Hacker News Reader

meroeson OpenAI claims gold-medal performance at IMO 2025

Math RLHF already has verifiable ground truth/right vs wrong, so I don't what this distinction really shows.

And AI changes so quickly that there is a breakthrough every week.

Call my cynical, but I think this is an RLHF/RLVR push in a narrow area--IMO was chosen as a target and they hired specifically to beat this "artificial" target.

meroeson OpenAI claims gold-medal performance at IMO 2025

https://work.mercor.com/jobs/list_AAABljpKHPMmFMXrg2VM0qz4

https://benture.io/job/international-math-olympiad-participa...

https://job-boards.greenhouse.io/xai/jobs/4538773007

And Outlier/Scale, which was bought by Meta (via Scale), had many IMO-required Math AI trainer jobs on LinkedIn. I can't find those historical ones though.

I'm just one piece in the cog and this is an anecdote, but there was a huge upswing in IMO or similar RLHF job postings over the past 6mo-year.

meroeson OpenAI claims gold-medal performance at IMO 2025

Latent??

If you looked at RLHF hiring over the last year, there was a huge hiring of IMO competitors to RLHF. This was a new, highly targeted, highly funded RLHF’ing.

meroeson OpenAI claims gold-medal performance at IMO 2025

They didn’t actually compete.

meroeson OpenAI claims gold-medal performance at IMO 2025

In the RLHF sphere you could tell some AI company/companies were targeting this because of how many IMO RLHF’ers they were hiring specifically. I don’t think it’s really easy to say how much “progress” this is given that.

meroeson AI therapy bots fuel delusions and give dangerous advice, Stanford study finds

DOA doesn’t mean forever and always. But certain claims like the advent of living beyond 200, humans on Mars, etc can just be immediately dismissed outright for several decades. What you’re talking about is unsupervised LLM therapy. Even when my dentist used AI to read my X-rays, he’s overseeing everything. I’m fine sticking my neck out to say LLM therapy is DOA for the foreseeable future.

meroeson My bank keeps on undermining anti-phishing education

My bank’s fraud department uses text shorthand like “Stop2end” and “call ph#” and their dates lack spaces “24Jun” in their texts to me.

Is this some kind of meta-level play to sound less fake?

meroeson AI therapy bots fuel delusions and give dangerous advice, Stanford study finds

There's no chance LLMs have a sufficient training set of effective therapists-patient interactions, because those are private. Ergo, there is no need to wait, it's DOA. Anything else is feeding into LLM hype. It's that simple.

meroeson US importers sued for 'greenwashing' Mexican avocados

Driving through the Central Valley I’ve seen signs for 10 for $1. No idea if they are extra small, but typically it’s $1-$2.50 per

meroeson Bad Actors Are Grooming LLMs to Produce Falsehoods

LLMs are “taught” two kinds of “truth”. One is 100% adherence to a reference text. If the text says the Coliseum is in Antarctica or 1+1=716, model must too. The other is adherence to reputable outside sources.

Not sure if it’s embarrassing or a fundamental limitation that grooming and misunderstanding satirical articles defeat the models.

meroeson AI agent benchmarks are broken

> arithmetic (why would use LLM for that ?)

Because people ask LLMs all of these things, including arithmetic. People were saying the same about the number of r's in strawberry. Why ask and LLM that!?!? But the big AI companies want LLMs to be better at these questions, probably because people ask them to LLMs. The big AI companies want this because there is no other explanation for the money poured into RLHF'ing these types of problems.

meroeson A non-anthropomorphized view of LLMs

> there exists a string of words (not necessary ones that make sense) that will get the LLM to a better position to answer

exactly. The opposite is also true. You might supply more clarifying information to the LLM, which would help any human answer, but it actually degrades the LLM's output.

meroeson AI note takers are flooding Zoom calls as workers opt to skip meetings

The problem is these meetings are so low information density even an AI summary is not worth my time. And it’s not some elitist mindset. It’s like the entire reason there are these regular meetings is to make some mid level person feel better. They like giving directions vocally because that authority is harder to question than if they wrote up a memo and all the receivers can poke holes in it. I’m convinced most meetings are to make up for poor writing skills.

meroeson The bitter lesson is coming for tokenization

I don't buy the token explanation because RLHF work is/was filled with so many "count the number of ___" prompts. There's just no way AI companies pay so much $$$ for RLHF of these prompts when the error is purely in tokenization.

IME Reddit would scream "tokenization" at the strawberry meme until blue in the face, assuring themselves better tokenization meant the problem would be solved. Meanwhile RLHF'ers were/are en masse paid to solve the problem through correcting thousands of these "counting"/perfect syntax prompts and problems. To me, since RLHF work was being paid to tackle these problems, it couldn't be a simple tokenization problem. If there was a tokenization bottleneck that fixing would solve the problem, we would not be getting paid to so much money to RLHF synax-perfect prompts (think of Sudoku type games and heavy syntax-based problems).

