This is more accurate. The productive fraction of professors still follow the Pareto principle or 80/20 rule. And even then, these professors aggregate into the elite institutions making it even more skewed.
To get an academic position, you need to have a star advisor either for your degree or postdoc.
I think it was The Atlantic a few years back that ran the numbers on professors and their PhD alma maters. But I can't find the article, so please accept my bad recollection.
Essentially, in nearly all of the humanities, if you did not go to a top 10 PhD program, you had a 0% chance of getting tenure. Not 'like' a 0% chance, an actual 0%. There are no professors at all, anywhere in the US, in nearly all the humanities departments that did not go to a top 10 school. The distribution followed a power law, of course.
However, most universities have PhD programs that will accept students.
The hubris (?) is just amazing to me. Both on the students and the advisors sides. Like, guys, what are we doing here? This isn't STEM, there's like no difference in the job market between a humanities PhD and a BA.
I've known some English PhDs. They were more focused on self-education rather than external rewards. Many of them were training to become high school teachers eventually, and they knew it. They saw no hurry to begin that career.
> Whoever made the original graphic doesn’t understand the scale and speed of smart high IQ people who can program, and what they can do in a moment when intelligence now on infinite tap using LLMs
Thanks. Eesh. Garry Tan's Twitter feed has certainly been an experience...
Perhaps the key for understanding the wildly credulous optimism on display in that feed is to remember that money 'saved' from the Treasury will likely fund promised tax cuts for the .1%.
Closest I found is this one: ```I'm generally sympathetic to what you're doing. But I hope you will take your time and do it carefully. This isn't just a company," Graham wrote on X during a back-and-forth with Musk on Tuesday.```
While not technically defending it's still pretty disgusting.
It's unbelievable that Paul Graham wouldn't understand how this forever and utterly compromises the Treasury. Nor is it remotely believable that he thinks Musk is capable of doing something like this "carefully" (as if there's a careful way to illegally tamper with the Treasury's innards).
Seems like he is just tweeting so that later he can say "Well I tried; I told him to be careful". Doesn't look smart tho.
Btw, afaik Paul G isn't CEO of YC, and hasn't had an "active leadership role" since 2014.
Fascists always are, that's how they gain enough support to gain power. Hitler had a lot of friends in the capitalist class, and the socialists and communists were the first to go under his rule.
How can anybody here like or admire musk (yeah, very small m) these days is beyond me. I would be properly ashamed to drive tesla, and have 0 sympathies to numerous keyed owners, its not like he became fascist overnight if you actually listened to him. I would stop giving any money to company doing any work with his companies.
Heck, there may soon be some new badges 'musk-free' as some sort of moral badge for businesses to attract and retain young customers.
Based on the scale and impact of fraudulent results, I wonder if some form of LLM based approach with supervised fine-tuning could help highlight the actually useful research.
Papers are typically weighted by citations, but in the case of fraud, citations can be misleading. Perhaps there's a way to embed all the known alzheimer's research, then finetune the embeddings using negative labels for known fraudulent studies.
The the resulting embedding space (depending on how its constructed; perhaps with a citation graph under the hood?) might be one way to reweight the existing literature?
Highlighting research that's useful is probably too difficult, but highlighting research that definitely isn't should be well within the bounds of the best existing reasoning LLMs.
There are a lot of common patterns in papers that you learn to look for after a while, and it's now absolutely within reach of automation. For example, papers whose conclusions section doesn't match the conclusions in the abstract are a common problem, likewise papers that contain fake citations (the cited document doesn't support the claim, or has nothing to do with the claim, or sometimes is even retracted).
For that matter you can get a long way with pre-neural NLP. This database tracks suspicious papers using hand-crafted heuristics:
What you'll quickly realize though is that none of this matters. Detection happens every day, the problem is that the people who run grant-funded science itself just don't care. At all. Not even a little bit. You can literally send them 100% bulletproof evidence of scientific fraud and they'll just ignore it unless it goes viral and the sort of media they like to read begins asking questions. If it merely goes viral on social media, or if it gets picked up by FOX News or whatever, then they'll not only ignore it but double down and defend or even reward the perpetrators.
I think you misunderstand my proposal. I am not describing a fraud classifier.
I am describing fine-tuning an embedding space based on papers with known fraud.
The content of the fraudelent paper which includes information such as authorship and other citations can be made to exist on an embedding space.
Supervised fine-tuning on labels will alter the shape of that embedding space.
The resulting embeddings that would come from such a fine-tuned model would generate different clusterings of papers than what you'd get than if you did not have the labeled fraudelent data at all.
Me neither. But it’s very much in keeping with other seriously-intended suggestions I’ve heard. Optimism is fine until it becomes just dreaming and wishing out loud.
Sorry, I guess you were not being sarcastic. LLM's are good at vocabulary and syntax, but not good at content (because nothing in their architecture is designed to do that). Since the kind of article we're looking for would be fine if written exactly the same, but it was true, an LLM is not a good match for finding this.
