Petition to change AI to AW. It's not artificial intelligence, but artificial wisdom. It's wise as hell, but dumb as hell, like that one grandparent who's experienced everything but is dumber than a log
@bowiemtl
2 ай бұрын
No there is no wisdom either, at the end of the day it’s just advanced statistics right? There is no genuine understanding of anything, there is no wisdom because it doesn’t know how to handle situations it has never encountered. Sure humans might not either in some cases but that’s more indicative of our own abilities
@SwordFishTheFish
2 ай бұрын
Big D&D player here knowing the difference between wisdom and intelligence 😍
@Mulakulu
2 ай бұрын
@@SwordFishTheFish that's where I got it from 😂
@yuzzo92
2 ай бұрын
Given it's a data-trained algorithm, Wisdom is a way more accurate word than Intelligence than most realize.
@Mulakulu
2 ай бұрын
@@yuzzo92 I agree. It doesn't really do any thinking. It's just guessing based on previous knowledge (although I guess we are too, just much more effectively). I wonder what our difference is. Maybe us being able to think before we speak and realize some paths were wrong, like "ok, this is the question, let's find an answer. Does this work? No. Does this work? No. What about this? Yes!" gives us the ability to self correct, but AI always tries to guess it first try with no self reflection, as if it's writing out a story or a paper. It's too formal. What if the AI does the same? It has one model that's made to find possible solutions, one that's made for finding flaws in those potential solutions and gives corrections and one that the user sees that gives its conclusion? That would actually be AI instead of AW. There would also have to be a model that's trained to find out the requirements for completing the task and determine whether or not a solution is actually sufficient. AI lacks trial and error
@clarcks2540
2 ай бұрын
It’s like how you could make a device that “sees into the future” when it’s only calculating every possible outcome.
@GabrielPettier
2 ай бұрын
With a good prompt, some people have gotten really decent chess games, not as strong as specialized AI (which are insanely stronger than the best humans), but getting close to grand master level. The prompt can make a hell of a difference.
@Jabrils
2 ай бұрын
With just using the ChatGPT API? I don't believe you
@GabrielPettier
2 ай бұрын
@@Jabrils Ok, i might be pushing it a bit with grandmaster :D, but you can read "Playing chess with large language models" by Nicholas Carlini (2023-09-22), and "Debunking the Chessboard: Confronting GPTs Against Chess Engines to Estimate Elo Ratings and Assess Legal Move Abilities" by Mathieu Acher, if you want (a lot) more details about quite successful experiments in that direction, with gpt3.5-turbo-instruct, playing at around 1750-1800 ELO, which is definitely more than many human players. It helps to make it understand you want it to emulate a good player, by suggesting what its completing, is a game between two strong players.
@imiebaka
2 ай бұрын
The problem is the model was build to win, but forgetting to also stop itself from loosing.
@aeyempire
2 ай бұрын
By the model type they were created, they work as autocorrect but for whole paragraphs. And it won't ever have human intelligence, at least the way this model of Ai creation works. Because it doesn't have ability to think, it's an archive + autocorrect in it's core. You want intelligent Ai? Then you would need to start over with a different and more expensive approach on Ai development.
@eypower8647
2 ай бұрын
Define human intelligence or thinking. What do humens have in their core?
@austinprice770
2 ай бұрын
"More dumb" is incorrect grammar. "Dumber" is correct.
@Sifar_Secure
2 ай бұрын
When I read the title of the video, I thought that it must be an ironic solecism. Then I watched it and realised I was mistaken.
@jamesdriscoll_tmp1515
2 ай бұрын
Dumb ai is scary too
@zackarhino17
2 ай бұрын
Yeah, many people rely on it as if it's completely accurate but it's anything but
@njnexgen
Ай бұрын
Yeah but language models are not exactly the AI meant to be doing these kinds of things. We have logic-based algorithms, reinforcement learning and such which can beat humans easily. Once LLMs understand what's been asked of them and then consult the relevant RL model, then it won't be the same
@ace.of.space.
2 ай бұрын
is this your king is so funny
@dawson3115
2 ай бұрын
AI is a great idea generator. That's it.
@jacoballessio5706
2 ай бұрын
I wonder if you could improve LLM logic abilities in general using reinforcement learning on video games
@bjkorb7272
2 ай бұрын
If you think that's bad, try playing a game of connect 4 😂
@WizardOfArc
2 ай бұрын
Chat GPT dumb as rocks but talks big
@samucabitim
2 ай бұрын
Try Claude
@LIONGOD
2 ай бұрын
even a 4yr old isn't sd bad as chat gpt
@ghosty5195
2 ай бұрын
Finally, someone seeing through this GPT bs. I agree man.
@Y0y0Jester
2 ай бұрын
Aye, it might be at the stage of a three year old but once it can improve itself it'll become a 1000 year old sage within the blink of an eye 🤕
@zacharygolden9797
2 ай бұрын
Just because it does not currently possess the ability to perform spatial reasoning does not mean that it one day will. Does the inability to play tic-tac-toe really mean that it is not intelligent?
@MindlessShowCase
2 ай бұрын
I think it's more important to judge a system where it's currently at then where it could be at. With your train of thought why not say it will be smarter and more able then God itself?
@Petch85
2 ай бұрын
But it is so cheap.... just wait till you see all the money someone with a lot of money can save. No lawyers no judgings no artists. AI will solve all rich peoples problems and only at the expense of poor people and that is a risk they are will to take. 🤣
@dr-wz4gf
2 ай бұрын
solving all rich peoples problems?
@alexanderst.7993
2 ай бұрын
I didn't know rich people had problems. Nor did i know that we should prioritize them. Also "cheap"? Power doesn't grow on trees, mate.
Пікірлер: 38