Imagine telling Yann LeCun what is and isn’t right when it comes to science
honestly LeCun should know better than to argue with a crazy person.
it doesn't matter how right he is, musk will turn everything around and have fun while doing it.
thanks to melon tusk, i don't have to
The Musk likely knows who and what he himself is, even if only in the darkest and most sleepless hours of the night, but on the other hand, his followers eat this shit up like candy. "Survival of the fittest" - caveat: in the current climate, or rather the one from the last few decades - has led to him being put in charge of way more than he should, in the same manner that a cockroach is "fitter" than humans since they will outlast our having caused WWIII (unless we make it to space, which seems increasingly unlikely at this point, at least within any of our current lifetimes).
Anyway, it is important to remember that he does not do this for reason of mere stupidity - he literally gets paid to dish out this kind of shade.
Edit: case in point, the fact that we are discussing this now, and also the title of this post. If Elon had said "I respect you", that would have been the end of the matter right there, but it would not have met his goals (or apparently, ours either).
Who he?
Neural networks, OCR, called one of the "godfathers of deep learning", etc, etc.
He's a brilliant, well known scientist.
I like the sentiment, but there are non-peer reviewed papers that are real science. Politics and funding are real things, and there is a bit of gatekeeping here, which isn't really good IMHO.
Also, reproducibility is a sticky subject, especially with immoral experiments (which can still be the product of science, however unsavory), or experiments for which there are only one apparatus in the world (e.g., some particle physics).
The things you’re describing are not science. This might seem nit picky but the scientific method as we know it today require that peer review and require methods of reproduction. Whether you can reproduce results is a different story.
The entire difference between research and science is whether or not you engage in the process of peer review and review often requires method of replication. So you usually can’t have one without the other. If you aren’t trying to have your paper reviewed by your peers, that’s fine, but that isn’t science.
To address the gatekeeping, I get it. We shouldn’t be using the word to demean people who do valuable research but don’t strictly engage in the scientific process. That’s really not important to do. However we should all be interested in preventing the scientific process from being muddied to include every R&D process under the sun. That’s all research, not science, and we call them separate things for a reason.
I think the word you're looking for is merit, publication which are cited and peer reviewed hold much more merit than those who don't.
Science is a rigorous, systematic endeavor that builds and organizes knowledge in the form of testable explanations and predictions about the world. 1
Nothing in this quote requires external publication. Following the scientific method, publishing, peer reviewing and reproduction can all happen internally in organisation using independent teams. Those private publications hold but a fraction of the merit of publications in recognised journals, but are science nonetheless.
This might seem nit picky but the scientific method as we know it today require that peer review and require methods of reproduction. Whether you can reproduce results is a different story.
Oooh, are we about to have a discussion on whether large portions of the soft sciences across the past several decades fail to be "real" science due to the reproducibility crisis?
Imo both in the twitter are stupid. Like, no way engaging with musk could go well.
The rules and conventions to do science today are quite well known and understood by educated people (including of course Helen Mosque) ... but any rules have exceptions :
Project Manhattan to produce the atomic bomb was secret science : in many countries military will have secret science development. Pharmaceutical companies will do as well.
People in those projects will not have recognition by the wider public but they will have recognition from their group.yes, but even within those "secret groups", there are SOP and conventions of intergrity.
Thanks ! ... if anyone else wants to know :
SOP stands for Standard Operating Procedure. Within secret scientific research groups, SOPs are established guidelines or instructions for carrying out routine operations to ensure consistency, quality, and compliance with regulations. These SOPs help maintain integrity, confidentiality, and efficiency within the research group.
Heck, I can think of a half dozen other examples of things that aren't published and/or can't be reproduced but would be considered science.
If I had an unpublished workbook of Albert Einstein, would I say the work in it "isn't science"?
If I publish a book outlining a hypothesis about the origins of the Big Bang, is it not science because it doesn't have any reproducible experiments?
Is any research that deadends in a uninteresting way that isn't worthy of publication not science?
I like dunking on Elon as much as the next guy, but like, "only things that are published get the title of 'science'" seems like a pretty indefensible take to me...
i agree because what I usually mean when i talk of science is scientific work even if this work doesn't result in proving that an hypothesis is right so that it becomes a scientific theory.
For me the main criteria is to follow the scientific method.If I had an unpublished workbook of Albert Einstein, would I say the work in it “isn’t science”?