Now there might be algorithms that could, for example automatically checking for photo doctoring or reuse of previously used images that are not attributed. These sorts of things would also not be an LLM's forte.
My apologies again, it's just that LLMs are the subject of so much hype nowadays that I genuinely thought you might be saying this in jest.
LLMs are good a producing embeddings which are latent representations of the content in the text. That content for research papers includes things like authorship, research directions, and citations to other papers.
When you fine-tune a model that generates such embeddings with a labeled dataset representing fraud (consisting of say 1000s of samples), the resulting model will produce different embeddings which can be clustered.
The clusterings will be different between the model with the fraudulent information and without the fraudulent information.
Now using this embedding generation model, you (may) have a way to discern what truly significant research looks like, versus research that has been tainted from excess regurgitation of untrustworthy data.
Hopefully the Deepseek news hasn't obscured the discoveries from this space mission. 14/20 amino acids found in life on Earth were recovered from the samples from that mission. 33 amino acids overall. Unlike with terrestrial amino acids found in living things, there are equal number of left and right handed amino acids.
This has big implications for understanding the rise of life in our solar system, and perhaps the potential rise of life in the cosmos.
> big implications for understanding the rise of life
That's quite a stretch. We've known since the 1950s that amino acids form spontaneously in a hydrogen-rich environment with ultraviolet light exposure. Which matches both the protoplanetary disk and the early Earth. This is just physical confirmation of what the blackboard nerds have been expecting for decades.
This was actually the year I was born. Also very auspicious if you're Chinese. I'd be the most asian white supremacist north of Tila Tequila if this was actually a coded Neo Nazi user handle.
(Also, don't look up Tila Tequila Nazi's you won't like what you find)
The presumption of innocence is a legal principle that every person accused of any crime is considered innocent until proven guilty.
Under the presumption of innocence, the legal burden of proof is thus on the prosecution, which must present compelling evidence to the trier of fact (a judge or a jury). If the prosecution does not prove the charges true, then the person is acquitted of the charges.
> The presumption of innocence is a legal principle that every person accused of any crime is considered innocent until proven guilty.
That should be "considered innocent by the legal system". People are still free to come to their own conclusions--and act on them--even without a jury rendering a verdict.
Rather famously, for example, OJ Simpson was acquitted by a jury of murdering his wife. But most people these days would agree with the statement that he murdered his wife.
It is also the case that prosecutors need to decide both the probability of conviction, the effort needed to do so and whether likely conviction on other serious charges are sufficient for the people to feel that justice has been done.
My understanding is they never brought the charges in the first place. The supposed online hitman and the victim were both FBI informants. They never filed any charges because it was clearly entrapment and no one was ever in any danger.
The prosecutors later used that evidence as support for their sentencing request after Ross was convicted of only non-violent offenses, which has a much lower standard of evidence. The allegations of murder-for-hire were never tested at trial. They may have evaporated under cross-examination by a competent defense. Our system of justice holds that Ross is innocent of those allegations unless convicted at trial.
You don't become guilty or innocent based on legal proceedings. The point of a case is to establish guilt for the purpose of punishment.
But an innocent person doesn't become "guilty" even when the evidence shows that in the court. A guilty person remains guilty whether a court can prove it or not.
He should be considered innocent by the courts - and he was (innocent of the murder for hire charges, I mean). In the public we aren't obligated to follow the same standards of evidence as the courts. I think he almost certainly did pay to have those people killed, and that can shape my opinion of him.
That's perfectly reasonable - but I don't think it should really have a bearing on whether he should be pardoned. That is not exactly a matter of the courts (by definition), but I think as an official public act it should be subject to the presumption of innocence as well.
What makes you think I support those people being locked up either? Also, afaik Ulbricht didn't sell drugs himself, he simply provided an unmoderated marketplace.
If the law is unethical then you may be pushed to do "bad" things. For example if you are a Jew living with family in a Nazi Germany and someone know your secret and he feels he need to disclose it to the authorities then you may consider... murdering him. Would you really be a bad guy?
My kneejerk response was to point to the incoming administration, but the fact Stargate has been in the works for more than a year now says to me it's because of tax credits.
I don't think the UK is allergic to industry, it just got worse and worse at manufacturing relative to other countries after WW2.
Just look at the UK's automobile industry... terrible quality.. terrible reliability, particularly in electrical components until the whole thing collapsed.
The general attitude that manufacturing is only for people who suck at school is the driving force behind this decline. You are indeed left with mostly not-that-bright guys who don't even consider themselves skilled workers, and definitely don't go the extra mile to produce quality stuff. A guy half-assing his work earns just as much as the guy who puts his heart to it, that's the harsh reality of modern industrial work. If anything the guy who cares too much is ridiculed and considered weird.
The exceptional craftsmen still exist but they mostly work for themselves, for obvious reasons. They really don't want to be "managed" and bossed around like cattle.
To get an academic position, you need to have a star advisor either for your degree or postdoc.
reply