- cross-posted to:
- news@lemmy.world
- cross-posted to:
- news@lemmy.world
Americans’ credit card debt levels have just notched a new, but undesirable, milestone: For the first time ever, they’ve surpassed $1 trillion, according to data released Tuesday by the Federal Reserve Bank of New York.
Just to be fair, chatgpt isn’t an artificial general intelligence.
GPT4 absolutely is the foundational block of the first AGI.
I’ve been working with it exhaustively since it was released. I’ve built frameworks around it to give it a memory and other supplements. I have ZERO doubt that this is the first AGI, or at least it’s the engine that powers real AGI.
If you haven’t read this paper, you should: https://arxiv.org/abs/2303.12712
That was researchers using gpt4 before it had guardrails put on its behavior, but also without any supplements to it’s functionally.
Hard definition to fit because it all depends on what you define as ‘general’. Its not great but tbh you can ask it a question about anything and it will give an answer, so I’d argue that’s general enough
It’s missing intelligence though.
ChatGPT uses machine learning to predict a sequence of words, but there’s no thought or understanding of the words.
Removed by mod
Another fairly soft definition, which to debate I gotta ask what is ‘intelligence’ to you? Defined as rigouroudly as you care to, cause you’ll find its pretty damn hard to get anywhere fundamental.
My point with this is usually if something has no definition, its probably not that good to use as a definition for other things, like AI
Realistically the only way you could consider AI not intelligent is of you specifically require aspects of humanity within it, as such if aliens existed and didn’t have anything analogous to “thought” would they then not be intelligent?
From google intelligence is
ChatGPT is not capable of active learning so no.
If aliens can then yes.
I’m not even talking about human intelligence, most animals have some level of intelligence that chatGPT doesnt have.
ChatGPT is just very good at appearing intelligent.
But of course the AI is trainable, hence how it got trained.
Actively learning doesn’t come up in the definition, but it being able to respond to multiple comments with context shows it does actively learn and understand the topic at hand.
Its for sure much more sterilized than natural intelligence, but tbh the main reason it doesn’t train on its input data is because it would turn to junk fairly quick with the mess of messages it must get.
I’m not sure I get your ‘appearing intelligent’ comment, either something does or doesn’t actively learn under your definition, so where does the appearance come from? Unless you mean because people are undecided of conversation is active or not, which would put it on the fence (again under the active definition)