005)"

red75prime — Sat, 16 May 2026 19:22:29 +0000

> With no need to build, does he spend time on design then?

He's a cyborg: "the thousand petaflops of distributed processing power running the neural networks that interface with his meatbrain through the glasses." Why he is special? Who knows. Maybe he has a talent of interfacing with the nets through the crude hardware of the era. Maybe it's connections you mention.

New comment by red75prime in "Waymo updates 3,800 robotaxis after they 'drive into standing water'"

red75prime — Sat, 16 May 2026 06:26:45 +0000

Also google "sensitivity-specificity tradeoff." Initiating emergency braking when at least one sensor detects a pedestrian has the highest sensitivity, but the worst specificity. The real trolley problem of self-driving: how many spilled drinks and read-end collisions society is willing to tolerate to save one life.

New comment by red75prime in "Heritability of human life span is ~50% when heritability is redefined"

red75prime — Sat, 16 May 2026 04:33:51 +0000

I have no problem finding whether this rock is too heavy to lift for me in these specific circumstances. But if we try to define if it's too heavy in general, it becomes not well defined. Specific traits have causal development pathways that theoretically should allow to predict genes/environment interaction, but heritability uses population-level statistics.

New comment by red75prime in "The sigmoids won't save you"

red75prime — Sat, 16 May 2026 03:55:32 +0000

> basically how well a universal function approximator can fit to a function we define

That's what you've got wrong. We don't define functions that an LLM approximates. Autoregressive pretraining approximates an unknown function that produces text (that is what the brain does). RL doesn't approximate functions, it optimizes objective by finding an unknown function that performs better.

New comment by red75prime in "Heritability of human life span is ~50% when heritability is redefined"

red75prime — Thu, 14 May 2026 10:59:58 +0000

There's no well defined "a rock too heavy for a person to lift" too, but we manage. So, what's the point?

New comment by red75prime in "Starship V3"

red75prime — Thu, 14 May 2026 00:14:29 +0000

"This company is overvalued" and "The market is not rational anymore" are a bit different claims.

One of the assumptions I mentioned is "significant overhaul of economy is impossible."