Has anyone else noticed Yann LeCun's daily LinkedIn/Twitter content bashing chatGPT and all LLMs? I understand that they have limitations, but can't help think that he's feeling salty for Zuck bashing him for missing out on the chatBot race altogether, especially with Google and Microsoft and even Quora now having functional solutions, among many more companies. Especially with all the news about OG Zuck returning and slashing jobs left and right. Yann might be a VP but still at risk of being let go? TC 180
The race hasn't even started yet. Cannot be salty and good to remind people that chatgpt is often confidently wrong, just like much if blind lol.
What are you people talking about. Yann leCun is definitely an authoritative figure in the AI community and his words should be taken seriously. All he is saying is nothing new. We have been echoing same concern for years. Most AI people knows the he saying the all the right things and just did not have the power and the guts to say out loud
OP is just saying that he is bad mouthing LLMs constantly since ChatGPT got attention. I’ve seen him defend Meta to the death when they were going through the misinformation scandals as well. Say what you will but you should not be listening to him to get an unbiased view.
I would not and didn't listen to him when he was depending Meta on the issue of privacy thats not his expertise. But on LLMs in particular, thats literally all he has done for the past 30 years. Vested interest or not, he is talking about real issues. Is he particularly salty now probably yes but that does not negate what he says.
Don't get the big deal with LLMs making stuff up. Minority Report showed how to resolve this a generation ago. Run 3 LLMs and check any double check any disagreements. Think Meta cannot afford the investment to do this.
How are 3 llms going to help when you have one corpus to train on. Also in the age of misinfo I think there will be attacks that subvert training data and this will make llms less reliable. Kind of like Google and social networks became less useful over time for finding reliable info.
Each LLM can be built differently although fed the same data.
Yes it’s possible that he is salty but he also has more context on AI that at least 99.9% of people on blind and will not want to put his credibility on the line by spewing nonsense. So I think it’s useful to pay attention to his comments.
Do you think opnAI and Microsoft and Google have put their credibility on the line if their chatbot gives you a wrong answer? Another case of researchers waiting too long to deliver a perfect solution and losing out to competitiors who released quick and captured market share. Heck, whatever happened to Zuck's move fast and break things?
Meta can't afford to break things anymore. G/MS have some leverage, so they are using it.
Meta will copy it eventually, zuck is busy looking at twitter blue for now
Classic case of being good at research and academia, while being terrible at product. Hats off to these AI researchers for their contribution to the field, but it doesn't always translate to building products that excite the masses. Hopefully these researchers are smart enough to recognize that they're not good at building products and let others do a better job.
I think he has a lot of good and valid points, I just don't know why he has to post the same rant every day
Exactly! Definitely feels it's coming from a place of hurt
Yeah but he's right. It just says stuff that sounds good with no idea if it's right or not, which is your average blind post.
None of these well known ai people have really done that much past what they got famous for. Another is Andrew ng. Very good about talking about ml, but no earth shattering innovations
True having one breakthrough is different from keeping up with industry trends and innovating along the direction the competition is moving. Making smart bets requires business acumen, which is not what these researchers are really known for having
LMAO. Yann and Geoffrey Hinton are the pioneer of deep learning. His achievements are way more influential than Andrew Ng.