AI Progress

[SaucedOnEm]

Well, intelligence is defined by the ability to adapt to an environment, so the absence of common sense in my opinion also implies a lack of intelligence...

I would argue someone who only has common sense is less intelligent than someone who has the skills to survive with less

 

[Preacher]

I would argue someone who only has common sense is less intelligent than someone who has the skills to survive with less

The ability to adapt includes (but is not limited to) the ability to survive.. Anyway, common sense and intelligence go hand in hand. One is not possible without the other...

 

[j.f.🔎]

Right, replying to multiple things can make things a bit confusing for me sometimes
The forced labor in Africa that tech companies profit from

See? Now I'm talking about AI in your thread. (African Indenture)

 

[SaucedOnEm]

The ability to adapt includes (but is not limited to) the ability to survive.. Anyway, common sense and intelligence go hand in hand. One is not possible without the other...

You can lack social ability to a strong sense and still be extremely intelligent. That's more so what I mean when I said that. Regardless, I'm trying to point at the fact that these machines are superhuman in multiple areas already. Once continuous memory is figured out, I believe things will start changing a lot faster

 

[thebully08$]

Looking forward to sex robots, until they rip my dick off and beat me to death with it.

 

[AdaBull]

I dont understand the nitty gritty.
No i dont consider myself an ai expert.
I understand the process of training these things go through. That's the self taught aspect. Understanding them as a concept.
Your analogy of it being a really big excel sheet is a good explanation of what i meant when i said experts dont fully understand.

My comment about you not understanding is a big foot in mouth moment now. This site has been bringing the worst out in me and that may have had an influence on how i replied to you. My bad.

I dont consider ai to be black magic. Just like i dont consider the human mind to be black magic.

The reasoning shown in the cancer example you gave is kind of my point. Why did it understand the label?
Next token prediction as a whole has a requirement of understanding why the next token is more likely, imo.

I'll acknowledge that the actual practice of using Ai is obviously important in understanding these systems. My ideas are based on ideas from people who actually work/currently work in the field, Geoffrey Hinton, Mo Gawdat, and others. So while i dont have any qualifications or published works centered around Ai, my ideas arent just "black magic"

"This site has been bringing the worst out in me and that may have had an influence on how i replied to you. My bad."
-No harm done

"The reasoning shown in the cancer example you gave is kind of my point. Why did it understand the label?
Next token prediction as a whole has a requirement of understanding why the next token is more likely, imo."
- The example I gave does not use tokens (word vectors etc) it uses filters, it does not see words in the picture, it sees levels of colors (but it depends on the input shape you use, this is where I work on but not for images). So when there is text on the image it sees just different colors shades, then using the filters it sees shapes like circles lines etc (I oversimplify) which would represent letters or partial words for us. And it sees only the link between different colors and shapes. It does not understand what this means, it does not understand that cancer is bad or good nor what good or bad is. The step further is recognizing words which can then be combined with the tokenization process etc this is how 'AI' processes text in a hand written documents for example. So it is a combination of multiple AI networks trained for different purposes.
I don't know if you saw it but recently LeCun said that for making smarter AI we should stop using LLMs as we arrive at a new stagnation, and I agree with him and as we piggyback more and more stuff on them the results get also worse and worse.

PS: sorry for the delay in responding and I find it good that you get interested more into AI than just using it like most people

 

[SaucedOnEm]

"This site has been bringing the worst out in me and that may have had an influence on how i replied to you. My bad."
-No harm done

"The reasoning shown in the cancer example you gave is kind of my point. Why did it understand the label?
Next token prediction as a whole has a requirement of understanding why the next token is more likely, imo."
- The example I gave does not use tokens (word vectors etc) it uses filters, it does not see words in the picture, it sees levels of colors (but it depends on the input shape you use, this is where I work on but not for images). So when there is text on the image it sees just different colors shades, then using the filters it sees shapes like circles lines etc (I oversimplify) which would represent letters or partial words for us. And it sees only the link between different colors and shapes. It does not understand what this means, it does not understand that cancer is bad or good nor what good or bad is. The step further is recognizing words which can then be combined with the tokenization process etc this is how 'AI' processes text in a hand written documents for example. So it is a combination of multiple AI networks trained for different purposes.

I should've guessed that tbh. Thanks for explaining that

I don't know if you saw it but recently LeCun said that for making smarter AI we should stop using LLMs as we arrive at a new stagnation, and I agree with him and as we piggyback more and more stuff on them the results get also worse and worse.

