There is no real AI. just algorithm based learning that's really good at processing data. It's interesting, but not that interesting as it's just a more refined version of concepts that are already well known.
The idea of creating actual AI, with original thoughts, opinions and individuality, is still a complete pipe dream and no one is anywhere close to accomplishing that. Consciousness is a big blind spot in science, we don't understand the nature of it at all, we can only observe the results. How do you program something that we don't even understand?
Just last night, I was catching up on BlackList and watched an episode where a guy programmed a robot sex doll to murder their "clients". Naturally, I Googled this to see how close to reality this fiction was.
Short answer - pretty darn close. I'd link something, but I'm sure that would cross lines. The curious can tarnish their own browser history. It's kind of crazy and a bit creepy. Or a lot creepy depending on your personal line of creepiness.
I've been testing out ChatGPT this morning and it's pretty insane. I threw a couple questions about web development code that I would come across occasionally and it spit out the code, with comments on how to use it in a couple seconds. Might not need to search Google / Stack Overflow anymore. This is WAY quicker.
My wife is a high school English teacher and I showed her. She put in a question she would ask her students and she said the response was spot on.
She then asked another question about a more obscure Canadian author and it didn't know what to do with it. So it's not perfect yet.
Teachers already have to change up how they're assigning work and this is going to make things even more complicated.
Last edited by KTrain; 12-02-2022 at 09:45 AM.
The Following User Says Thank You to KTrain For This Useful Post:
I guess it depends on the definition - but isn't this just google with narrative responses instead of links?
I guess is google search considered AI?
It's not just copy-pasting content from the web. It's synthesizing answers based upon a totality of content that it has scanned. The prof in that thread I listed ran the responses thorough plagiarism software and it all checked out.
It also remembers earlier parts of the conversation and makes use of that context.
__________________
"If stupidity got us into this mess, then why can't it get us out?"
I think people are getting caught up on the term "AI", but you're right. The term AI is being thrown around for all these things and it's not truly AI.
That said, it's still insanely powerful and fast. Faster than a Google search and can't currently be gamed by people with good SEO. It also seems to produce "original" work when you try to get creative.
Dismissing stuff like this because it's not 'real AI' is like dismissing a powerful car because it's not using real horsepower. People can call it whatever they want. The important thing is what it can do.
__________________
"If stupidity got us into this mess, then why can't it get us out?"
The Following 3 Users Say Thank You to JohnnyB For This Useful Post:
These large language models are certainly impressive at the moment. I do wonder how much further they can go - you can only increase the number of parameters so much before you hit physical limits. There is a whole category of human work which is going to become redundant in the next few years. Hopefully, that pushes companies to start giving more important work to human beings and let AI take care of the report writing and HR memos.
Alexsei supports legal teams with high-quality research powered by artificial intelligence. Effortlessly turn any research question into a comprehensive memo.
I guess it depends on the definition - but isn't this just google with narrative responses instead of links?
I guess is google search considered AI?
What's messed up is this thing can't access the internet right now. So you're totally right, most people will see it as a cleaner Google, but what it's doing is so much more impressive.
KTrain's example is important not because it got the answer but in how it delivered it. If I have a coding issue, I might go to a site like stackoverflow and search to see if somebody else has had my issue. If they have, it won't be identical, but I can probably adjust mentally to see what's going on. The answers are often full of needlessly complex language and superfluous detail.
This chat method gives an answer immediately, but goes beyond that in explaining it in the clearest way I've ever seen, as well as what's going on and how to implement the fix.
I saw somebody say a few days ago that this could be the first time Google as a search platform has ever been challenged. Many make a huge deal about how AI can get stuff wrong (which is correct), while also neglecting that Google results can easily serve up incorrect information, or information manipulated by ad spending.
One thing I've noticed after reading a lot of these is the, uhm, tone and style of it. I'm not sure if it's just that I am aware I am reading AI output, but it has sort of an uncanny prose quality to it. It just seems a little off, but I can't really say why.