No, why models are better are these problems now is because of RLHF. And before you say, well now models have learned how to count in general, I say we just need to widen the abstraction a tiny bit and the models will fail again. And this will be the story of LLMs forever--they will never take the lead on their own, and its not how humans process information, but it still can be useful.

meroeson ARIA, the UK's Bet to Build Scientific Revolutions

And the way to claw back rank on the world scene surely takes investing, and investing in science is among the best options for long term returns.

meroeson YouTube's new anti-adblock measures

The most disagreeable thing is they are psychologically insidious.

meroeson Congestion pricing in Manhattan is a predictable success

You can cut out auto foam cups without much change compared to redoing the entire country’s transportation infrastructure that was built with unprecedented federal funds.

meroeson Peano arithmetic is enough, because Peano arithmetic encodes computation

The argument is more like our most successful sciences like physics use our most successful math, which is ZFC. And thus incompatible reals are part of this package deal. Maybe it’s possible to do physics without any packaging of lots of these mathematical objects, but to my knowledge it hasn’t been done (physics based on a paired down math, from top to bottom). It’s a de facto require, that despite effort like intuitionism or Science without Numbers has not replaced classic ZFC as our most successful math.

meroeson Peano arithmetic is enough, because Peano arithmetic encodes computation

> absolutely nothing requires an incomputable continuum of reals.

In a sense this is wrong. Some say the indispensability of modern math, which includes the incomputable reals, shows that such abstract objects are required for science.

meroeson Meta invests $14.3B in Scale AI to kick-start superintelligence lab

Doesn't scale contract out/aggregate many smaller companies, and that's where the bespoke solutions are?

meroeson The Gentle Singularity

What? AI proofs are still atrocious.

meroeson What happens when people don't understand how AI works

Solid article

meroeson What happens when people don't understand how AI works

There are middle-men companies like Scale that recruit thousands of remote contractors, probably through other companies they hire. There are of course other less known such companies that also sit between the model companies and the contracted labelers and RLHF’ers. There’s probably several tiers of these middle companies that agglomerate larger pools of workers. But how intermixed the work is and its scale I couldn’t tell you, nor if it’s shifting to something else.

I mean on LinkenIn you can find many AI trainer companies and see they hire for every subject, language, and programming language across several expertise levels. They provide the laborers for the model companies.

meroeson What happens when people don't understand how AI works

Ya I don’t think I’ve seen any article going in depth into just how many low level humans like data labelers and RLHF’ers there are behind the scenes of these big models. It has to be millions of people worldwide.

meroeson Washington Post's Privacy Tip: Stop Using Chrome, Delete Meta Apps (and Yandex)

Hmm how can I use being forced to use Chrome for work, for me tax wise…

If I’m a contractor forced to use Chrome and mobile devices, can I deduct a separate work phone?

I really hate having it my iPhone, at least maybe I can claw something back this way?

meroeson The Illusion of Thinking: Strengths and limitations of reasoning models [pdf]

Yep. This is exactly the conclusion I reached as an RLHF'er. Reasoning/LRM/SxS/CoT is "just" more context. There never was reasoning. But of course, more context can be good.

meroeson Just how bad are we at treating age-related diseases?

Ya for Alzheimer's, the two people in my family who've had it had bad diets (think dining out lavishly every night or sugar-addiction) and poor hygiene (like having every single tooth pulled). I also suspect chronic lack of sleep is my dad's worse health trait. I don't get why they don't work on this stuff personally.

meroeson AI makes the humanities more important, but also weirder

Sounds like all ways to skirt around the actual learning bit. Consuming the classics while gaming? Really?

meroeson My AI skeptic friends are all nuts

> LLMs can write a large fraction of all the tedious code you’ll ever need to write.

But, you still have to read it:

> Reading other people’s code is part of the job...I have to read the code line-by-line anyways.

So instead of writing the tedious code, I only have to read it. Oh but don't worry, I don't have to read it too carefully because:

> Agents lint. They compile and run tests. If their LLM invents a new function signature, the agent sees the error

But remember...

> You’ve always been responsible for what you merge to main.

So now I have to oversee this web of agents and AI ontop of coding? Am I doing more now for the same pay? Am I just speedrunning myself toward lower pay? Is AI adoption a prisoner's dilemma toward lowing my wages hardest?

Because is good at coding compared to many other disciplines (e.g. math), it makes the internal AI politics among programmers more of an issue. Add fuel to that fire baby!

meroeson Adam Riess and the Hubble tension

I don't think the big bang or the universe ever had a center.