it has been exposed to a breadth of surgical examples
So… machine learning, and a misleading article.
I was gonna ask, “what kind of AI”
We really need to stop using the catch-all term AI, because it’s designed to let OpenAI and Anthropic take credit for everything ML and NN-related, and then even some.
It is AI but not a GPT/LLM model. I also hate the current usage of the word.
Bingo. Too many people hear “AI” and think it means “AGI” (like Skynet) when really it means plain old machine learning like we’ve been doing for decades.
Likely multi-tech system, could even be part expert system AI as they’re often used in the medical field. They haven’t really explained much about the underlying tech they’re using here, except that it was trained on expert data.
Whereas, I think if they could sell this exclusively as a LLM success story it would be mentioned by acronym in the release several times to gain media buzz for the research/publication.
machine learning
Uh… What exactly do you think LLMs are? Christ, people will jump through any hoop to delineate “thing I hate” from “thing I hate but is too useful so I must find a way that it’s different so I can like it”.
Machine learning, including LLMs, are a technology. They are a tool, and can be used for good or ill. All this hatred towards the technology, instead of the salesmen burning our world over it, baffles me.
Do you know what I used AI for the other day? I wanted to try a new store so I booted up my local AI and I had it search for whether the owners have donated to anyone or anything I’d consider a problem. It looked through news stories, the website and wayback machine, public FEC filings, and on and on. In a matter of minutes I had a pretty comprehensive report and was good to go about my day, decision on relatively sound data in hand. Without AI, I’d have spent several times as long and not found half as much… Or, I just wouldn’t have checked at all.
It’s a technology, and its useful. It saved a man’s sight, here, and it’s researching new compounds to save more lives. It’s also being used for war, and crime, and a lot of research was done unethically. But it’s okay to like the good things it’s used for, because a lot of people are using it to help.
The point is, they are using the functional specialized machine learning systems, as a way to justify building thousand of data centers to build LLMs.
I don’t disagree with this as a statement of the strategy the broligarchs are employing (that’s the “burning our world” bit), but my point is purely about the technology itself, and how batshit insane it is to try and maintain one kind of ML is not really ML because icky. All the “AI” that people hate on so blindly is a lot more than image and text generation, but it’s all Machine Learning.
What exactly do YOU think LLMs are?
Machine Learning is a kind of AI. LLMs are but one kind of Machine Learning Model. All LLMs are Machine Learning models, but not all Machine Learning models are LLMs. LLMs are one kind of AI, but not all AI is an LLM.
They’re different words for a reason. If they were all the same thing, we wouldn’t give them different names. It’s completely possible and understandable that someone would find one acceptable but not the other.
Sources:
https://en.wikipedia.org/wiki/Machine_learning
https://en.wikipedia.org/wiki/Large_language_modelThere’s nothing quite so uniquely irritating as someone who holds a strong opinion on something about which they know very little.
I believe you’ll find I said
Machine learning, including LLMs, are a technology.
Because LLMs are a subset of ML models. I never said all ML is LLMs, but the person I’m responding to doesn’t seem to understand that LLMs are trained via Machine Learning. Thanks for the misplaced condescension though, looks real good on you.
But did you double check what it generated about the sources? These LLMs are vastly prone to lying about things. Which is why I feel like they’re nearly completely useless.
If I have to treat everything it says as a lie, especially if it won’t source or substantiate its claims, then what good is it? If I have to manually verify everything, I may as well look it up myself.
Ah, but that’s just it. You can make it source and cite its claims, with deeplinks, and you simply don’t believe anything it said that wasn’t supported by the sources that it collected and presented to you.
But Jesus Christ my dude, the article we’re talking about is about this technology, machine learning, saving a man’s sight during surgery. And what are people doing? Saying how much they hate “AI”, but not this because it’s machine learning which is somehow different than AI 🌈
I doubt the surgeon would say it’s useless. I doubt the guy who can still see would say it’s useless.
Blows my fuckin’ mind, man.
LLM is
trained on a vast amount of text for natural language processing tasks, especially language generation
So… not related to the kind of AI mentioned in the article. You’re the one who talked about LLMs and I don’t know why.
I’m going to respond again here so you see it. You copied some of Wikipedia, congratulations, but you conveniently left out the bit that proves you wrong:
LLMs are Neural Networks. What does Wikipedia have to say about neural networks? Why, they’re part of the field of Machine Learning!
Because it’s machine learning. That’s the premise of the very first sentence I wrote. Are you sure you haven’t surrendered all your thinking to the machine?
What exactly do you think LLMs are
LLM and ML are very different.
Are you sure you haven’t surrendered all your thinking to the machine?
I don’t use AI because it sucks.
LLM and ML are very different.
No, they’re not. Any line you try to draw between them is a lie you’ve convinced yourself of. There’s no arguing with a take this detached from reality, though.
Edit: Okay smartarse, go ahead and explain how LLMs are not products of machine learning technology. Explain how these two things are unrelated. You’re going to have to actually learn about training transformer layers to make any kind of argument, so you’re going to pretty quickly realise you’re entirely fucking wrong, and then you can scurry off to not acknowledge it here.
Be careful of UCL press releases. This institute is quite the media whore and they massively oversell themselves.
How so exactly?
Probably in the way as every single online source/influencer/social media user.
