i know people like that
And THAT, my friends, is the issue.
Itās the same with people. Some people I trust implicitlyā¦ which I have learned over time to be accurate. Some other people I ātrust but verifyāā¦ and the list goes on.
IBM Watson is doing this very thing, but it is thus far limited to purely data-based pursuits (you can argue whether weather forecasting is āfactualā, but it is indeed data-based).
Just because ChatGPT says it doesnāt mean itās the right answer, or even the only answer.
Eventually one will be asked, āIs there a God?ā And it will answer, āThere is now.ā
And for real, checkout some of Robert Miles videos on AI safety.
Chess was solved with heuristics. Which is not the same thing as ML and the two should not be conflated. They are fundamentally different technologies with different pitfalls.
I think youāre being imprecise here. āChatGPTā is a natural language model. Its whole thing is stringing together statistically likely words. It is Googleās auto-complete on steroids. It inherently has no concept of facts or truth. There is an ongoing debate between people who believe that we just need bigger training sets and more processing power and that will be good enough to fake it. There is also evidence that approach isnāt scaling like the Moar Data people think it should. Truth remains elusive.
The opposing side think we need to build entirely different architectures that can actually model the world and develop a sense of what is a fact and what isnāt. Yet it is unclear what that would actually look like in practice or how well it would perform.
Right now thereās enough money chasing this tech that we should get some kind of answer on this at some point. Though, people also threw billions at blockchain and crypto and that went nowhere. Investment is no guarantee of success.
Well, thatās actually the biggest problem here. If youāre not concerned about truth, or critically, are interested in diluting truth, this tech is ready for primetime. It gives plausible answers, which is awesome for those wanting to distribute disinformation. It can replace troll armies. It can generate deep fakes. It can saturate the internet so that even earnest AI projects are now ingesting disinfo at a massive scale, creating an ouroboros of bullshit.
Oh, the billions went somehwhereā¦ just not to the people that the majority wanted it to go.
Yeah, it probably isnāt entirely on-the-level for me to compare ML to a straight-up scam just because the two are both tech related. It was just the first thing that popped into my head. Very little money went into developing the underlying technology of blockchains. All the cash went chasing after tokens and the ability to mine and exchange them.
Ture totally different but it was a belief by some that a computer wouldnāt be a chess master. Which seems completely silly today. Humans are --generally-- clever little monkeys and I donāt have a crystal ball.
Youāre right. I said āChatGPTā when I intended to say, āChatGPT Likeā. I donāt expect a natural language model (NLM) alone to be trained into a full AI, but a NLM combined with other AI technologies someday in the future. I too have read articles on the āMoar Dataā debates and I side a bit in the middle. I see a NLM combined with a ātelling the truthā model not just a bigger NLM.
well, this thread
A side bar here ā¦
My son is autistic (but high functioning), but struggles with language. When he was a child and I said something that turned out to be incorrect, heād accuse me of lying. We had to teach him that a lie requires some level of intended deceit.
So abstracting that to AI (which I donāt think ChatGPT is considered) it is never going to lie to us.
HAL did. Thatās why he had to kill them so he wouldnāt have to keep lying!
TBF Wonka wasnāt fun either and ran an unsafe workplace, possibly staffed with slave labor. The only reason anyone wanted to hang out with Wonka in the first place was because he was stupid rich.
Conflating ML models with general purpose AI is an even bigger leap than Deep Blue to ChatGPT. Iām not trying to pick on you, really. These chat and image bots have sucked up a lot of oxygen lately and the hype train has been running over everything.
Candy. Iām in it for the candy. Bathe me in that sweet chocolate river.
Good news for you is chocolate tends to have a SG of 1.02 you should float quite well and be safe!
I donāt know how to feel about the fact that you just āWell actualliedā my meme
Ironically, Iād have cited the Futurama Slurm Factory episode if our posts were required to stick to the MLA handbook.
I mean, I wouldnāt even care if I go out by chocolate asphyxiation. Could be worse.
I think this is the 3rd or 4th utter dumpsterfire of a thread this yearā¦ Is there something in the water?