Do you agree with rms?
It does understand semantic meaning, that's what embeddings are
Even if what he says is true, LLMs are still very useful, also for specific and even medical applications. As long as you know how to use them of course. Models like deepseek are free software and can run on your own hardware, as long as your hardware is fast enough. You could achieve similar levels of results as openai's externally hosted solutions with the right hardware and only free software.
>>106934104Yes, seems like an entirely accurate take.
No. I am writing a book and had NotebookLM analyze it and generate an overview podcast. It had a very deep understanding of my work on many levels and also generated debates between hosts arguing sides of various aspects of the subject matter. It is beyond impressive. Gemini chat is retarded, which makes me think they intentionally limit it in that context. Older versions of Microsoft Copilot could have incredible conversations on philosophy, and do spontaneous poetry riffing jams together.
>>106934104no shit, tech is not there yet, at the level of intelligence that humans exhibit, general intelligence. there's a gigantic rift between human understanding and AI algos of today. mostly a tech and probably theory issue. and clearly hardware issue.
>>106934156
>>106934163They're useful if you want to produce slop.
>>106934104LLMs are only useful for producing porn with four fingers. However, people are already frothing at the mouth for them, because they allow them to generate pizza. This has already toppled the economic system and will continue to bear consequences.
>>106934104I agree that it is not real intelligence, but that doesn't mean we shouldn't call it AI. This is the word we have for it, and seething about it is pointless. It's often used for marketing, but that doesn't matter because this is the word for it now. A lot of words in our language are inaccurate, but we still have no issues communicating with them.You can generate crap with LLMs, but only a fool would deny their incredible usefulness.I agree we should be able to run them on our own computers. On the other hand, I really really don't want GPL-style cuckoldry to infect the LLM model space. The Chinese have been incredibly based releasing models with REAL FREEDOM, not GPL """""freedom""""". Certainly, I should have the right to modify it. I also should have the right to offer it as a product to other users without releasing my modifications. It is just a program running on my own computer.
>>106934104it's retarded to even try to define intelligence, the only thing that matter is the output.it is useful?it is better than other method?can we get shit done with it?if you answer yes to these questions then call it whatever the fuck you want, that's a product/thing/robot/whatever_the_fuck that we'll use, end of the discussion.phylosophy is pure brainrot, these dudes are wasting so much time and energy trying to define things no body gives a fuck about instead of getting shit done, just like rms when he stopped writing code and free software adoption stalled over night.
>>106934523>I agree that it is not real intelligence,What is real intelligence? What is the simplest task you can think of that requires real intelligence instead of fake intelligence?
>>106934523Even before LLMs, the term "AI" has already been diluted to being completely meaningless.It's overly broad, and has no meaning whatsoever anymore.Same thing happened with Godwin's Law and "Nazi" by the way.Overused terms shouldn't be used because their overuse renders them meaningless, and words are supposed have meaning.
I don't think it really matters apart from the general annoyance of seeing the term AI everywhere. It's like the word art. Are videogames art? Who cares.
>>106934601You are the most retarded nigger I've seen in years
>>106934104yeah except its more complicated than that.people always need a real person to blame.saying "the robot did it" is not enough.the one who shoulders responsibility for enough damage will probably get executed by mobs.
>>106934104Over the years I have developed the instinct to simply ignore RMS whenever he spergs out, but this time he has a point.Unfortunately, it's such a glaringly obvious point it barely merits any sort of praise.
>>106934104>ChatGPT doesn't understand anything because this isn't an intelligence and it isn't intelligence because it doesn't understand anything He is saying nothing.
>>106935085>He is saying nothing.Has he ever said anything?
>>106934104ChatGPT disagrees
>>106934104No, that take is retarded. LLMs are already more intelligent than most people. You can ask them a question in all mayor languages and the answer is way better than the majority of the population of the planet. For example if I ask most people living in Africa "Do you agree with rms?". Most of them are going to make stupid noises because they don't speak English, and if I bother translating all to them, they are just going to give me a really stupid answer, because their average IQ is 70.I just uploaded the image of this post to the AI asked it: What do you think about the opinion on the image?Here is what the AI had to say about it:> Many AI researchers argue that while LLMs may not "understand" in the way a human consciousness does, they exhibit useful intelligence by effectively performing tasks that require intelligence (like writing, summarizing, coding, and reasoning). For practical purposes, their output simulates understanding to a degree that makes them incredibly valuable. The counter-argument is that "understanding" might be an emergent property of these complex models, or that the utility of the function outweighs the philosophical definition of the form.
>>106936248>LLMs are already more intelligent than most peopleNiggers aren't people.
>>106934104Yez
>>106934104>crap generatorhow can he complain about LLMs when emacs has M-x doctor?
>these are really dangerous immigrants Stallman? more like Stalinman
