Hacker News Reader

Comment by evolarjun | Hacker News Reader

It is odd (or maybe not) that they decided to publish a marketing whitepaper rather than a more traditional journal submission + preprint. The work does appear to be sufficient for a publication, though there's a good chance a reviewer will rip into them for some of the assertions they make, but given the topic I'm sure the paper will be accepted regardless.

The market for entry-level programmers has already declined, but at least they were somewhat in demand and made reasonable salaries. Now what happens to post-docs who already make almost nothing and often get treated like crap?

View on HN

Did anyone read the blogpost? They did publish a pre-print:

  > Our work to understand the primary function of ARTs is ongoing. However, we think it is important to share such findings early, both to demonstrate Claude’s capabilities and to give the broader community insight into what we’re working on. We have released a pre-print (here) that discusses this in more detail.

https://www-cdn.anthropic.com/22573675ada52a8ca8a97a1a4b4326...

Just going to post this. And I think it's safe to say that no, they didn't read it.

What's with all this pre-print business. It became very prevalent during covid, where it felt like every week some new pre-print was published that discusses some new aspect of the virus. These papers would then be used in arguments and put forward as proof of whatever claim the arguer was making.

Every man and his dog can publish a pre-print and in my opinion it's academically worthless.

Journal articles are usually behind a paywall. Preprints are a fully open workaround allowed by most journals.

> Every man and his dog can publish a pre-print and in my opinion it's academically worthless.

Sure but if you look at the authors names and see they have 50 other published papers, you can get a rough idea that it's probably equivalently good to their other work.

Until you've done it yourself, it's hard to grok just how bad the peer review process is. It's like...5% better than nothing.

Honestly you could argue peer review is worse than nothing, as it also filters out actually quality work that violates some dogma of the field.

The review time on top journals is multiple years now (your paper will go through many review loops each of which takes months). It's just totally unworkable for active research, whether you're a student or a corporation.

Preprints are just a way to sacrifice rigor for accessibility and velocity. You can throw out "here, this is what I'm working on, here are the quick and dirty findings" really fast and with little friction.

This does skip the academic "checks and balances" like journal selection and peer review - but it can also help anyone else who's working on the adjacent topics.

If a field is moving fast, and you think there can be some value in your work for others in the near term? Preprint. If your work is too incomplete or too minor to warrant trying to polish and publish it, but you don't want to table it? Preprint. Too deep in corporate structures to care about academic "street cred", and want your work to be accessible? Preprint. Have an exciting early finding that you want to push out there, and are willing to take the rep risks of being wrong about it? Preprint.

There's a reason why preprints came to be the lifeblood of ML.

Academia isn't my thing but I also wonder if there isn't an aspect of putting a stake in the ground? So that if someone beats you to publishing you at least have some record of being on that track.

That is definitely a big motivator for publishing preprints. Journal submissions can take up to a year. Comference submissions take months. If the field is moving fast, claiming a finding early can become an important career move.

In older days, academics would just share notes on their work and word wouldn't usually spread widely before publication.

Preprints may be the better model. But public visibility means that non-experts now get to see the good and the bad research equally, but they won't have the domain knowledge and skill to distinguish one from the other with confidence.

I have a similar to pagerank method I use to evaluate such papers. I look for the references to see how many authors are using their own references (past work), the idea being that people do not jump too far, they make incremental progress.

For the pre-print I could only find only one author who has a single referenced article.

The pre-print authors all have long publication histories.

Pagerank was inspired by academic citation networks; it just turned it in a recursive matrix problem (of which there was some prior literature).

> references to see how many authors are using their own references (past work), the idea being that people do not jump too far, they make incremental progress

The authors are not using their own prior work in the paper, thats the point I was trying to make. I have worked in biotech lab for couple years and its one of the criteria's people use to consider some ones work useful and worth the time.

I think you fundamentally misunderstand both the results published here, and how scientists operating at the highest level of academic research operate. None of these authors has to worry about citing previous work to get the attention of biotech labs.

I don't recall that being there when I first read their post a few hours ago... No way to confirm, sadly, as they've posted it to their own domain

Earliest snapshot is from 3 hours ago, more or less, it already shows the pre-print: https://web.archive.org/web/20260923180804/https://www.anthr...

archive.is has a snapshot from 18:47, showing the link was indeed there a few hours ago. https://archive.is/XM0Nw

I just looked at it. I really hope they're not thinking of sending that to an actual bioinformatics, computational biology or molecular biology journal! So embarrassing...

(I love how Anthropic boast about building a lab, but don't seem to realise that you have to test your hypothesis in the lab! Right now, all their "spectacular" assertions are untested and unproven.)