LeCun is a heavy skeptic on this subject. Maybe he's right tho. I just disagree with his timelines due to how much stuff has been accomplished. Ouu, what do you think about gpt 5?
Also, do you think llms have surpassed the average human in knowledge? I heard that we surpassed the point where normal people will notice a difference in the intelligence of the model and now phds are saying that the models are solving grad student level problems. To me that says a lot.

sorry for the delay in responding and I find it good that you get interested more into AI than just using it like most people

No problem, we got lives lmao
And yea, ever since I talked to get 3.5, I've kinda been focused hard on Ai. More in a philosophical sense than technical.

 

[AdaBull]

I should've guessed that tbh. Thanks for explaining that

LeCun is a heavy skeptic on this subject. Maybe he's right tho. I just disagree with his timelines due to how much stuff has been accomplished. Ouu, what do you think about gpt 5?
Also, do you think llms have surpassed the average human in knowledge? I heard that we surpassed the point where normal people will notice a difference in the intelligence of the model and now phds are saying that the models are solving grad student level problems. To me that says a lot.

No problem, we got lives lmao
And yea, ever since I talked to get 3.5, I've kinda been focused hard on Ai. More in a philosophical sense than technical.

Q: What do you think about GPT-5?
I tested it when it first came out but initially preferred GPT-4. I think they’ve improved GPT-5 since then. At the moment, I use Mistral Le Chat, which you could call the “French ChatGPT,” mainly because it offers better transparency about what happens to your data. Gemini is also very good. From what I understand, some of these models share techniques or data sources, so there’s some overlap between them.

---

Q: Do you think large language models (LLMs) have surpassed the average human in knowledge?
That’s like asking whether Wikipedia contains more knowledge than the average person — or whether a dictionary knows more words than we do. So yes, LLMs contain a vast amount of information, but whether that really counts as knowledge is another question.

---

Q: I heard that we’ve passed the point where normal people can tell the difference in intelligence between models, and now even PhDs say they’re solving graduate-level problems.
They can indeed handle complex exercises, but not always reliably. For example, if you ask them to derive or develop a difficult mathematical concept that isn’t well-documented online, they often make mistakes. They also struggle with open-ended theoretical discussions. Even though they’ve improved their ability to cite references through web searches, they still tend to mix or invent sources — which is understandable, given how they work.


For scientific research, I wouldn’t trust them fully. Their real strength lies in synthesizing and reiterating information, or blending styles — that’s why they’re so impressive in the arts. They can do creative “fusions” like Michael Jackson sings Frank Sinatra or Van Gogh meets Picasso painting a horse. They also excel at common knowledge tasks, like “I broke my finger — what should I do?”


However, for more context-heavy questions (for example, “Should I invest in this house?”), the answer depends on many real-world factors — location, bank policies, personal situation, etc. The AI can help structure your thinking, but it can’t make the decision for you.

---

Q: Do you think AI will replace humans?
AI isn’t meant to replace humans — at least not right now. Its main goal is to simplify our lives. Take doctors, for example: they often have only five minutes to see a patient, then spend another fifteen doing paperwork (even with a secretary). If AI could pre-fill medical forms or handle the repetitive admin work, doctors could spend more time actually talking to patients, which would make healthcare more human, not less.


The same logic applies to other jobs. The goal isn’t for AI to steal work or replace artists. It’s to handle boring or repetitive tasks, freeing people to do creative, meaningful, or human-centered work instead. Ideally, AI should help us eliminate tedious factory-line jobs and create more interesting, rewarding roles for everyone.


PS: As you know, AI is great for improving text (as it did here 😉), but the content and ideas remain entirely my own — they reflect my personal views, not those of any language model. But I think it would agree.

 

[AdaBull]

 

[hedgerow]

the thread of dweebs

 

[trifling twat]

AI is wayyy too good. it crushed my dreams of being an animator and an artist.

 

[D.O.A.]

the thread of dweebs

Zoomers have reached the pinnacles of shitposting

 

[Mr. Meat Sled]

AI, this breed of AI anyway, merely makes stupid people even more stupid and lazy. In a world where retards are obsessed with Tikshit, there's is very little creativity and originality left....and it will get worse.
It is comparable to how music has gone downhill since the late 1990s to present day - there's no originality and "musicians" are now just products because they do no musicianing; there're no exceptional music that has been churned over the past 30 years.

We've been hiding all the exceptional new music in places like NPR and PBS. So go out and Support your Local Public Radio station! (Mine is 91.3, WYEP). Find out when the next Austin City Limits program is being aired in your demographic and give er a listen, or a watch.