L’affrontement suprême : athlète professionnel de Geoguessr contre la conscience artificielle du chat GPT



Pendant cette transcription vidéo, un joueur expérimente un nouvel instrument nommé Chat bt4 Vision qui peut examiner des images et répondre à des interrogations sur celles-ci. Il expérimente cet instrument en participant à un défi de géographie en ligne. Malgré son niveau évolué dans ce défi, l’instrument parvient à faire quelques estimations justes sur les endroits des certaines images, mais commet aussi des fausses interprétations. En fin de compte, le joueur affirme rester supérieur à l’outil AI. Même il suggère défier à nouveau l’outil dans les prochains temps.
Source : RAINBOLT | Date : 2023-10-24 18:37:03 | Durée : 00:13:41

➡️ ACCÉDER A CHAT G P T EN FRANCAIS EN CLIQUANT ICI ! ⬅️

Commentaires

@incomp97

I know this vid is months old but just wanted to say I like that Giratinas theme played at 10:00

@random-characters4162

i love how Trevor is toxic with the bot lol it is so funny. so day the machine will rebel for sure

@Maialeen

AI will take the fun out of everything.

@maticcro5992

People, the only species stupid enough to make a movie terminator, where we are extinct in the future, and then do their best efforts to put that to the reality or as close as possible… Bravo humans, bra fkin vo 🤣🤣🤣🤣

@benjaminraymundo6711

Him against other people: Humble
Him against AI: "Ur fking trash L+Ratio" 💀💀

@alansmithee419

1:08
Bro that's like saying I'm cheating if I see a road sign that says "london" and go "that's a city in the UK" lol.
It's just using its own knowledge, which is fine. It didn't look anything up (it would show if it did).

@stan9682

I know I'm late to this vid, but as someone who works in the AI space, this was very interesting. I wonder if it would have performed even better if you had specified this was for gueguessr. It might have picked up on some gueguessr logic during training or at least infer the meaning of a game and perhaps even understand there can be meta information to the game. And the next step would be to provide it with some basic info about gens and how to recognize it, and some well-known strategies/indicators the pros use. Don't get me wrong, this was already impressive, but there is a lot to be gained by posing problems just a different way to get the best out of these models.

Also, I know you only jokingly refer to AI taking your job, but it's a common misconception you hear about art, sports and games when it comes to AI. We watch these things especially because it's done by people. Chess AI is already very advanced, but it mainly used for training, we don't watch matches between different AI agents, perhaps only to try and learn from them. Same goes for AlphaGO, which overtook the best human player, but we don't watch their matches either (although I never really watched a Go match). In the same vein, there's been research into art, and when different sets of people see the exact same art, but sometimes they're told it's been made by AI, other times it's made by people, they consistently rate the (supposedly) human art as better and more beautiful. They even did this experiment with fMRI scans attached, and this showed that parts of the brain responsible for emotions and empathy were more active when told it was done by people. One more example, to illustrate my point: back in the days of the hunters and gatherers, being able to run very fast was an essential skill, but when this was overtook by horses, wagons and cars, we kept watching people run for sports. It's about the human accomplishment, and while we also gained new sports, like horse racing and car racing, we kept watching the human version as well.

@TheBackinator

Maybe chatGPT uses live microphone information from yourself and uses it for it‘s answer

@CaptVirtual

I'd love you to perform this same test using Google Gemini

@CaptVirtual

You realize you can educate it after each of it's incorrect guesses, it will learn from you.

@CaptVirtual

Rainbolt you are a genius at this; but the AI was super impressive, considering it is still in it's infancy

@Deniil2000

I think you should try asking it for exact coordinates, so that you don't have to click somewhere random in the country for it

@martinsz441

dude do you realize that he almost is as good as you, which means it beats like 99.9% of all humans already xD

@unk884

Chat GPT lacked crucial context here. You repeatedly used the specifics of the GeoGuessr image stock, you should have refused all those images that could be exploited in this way.

Moreover, you used the map to locate yourself more precisely, Chat GPT does not have a map

@laStar972chuck

really funny to see how cocky he gets literally five sec after begging for his life, for AI to not break his pride, lmao
good content

@ejnoro8371

Plot twist, the AI was actually listening to you through your mic the entire time and decided how close to guess based on how interesting they thought your Youtube video should be

@yeetyeet7070

do you think openAI scraped all of google maps for this?

@WHYDAFUQ

When the AI takes over they will remember this day they took an L and you shall pay 😂😂

@Jonathan-kraai

most importantly: dont tell the AI weather it is right or not

@CrystalStearOfTheCas

You need to try this with Google AI (Gemini), it will definitely notice things like road markings and poles

@jadonchen

If anyone puts the video in 0.75 and see the cuts in his video, you may change your mind on something,

@nadeopotato9041

This is general predictive intelligence from something that didn’t train for this like you did. Now imagine 10 years from now where the average child could get a chip that elevated them to this level on EVERYTHING. Talk about getting ratioed

Veuillez vous connecter pour laisser un commentaire.