I would say it isn't science yet. I'd say once you published it and other people confirmed he was right, then it would be science. Until then it's just research. Stating that it must be right just because Albert Einstein said it is disrespectful to the work of a lot of people, not least of whom is Albert Einstein
She's wrong though, everything following the scientific method is science. The fact that you didn't pay out of your ass to publicize your research doesn't matter. Of course it reaches less people, but that's a separate issue.
Yann LeCun is a dude
With all these "she" talk in this comment section, I was like when did LeCun change gender?
I don't even do anything remotely related to AI, but I know LeCun is a dude.
The scientific method includes peer reviewing.
You don't have to post it on a commercial database, only free one will do. But it needs to be accessible by the world.
Does it require independent peer review though? How do you achieve that
withwithout publication? The predatory publication system is a different point.Edit: fix without
everything following the scientific method is science
I'm fairly certain "report conclusions" is a pretty big deal in the scientific method. Principle of verifiability and all that.
True a lot of science is done in industry and the corporate world and not published to keep it a trade secret. It is still science but not shared.
There's no such thing as a scientific method
The scientific method varies from field to field. In medicine you usually need to proof it by taking a significant amount people. Then create a control group and a testing group. Then test your medicine on the group and give the other placebos.
When you can measure health improvement for one group over the other there is a reasonable amount of proof that the medicine works.
The scientific method has one major goal. Reduce human made errors in science. Humans do not work objectively. Humans always have an bias. Things like reproduceable tests and peer review try to reduce the bias.
I am a bit worried the response to this here is not a unified everyone's an asshole in this screenshot.
Academic publishing is in a very sorry state for a long time by now. A lot of research that is published is not reproducible. A lot of actual research is also in fact never published like that because companies base their products on it and publish those results only as patents.
So just by trying to be smug and oppose the Muskie you show yourself to be an idiot. Well done.
It's worth saying that ml is in a very different position to most of academic publishing.
All of the serious journals are free to publish and fully open access and a significant amount of publication includes enough code that things are mostly replicable. GitHub has done wonders for our field. Also many tech companies use publications as an indication of prestige and go out of their way to publish stuff.
We're still drowning in too many papers and 95% of everything is shit, but that's every field really. Talking to musk on twitter is the not right place for a nuanced discussion about publication.
"Nuance" and "discussion" did not appear to be part of either participant's intent.
Btw his bio is:
- Professor at NYU
- Chief AI Scientist at Meta
- ACM Turing Award Laurete
Yeah he's a legend in CS... Muskrat just further rides himself into being the fool of centuries
Academic Journals frantically spinning up botnets to retweet this
If you science in a lab and no one is around to review it did you make a science?
well it still goes back to the original arguement that it has to be published and reproducible.
else it would be forgotten and re-discovered again at a later stage.
some scientific discoveries of the mordern era were actually discovered by earlier ancient people before mordern science started recording such discoveries.
The difference between science and fucking around is writing it down - Adam Savage
The second, implied part, is that writing it down is for OTHER people to learn from.
So, although I hate the eletist gatekeeping language.. I think I agree more with the professional scientist than I do the professional clown.
Corollary: If it can't be reproduced you've failed to write down something critical.
That is to say: it's fucking around and, if you're good at it, finding out.
There are differences between "experimenting", "research", "analysis" and "science". You can do the first three at your home, scribbling some notes that no one will ever read or know about, but science, in its hard definition, is a methodology that requires the specific dynamics that are expected of the scientific community, where plenty of people check each other's work for faults, blind spots, biases, lazy interpretations and so on.
This is fundamental because everyone, including universally recognized geniuses, do sometimes fuck up. Have you heard of Einstein's famous phrase "God does not play dice with the universe"? This refers to his conviction that the laws of physics were fundamentally deterministic, which was put in question by the early experiments that were opening the way for quantum physics. Einstein found himself at odds with a new generation of physicists that weren't as inflexible as he was on this issue, and whenever there were indications that extremely small particles may behave in a non-deterministic way, Einstein would argue and push for the most hostile interpretation possible, which did lead other physicists to put his interpretations to the test, which did ironically further prove the non-deterministic pillars of quantum physics.
Science is necessarily a social endeavor because it is meant to help us overcome the fact that each individual human is doomed to be, sooner or later, at one specific issue of many, an inflexible idiot.
Your last sentence is pure poetry.
This refers to his conviction that the laws of physics were fundamentally deterministic
What was the bit about quantum mechanics yesterday? "Embrace the chaos and you start to see a pattern."