I realise that this will only improve from here, but gods Anthropic has no idea about the biological sciences right now.

Um. Take a look at the authors. Every single one of them is an expert in this field. They did test this in a lab.

If you want to complain about things like this, it really helps to be specific. Given the author list, it's unlikely they made any truly spectacular errors (and also possible the system they studied is not interesting).

This comment chain is depressing. People who hallucinated things they really wanted to be true. Guess it’s not a uniquely AI problem.

Can you elaborate on what you're thinking? I don't see any support for this being a hallucaination; from what I can see, it's a pretty typical "early biological discovery".

I was poking fun at evolarjun and epihelix for "hallucinating" that this was just a marketing whitepaper (they did publish a pre-print) and that the preprint was "so embarrassing" (you said that the authors are actually experts). It kind of seemed like they had some opinion about AI and this paper, and wishfully concluded things that were not true to support their opinion.

Medical advances are more important than post docs

Absolutely. But without post docs, it won't be long before we are unable to understand what the LLMs are suggesting.

once everyone lives to 200 yo there will be no need for post-docs - afterward we can just upload PI's brain to some AI.

I wonder who's going to pay for that extra 120 years of living we're gonna be doing?

Musk said we will all be rich so we will?

I have no idea what prompts questions like these.

We got to 80 without inventing money. Living that long back then was hard work every day.

Then the industrial revolution happened, and we got state pensions at one end of life and extended childhood a few years past adolescence on the other. We currently pay for this… by taxes funding both education and a pension.

Absent the radical transformations of an AI driven economy, we live 200 years in exactly the same way.

With those transformations, all bets are off unless they violate the laws of physics.

But the majority of people in the world are still working just as hard now as before the industrial revolution? Most of the labor rights activists of the 19th and 20th century spent their time advocating for working hours closer to their pre-industrial grandparents. Industrialization increased working hours with the promise that eventually everyone will have to work less hours once the equipment was built. And yet many generations later countries are talking about increasing retirement age, reducing worker rights, and wage theft is more prevalent than all other forms of theft combined.

What is AI going to do that industrialization and automation hasn't already made the same promises for?

This is a horribly wrong understanding of history. If you think life today is in any way comparable to the Industrial Revolution, you need to pick up a history book because your history classes failed you.

If you think malnourished subsidence farmers could work non-stop before industrialization then you need to think about that a bit more and aread a bit more early and pre-industrial works.

Before the industrial revolution, you started working as soon as you could assist with anything (cooking meant firewood, washing clothes was by hand, etc.); you stopped working when you were too feeble, and only if you were lucky (or from another perspective, very unlucky) a lord's almshouse might look after you.

Life back then was a never-ending quest to make more calories, and you had to consume about 90% of what you made just to not starve (the other 10% went to the lords, the army, and very young children; though I'm oversimplifying here because farm animals also eat and you had to feed them). As total production was lower (and because "preservative" meant cats, alcohol, salt, and grain silos on mushroom-shaped pillars so rats couldn't get in, not industrial refrigeration and sodium benzoate etc.), this meant very different work schedule compare to today; but people were working at the limits of what biology would support, even if hours were fewer (no affordable artificial light to work at night) and "holy days" more plentiful… but on that front, most pop reporting on that seems to forget that today we have two-day weekends, while medieval European communities often only rested on Sunday (and even Sunday-is-rest-day was relaxed somewhat to avoid crop spoilage).

The modern equivalent would be if everyone's job was to hit the gym for 10 hours a day in summer and 4 a day in winter, and still sometimes had mandatory overtime. Some people do labour-intensive work today, but pre-industrial this would be 90%+ of the population and not by choice.

What we actually have in developed nations today, is no significant labour before 18 or over 68, only about 70% the people of working age* are in work at any given time, and the "work" is far less intensive. Less than half of us are employed today to support the whole population. I say "are employed" rather than "work" because childcare and domestic work is still work, but this too is much easier than pre-industrial life.

* "working age" means different things in different surveys: https://en.wikipedia.org/wiki/List_of_countries_by_employmen...

A malnourished subsidence farmer could work non-stop at the same pace as modern workers then I got a bridge to sell you. 19th century worker's advocating for worker rights were literally calling for working standards closer to their subsidence grandparents.

Industrialization increased working hours, not decreased them.

> A malnourished subsidence farmer could work non-stop at the same pace as modern workers then I got a bridge to sell you.

Way to misread what I wrote.

A subsidence farmer necessarily spends their lives doing as much work as they can eat food, because the energy to do the work is that food.

Their work cycle was arranged differently than ours, with harvest season being longer days because letting crops spoil in the fields meant starvation come winter, and winter hours being mostly limited by sunlight and moonlight because artificial illumination was far too expensive.