If we're talking about gatekeeping what is and isn't capital-S Science, I'd really like to know where these "hard definitions" are coming from.
Wikipedia's page for the Scientific Method seems to get it wrong when it describes it as "a general set of principles," the core of which is forming falsifiable hypotheses and testing them... and the details vary from field to field and across different time periods. Sounds like you can do that at home.
The page for Science appears to also contradict the "hard definition" when it describes science as spanning most of human history, long before the modern institutions of formal publication and peer review, and doesn't describe them as mandatory at all. Definitely doable at home, as far as I can tell.
That's not to say that scientific collaboration isn't valuable, btw... I just can't find any basis to support the idea that if it's not published in a formal academic journal, then it's definitely not science, and that science CAN'T happen without the involvement of the institutions.
So like... Where does this "hard definition" that people keep talking about come from, and why doesn't Wikipedia seem to know about it?
It comes from literature I did read over a decade ago, which titles I no longer have, which argued (in a very summarised sense), that science as we know it today is only possible due to the development of social institutions and methodology that have been refined over centuries (and arguably, are currently in an evolving process), in ways that make it fundamentally different (in its workings, its results, how it is envisioned and how well it procures reliable knowledge) from what the natural philosophers of antiquity did, ultimately requiring an ample social system for it to even be viable. You will notice that the majority of attempts to schematize the scientific method include either reporting or publishing, or delegate the task of replicating experiments to third parties.
The page for Science appears to also contradict the “hard definition” when it describes science as spanning most of human history, long before the modern institutions of formal publication and peer review, and doesn’t describe them as mandatory at all. Definitely doable at home, as far as I can tell.
Sure, you can see sketches of what we currently consider science in the historical development of astronomy across Mesopotamia, Egypt and elsewhere (lacking the modern core methodology), or in Newton's writings about alchemy (lacking communication), but most of it was intertwined with mysticism and esotericism. You can use a more lax definition if you want, but I think that in doing so, you're making the concept lose meaning.
That’s not to say that scientific collaboration isn’t valuable, btw… I just can’t find any basis to support the idea that if it’s not published in a formal academic journal, then it’s definitely not science, and that science CAN’T happen without the involvement of the institutions.
150 years ago, the contemporary institutions of formal publication and peer review didn't exist, but equivalent processes were already starting to take form. These contemporary institutions aren't as important as it is that the tasks they fulfill do get done in one way or another.
I've seen published scientific papers that were written by chatgpt, complete with prompts.
okay so there are requirements besides getting accepted for publication
Were they peer reviewed?
Sounds like we need to define "peer". What happens when ChatGPT starts peer reviewing ChatGPT papers?
Couldn't science papers be hosted on a git-platform for review? Instead of costly publishing and the reviewers have to buy it then...
There are open access platform that is more reputable than git, like arxiv or hal.
Plus most conferences, at least in my field, support open access. But unfortunately for some of them, you do need to pay a fee in order to get the article to be open-access.
The prestige of the conference/journal is still the best way to get your article known, so that others can review and built upon your work, as of now.
Seems like a very elitist and gatekeeping perspective, specially considering how closed off the academic world is for the rest of society in some places, never mind expensive to publish. It's also basically saying that if you, say, come up with a groundbreaking hypothesis, that that's not science until you get a research paper out, and that might require mastery that goes beyond the hypothesis.
Sure, this might stop most of the looney theories from being called Science, but it also prevents public access in favor of those with the means and capacity to sustain an ever more complex geocentric model of the fashion of the times, from which any divergent theories must generally part from or involve renown in.
You think the person who made that hypothesis will die bitter and forgotten? Is that the general view of people who are not Scientists by Scientists? They might know what's up, and might not want the gatekeeper to take all the credit, as is often the case in academic circles, and might just feel satisfaction in seeing their hypothesis gratified. They might place more importance in exploring and understanding reality than compensating for personal insecurities. Perhaps it is science itself that might stagnate by stalling until it itself is able to discover these hypothesis under the properly accepted emeritus when they are eventually able to get to it.
Mostly it's just looney theories, but given Musk is involved, I imagine this discussion involves proprietary patents that do have a lot of research involved and under peer review of teams under non-disclosure agreements. Then again, it's Musk, could be mostly looney theories too. But the fact that it involves Musk, the man living off of Nikola Tesla's fame, a man whose demise could have been described to have occurred under the circumstances of a bitter and forgotten end, makes the gatekeeping particularly ironic.