> 19th century worker's advocating for worker rights were literally calling for working standards closer to their subsidence grandparents.

And? Those workers got those rights, past tense. The fact they got them is a big part of why less than half of the living population in OECD nations needs to work today: their efforts 200-100 years ago are why it is taxes paying for schools and pensions, not the largesse of lords limited to almshouses.

If we suddenly get an anti-aging treatment that has us all live 200 years*, all the governments can trivially handle this just by adjusting pension ages.

* somehow without any of the other things implied by the tech that can make such treatments; add those things in and you have to ask "why only 200?" and "what else can this tech do?" and this is all about AI having been a big contributor to some bio research, so there's a lot of "what else" already and we don't even have the anti-aging treatment yet.

Yeah harvest season has longer hours, I grew up on a farm. But you aren't putting in anything like that amount of work most of the year. You don't work the same fields all year long. You got plenty of other things to do but not a full modern work schedule.

Heating can be done in a few weeks even with handsaws and an axe. A chainsaw makes it faster.

If 12th century peasants had to work anything like a modern work schedule, how would people a 1000 years or more before then survive at all with less technology, tools, and knowledge? How would clothing exist at all without a loom? How did ancient people have any time for discovery and innovation at all if their lives were nearly so grueling.

Those workers rights activists didn't get all that they asked for. And why has it not gotten even better with another 100 years of productivity increases after that?

To me your explainations seem like repeates of what robber barons claimed. The luddites didn't form because people thought industrialisation was allowing them to work less than those before them.

I can't understand what the post-docs are suggesting.

True, but it's also not your job to understand what post-docs are suggesting (hopefully), unlike post-docs who kind of need to understand what LLMs are suggesting.

Not to get back into the AI safety debate from last week, but I really, really want a team of human experts to understand what the AI is suggesting before we administer new medical treatments or synthesize new lifeforms in the lab.

At this point it should be clear that Anthropic are determined to create the doomsday scenario as the most idiotic "I told you so" gotcha in history.

If you can't, then you aren't in the field, and it doesn't matter if you could or not.

There will always be a %age of humans who will be curious how the universe works including LLMs. Delusional to think those humans are only incentivized by degrees, or money.

Humans were curious and started the intelligence / learning explosion much much before money and degrees were invented.

Sure. But unless they are born into a very wealthy family, or decide to pass up on having a family and other hobbies, they won't be able to satisfy that curiosity beyond a very basic level.

For most of history we died of easily preventable diseases because we lacked the scientific institutions to learn how to prevent them. The fact that a percentage of the population wanted to know more is why we live so well today, but it didn’t stop many of those people from dying awfully. It wouldn’t take a lot to go back there, or at least to a civilization that would go back there the second the AIs have a major outage.

many drugs were discovered, evaluated and approved without understanding at all how they worked.

iirc back in the day chemists synthesized a whole bunch of random compounds, observed their effects (in mice etc., or even the chemists tasting them!) then did clinical trials to measure safety and efficacy.

high-throughput screening of chemical libraries on in vitro assays is the modern version of this. "rational" drug design, which uses understanding of mechanisms to design chemical structures for a specific purpose, largely failed back in the '80s.

Well I guess who gives a shit about knowledge

Ive been thinking about this a a lot, like now if LLMs can just try a bunch of things, you can just try them all now. Are there some applications where we can just brute force and test at scale and no longer need to understand? Where understanding is now in essence that it “works” and passes our acceptance tests.

I give a shit about knowledge! I think it's good entirely for its own sake.

I agree! I'm also not going to turn down an Alzheimer's vaccine that's been demonstrated to be safe and effective even if we don't know exactly how it works. (And if you think that's a contrived fantasy, check out the shingles vaccine).

There's been a decent amount of drugs that have been made via a brute force method, but it's a pretty physical process, usually pipetting a few hundred thousand compounds and samples. There's also a lot of drugs where we don't fully understand how they work.

The way I ask it in public situations with scientists is this: If you had an AI that could output cures for diseases (IE, drugs that pass phase III clinical trial, get approved by the FDA, and are highly effective), but it couldn't explain how it worked, would you use it?

Opinions are mixed. Some folks will say that it's morally imperative to cure people even if we don't understand the specific or general principles. Other folks will insist that it's a terrible idea to hand over the comprehension of medical treatments to LLMs, because in the long term it will leave us helpless and dependent.

Is it guaranteed? The most likely outcome is you get rid of both the humans to understand the research and the artificial wizard of Oz was actually just outputting nonsense, and it took killing a bunch of people to find that out.

Post docs is the base that AI was born from...

We still need post docs. What will change is their specializations.

That's why nobody writes their paper on gravity or polio in 2026.