It doesn't need to be published in a scientific journal. Publication in journals is the most streamlined way to go through the process, but you could publish your hypothesis and methodology to a blog and potentially get the same benefits.
Even patents need to be published. Publication is how discoveries are shared and verified.
You would still need to be recognized before someone more recognizable takes it and sticks their name on it the moment they see any validity in it. Plagiarism isn't a myth, and good luck getting recognition even just for a hypothesis without a master and just as a hobbyist.
Academics want a well prepared research paper without evidencing crude freshman mistakes, and by its nature yours might be far cruder than academic standards. Even if you do end up releasing it and if it does by some miracle get acknowledged, it will by its nature take longer and run more risks from a lack of peer review that might discard it due to simple but correctable mistakes while running the risk of getting it plagiarized by someone capable of fixing it up, and no one is going to take a random blog as the proof of a preexisting theory over a research paper with a name with some masters to it that claims the idea was entirely theirs shortly thereafter. And if all you care about is the study of reality and science, why risk the heartbreak of getting personally involved?
Patents don't need to be a full comprehensive research pieces, they just have to be enough to define and identify particular intellectual property.
It's free to publish.
never mind expensive to publish.
Academic world is very not happy about it either. Academic world hates journals publishing corporations.
See lawsuits against ResearchGate, lawsuits against Sci-Hub and lawsuits against many students and academics that shared scientific papers.
This is why the machine learning community will go through ArXiv for pretty much everything. We value open and honest communication and abhor knowledge being locked down. This is why he views things this way. Because he's involved in a community that values real science.
ArXiv is free and all modern science should be open. There were reasons for publications in the past, since knowledge dissemination was hard, and they facilitated it. Now the publications just gatekeep.
ArXiv, bioRxiv and cyberleninka.
EDIT: Today I Learned that cyberleninka now has English version of it at cyberleninka.org
Science is a specific social activity that humans engage in (emphasis on social). Science is not the same as fact-finding, or philosophizing, or reasoning. It’s a particular method of peer review that generates shared public knowledge.
Again, “science” is something humans do together. Experimenting, investigating, puzzling, hypothesizing, intuiting, discovering, and knowing are all things you can do alone.
Science is a particular method of peer review...?
This thread prompted me to revisit what I think "science" means, and I've been through a number of different Wikipedia pages, dictionary definitions, etc. but that inquiry just reinforced that this "science == participation in the institutions/communities of science" idea just doesn't seem to hold up.
Where does this idea come from? I keep seeing this "science is this very particular thing, it's not just forming falsifiable hypotheses and then testing them," but then when I look it up, the sources I find say exactly the opposite.
EDIT: To respond, backwards, to the edit below, I guess...? That's not really a gotcha, and not really what I was saying, lol. Please read the whole thread.
Fuck, I really hate to agree with Elon on anything, but that is a ridiculous argument. LeCun must also really believe that trees only fall in the woods when someone is around to see it happen.
Science is strictly a social activity. You can’t have a social activity without the social component.
Again, fact-finding is not the same as science.
Yeah, they're both pretty wildly off base. Publishing papers that are vetted and used as a foundation for other work is science. Also, sorry, but developing advancements behind closed doors is still science. Oppenheimer's secret research for the government is pretty fucking foundational. Thomas Edison wasn't interested in sharing his ideas, but rather in selling them. Everyone remembers him.
This argument reads like two people having an ego trip past each other.
You were correct in the first, but the things you're describing are product research and development.
All super important, but not exactly what I call science as a socially beneficial activity humans do specifically to learn the truths of the universe.
Science is just the process of testing things in the world in a reproducible way.
LeCun’s argument is good career advice (you only get credit for what others know you did), but it’s not factual correct.
Also how transparently published and reproduced was Pfizer's vaccine trials, considering a judge had to force the contents released, yet it was science right away. You can't have cake and eat too.
how about you figure out how to make a gas pedal that doesn't try to kill people before you talk shit?
They'll have to publish the science to a public board, which doesn't make him money.
There's private company r&d science and military science as well, even though those aren't academic science with it's peer review and publication.
I agree that for it to be science it needs to be reproducible, but obviously publication could happen internally. It just ends up as science that no one else can benefit from, which is contrary to what most scientists actually want.
Musk is just an ass who doesn’t want to share his toys.
I'm too lazy to edit it, but you get the idea:
Is the Tweet still up or did it not sit right with Elmo so he took it down?
Yann LeCun is not a good person, but at least he's significantly smarter than Elon
I wish he was even smarter so that he would choose not to play chess with a pigeon.
I really don't mind him making Musk look like an idiot
I know nothing about him, what did he do?
For starters, he’s both French and works for Meta. 😏
But beyond that, from a professional perspective this guy has pioneered so much stuff that we take for granted today; like check scanning and OCR. I mean he is touted as one of the godfathers of ai.
He bought a blue checkmark.
They both come across as pompous asses in this one.
If you develop a product in secret, take it to market, and make a fortune off it, far more people will know your name than almost any scientist.
Spot on
They've been feuding for a while so I'd be surprised if they both didn't try to one-up the other insufferably.
Sure, but then you aren't considered a scientist nor your work science.
I think most people outside the scientific community would consider that science.
Every time I see Elon's picture in one of these posts, my brain tells me he's wearing an Ant Man costume.
it's worse than that, he's in a WoW costume or something similar.
Well somebody is about to find his Twitter account closed for a spurious reason.
But my freeze peach!
The best emoji for defending your fragile self esteem: 🤣
For depicting your interlocutor as the soyjak and yourself as the chad, it's almost as good as 🤡
I'd say someone should get Musk some burn cream, but he can afford to get it himself.
I've got some Icy Hot I could maybe give him. I don't use it.
also some OC spray for his face.
Don't let love or hatred blind you to the truth
Now gather round chillun, sometimes, I say sometimes, you know, sometimes ... one should shut up and be rich.
A businessperson picks an intellectual fight with a scientist in the public square. We humbly suggest for due consideration, to 'take under advisement', or 'run through the handlers', that perhaps, possibly, although we could be wrong, or locked onto the wrong VOR while navigating this latest PR disaster, but just maybe, the global reputational maximum (don't even need gradient descent for this one brah), is to be quiet with ones insecurities, rather than ham-fistedly operate the mouth, removing all doubt, and thus broadcasting the spectacle to the internet (a series of tubes), which will still hold said incident in its memory banks longer than any wetware.
Plus, as an added insult + injury bonus, AI models will slurp this incident into their learning like a line of Bon Ami snorted off a 3-day old third-pan of 'temalees' in a gas station 'buffet' (please avoid the sushi) on the way to nowhere.
All nibbles and bytes are immortal now and forever more!
DWord to your motherboard.
I like the cut of your jibes, sir or ma'am.
I've read plenty of times about bullshit published papers that disprove it must be correct and reproducable to get published.
Edit: Where did I claim it was or wasn't science? I'm pointing out the statement that "to be published it must be checked for correctness" simply isn't true.
Some published papers are not reproducible. All unpublished papers are not reproducible. You’re creating a dangerously wrong equivalence.
I feel like I'm missing something here so I'll be the devil's advocate, why can't unpublished papers be reproducible? Multiple teams could independently be verifying hypotheses and results under the same organisation, adhere to the same standard but never publish, that would still be science no? Not doing humanity any favours, but science nonetheless.
They never claimed that all published papers are science. They said if it's not published, it's definitely not science.
Then goes on to say that to qualify as science, here are some things it needs to have done. They don't say that it has to be current and reproducible to be published, they say that if it's not published then it's not correct and reproducible. Those are different claims.
Never thought I’d say this but I’m on Elon’s side this time. If you’re seeking new information about the world and generally following the scientific method you’re doing science.
Is it good science? That’s a different question.
By that logic you’d say that dianetics counts as science? Lmao
And you’d say that some brilliant student project isn’t. Publication is not the definition of science. Does the work have to be correct to be science? No. Most science is eventually shown to be incorrect in some way. It’s the search for answers that defines science not some journal that openly exploits the people who are searching.
What the fuck is LeCum thinking about? I work in academia and I couldn't give a shit about being remembered, I just want to live to fight another day like the next guy.
This feels like billionaire banter.
As a trans person, I fucking hate Elon Musk. But as a chaos magician, I gotta take his side on this debate. Mainstream scientific publications won't take research that's sufficiently "silly" or "esoteric", and magic falls under these even when the scientific method is used. Now, many people have published works on chaos magic such as books, but not so much with publishing in reputable scientific journals.
Science, real science like Elon is describing, happens when you write stuff down. "Published science" is where the glamor is but that's, quite obviously, not what Elon was talking about.
So sad to see bitter people lash out at the successful. (projection is also a classic trait)
Science requires peer review, so just keeping it all private isn't doing much for the scientific community as a whole
Published science requires peer review. Big difference.
"good for the community" isn't relevant to being science.