But do we get any without them?

Then we end up like that anti-aging guy and end up older and giving ourselves untreatable diseases

Perhaps if you assume we will continually be offered highly valuable medical advances into the future without humans who understand it.

That's a big assumption to make, so I hope you at least have some proof to back it up.

This is a bit like saying that shoes are more important than shoe factories. Yes, sure, I can't wear a shoe factory, but we'll all run out of shoes if all the factories are gone.

cobblers used to make and mend shoes before shoe factories existed..

Medical advances require post doc levels of education and knowledge to advance. Otherwise we are pretty soon unable to determine what is trash and what is useful. It is easier to generate trash data than good data, and most data generated will be trash, so if nobody can sift through the good and bad the next level is going to be ingesting nonsense and getting worse every time.

> The market for entry-level programmers has already declined, but at least they were somewhat in demand and made reasonable salaries. Now what happens to post-docs who already make almost nothing and often get treated like crap?

Waiting for frontier labs to get into Political Science to show that SOTA models can be vastly better politicians...

It's not odd, it's a for profit company that's using it for marketing.

Why do you think they're going to be treated badly? Right now, I think it's kinda accepted that the people best suited to directing AI for programming tasks are programmers - only we operate at a higher level.

Claude's going to be a similar productivity booster to researchers and postdocs.

I'd be totally lost talking to an AI about biochemistry.

Novel discoveries are now marketing ads.

If individuals had power to allocate funding or not to public research projects rather than get a blanket tax, there would be a lot more conventional marketing in the public sector as well.

No, it's not odd; IMO. It's by design.

I see all of this leading to a setup for: We did cure Cancer, everyone else (Healthcare, Gov., Rx) etc... has just not caught up or even worse; "you just don't have access top that model/version".

I have seen several times on HN recently how people don't see the impact of AI/more code etc... and I believe this is because its following the K-shape of the current economy.

At the top where most of us aren't but CAN see via stock market news etc...; they are making more money by adding efficiencies etc...

At the bottom; efficiencies are being applied at a scale that they could not before such that social and Gov. programs are more manageable and optimized at scale.

> Now what happens to post-docs who already make almost nothing and often get treated like crap?

At least in the US, that particular brain drain has already been happening due to Trump's administration. The best of the best are exiting to other countries that will gladly have them, and then there will be far fewer people getting into the field. Science in general has taken a massive hit under the current administration and it going to take decades to fix if it's even possible.

Why decades?

Need risk aversion to reset after risk was raised temporarily. Humans are cautious creatures

Less risk aversion and more about momentum. You remove billions in funding, people leave the industry that took decades to obtain the education to be functional in, and at the same time now there are less people to train those who will replace them, and you still lack the funding.

If tomorrow you just inject 50% more funding, it doesn't mean 50% more science gets done tomorrow.

I’m curious, what differentiates this from a preprint given the assumption it’s sufficient for publication? It didn’t read like marketing, they don’t seem to sell anything, and there's a link to a not-anthropic.com hosted paper.

> they don’t seem to sell anything

i'll give you a hint: they're selling something

Let's say "anything talking about your product is selling something instead of doing science"

Then we're faced with "why would a (insert whatever makes this a preprint) mean they're not selling something"? (well, at least OP is faced with that, FWIW I think there's ~infinite snarky replies available, but they're sort of uninteresting, no? :)

Well because it is not talking about your product that is what would've created the publication. But what is argued is the research should just be properly published. That publication would have been some advancement in enzyme research and not "Hey we pretend our fancy model created new research and we forget to mention that it was queried for weeks by experts in the field constantly correcting the model whenever it did something wrong until it resulted in something that the researchers could also have created on their own"

Tbh I might be misrepresenting the original post, because in this case I did not read it, but for your point I feel like I also don't have to

> Let's say "anything talking about your product is selling something instead of doing science"

you are gonna be so stunned to find out when more than one thing can be true at the same time

are they doing science? yes. are they selling something? yes.

"Look at what our product achieved" isn't selling anything? Do you want to take at this bridge I'm selling?

Sure! (presuming it's "take a look at" :)

After I entertain you by doing that, is there a steelman version of my reply you're interested in entertaining me with, by replying? Or, just the strawman?

> Now what happens to post-docs who already make almost nothing and often get treated like crap?

This sort of discoveries are what gets postdocs funded lmao.

Every new idea like this creates several years worth of highly specialized work to test out derivative ideas, productizing it, and connecting dots to existing work.

It's not odd at all. Every single "AI did this cool thing" type post is an Ad. Remember AI outputs slop and never produced anything valuable that wasn't heavily assisted by humans or is a lie.

Reply on news.ycombinator.com