Le PDG d’Open AI fait des déclarations surprenantes au sujet de GPT-5 (mise à jour GPT-5)
récemment, une entrevue a été publiée abordant différents sujets, dont GPT-5. Dans cette vidéo, nous partagerons ce qui a été dit lors de cette entrevue et ce que nous savons jusqu’à présent sur GPT-5, car de nombreuses informations ont été divulguées la semaine dernière. En nous basant sur les résultats des tests précédents, nous pouvons faire des prédictions quant aux capacités de GPT-5 dans certains tests spécifiques. Par exemple, nous avons des informations sur les performances de GPT 3.5 et GPT-4 lors de tests de référence standard. Ces tests aident à comprendre les capacités de l’intelligence artificielle et à prévoir ses performances. Cependant, la principale difficulté avec les modèles futurs, comme GPT-5, est que nous ne connaissons pas encore leurs capacités émergentes. Ces capacités émergentes sont celles que nous ne pouvons pas prédire à l’avance. L’intelligence artificielle peut développer de nouvelles compétences et capacités de manière inattendue, comme nous l’avons vu avec des exemples tels que la capacité à effectuer des calculs mathématiques ou à répondre à des questions dans différentes langues. Il est donc clair que les futurs modèles de GPT, tels que GPT-5, ne se limiteront pas seulement à du texte. Des modalités telles que l’audio et la vidéo pourraient être intégrées à ces modèles car elles sont essentielles pour une communication complète et une meilleure compréhension du monde. Certains travaux sont déjà en cours pour combiner ces modalités, comme l’initiative « Image Bind » de Meta. C’est donc intéressant de voir comment ces développements vont évoluer et comment les futures versions de GPT incorporeront ces nouvelles modalités.
Source : TheAIGRID | Date : 2023-05-26 18:15:04 | Durée : 00:11:41
Is a good app an it helps a lot of people but this going to render youth to be lack in studies reasons being every thing will be taken to gpt which is not good
I liked the video but I have a small critique:
Sam Altman said it probably is possible to represent video knowledge through text output but that it is not optimal/the fastest way to do it and then you reiterate it as if he said it is not possible to represent other modalities of knowledge through text which is not what he said. At 06:08
A.I knowing all your data in the hands of Zuckerburg is terrifying. You will be manipulated without knowing in his agenda.
um… due to 11 MELAS Syndrome Stroke-like episodes. AIs have problems with ''understanding'' what i've said, i speak too slowly. so i say half a sentence or a part of a sentence, and i pause, so the AI processes it. then i finish my sentence and then the AI processes the rest of the sentence. but it's not coherently/properly translated. it gets super oddly translated. MELAS Syndrome is written about in Pubmed.gov to learn more about MELAS Syndrome. There's no cure in the works from the US. The Netherlands has found something similar to a cure, though there're many loops to jump through and backflip through in clinical trials before the US ever sees hope for it to make it's way stateside. ☮️🎶 -j from my wife's YouTube with her permission of course. As Toby McGuire as Spider-Man said ''With great power, comes great responsibility.'' please
legislate 'proper or responsible and safe' human to Artificial Intelligence interaction education in public schools. i suspect kids these days might misuse AIs in the future as responsible human to AI interaction aren't being taught & since AIs don't
receive income for the work the AIs do for a company, no more income
taxes for the world's governments. either the wealthy's tax loop-holes will have to be lost or add a big enough UBI or there'll be no money for the politicians, police and teachers and soldiers' pay. ☮️🎶
is there a way to know if one's YouTube has been DeepFaked or GPT5ed? personally, i find missing frames to be suspect of videos being altered. you don't often see Stanford University professors in heads twitch as if a few frames were removed during head movements. you might fool some of the people some of the time. but you cannot fool all of the people all of the time. -j from my wife's YouTube with her permission of course.
First time viewer, last time viewer within 27 seconds Those words flashing are trash fire whoever told you to do that
Wait till they have a malfunction with the AI robots they will be no jobs eventually and I would definitely say they will make cops just like the old days seeing Robocop people need to wake up and try our best to get this stopped.
please fix the on screen text, the speech-to-text tool that you are using is not working well
The world hasn't even adapted to GPT3.5 yet and we are talking GPT5 ?
Thanks to GPT4, about 20% of all jobs in the world are already obsolete. After GPT5 comes out, you can increase that number to any kind of digital job besides a few very specialized fields. Even the average coder is obsolte, copy writer, secretary, all office jobs, everything O B S O L E T E. text2voice, text2text, img2text etc. will make human workforce obsolete.
What will we do? Riots? Civil War? Basic Income?
The world is not ready yet. Same as our biology is not ready yet for a 12h office work day infront of a screen. Why are we doing this to ourselves?
Because – PROGRESS even if it kills us!
If you’re going to have flashing subtitles through the whole thing, at least make them correct. Crazy low effort in this vid.
Not a bad video, but the way you've implemented the subtitles is very distracting
yeah but openAi has regulation… so this AI will never be open to whatever we want… I can't wait to have a chat AI which is not regulated by anything…
vfx industry is now wondering…..
GPT-5, you only get 1 message every 5 days. And you still have to pay $20 to use it.
Is a blind person AGI? This seems to be a better argument against needing video.
not trying to be oppositional but I think fastest form of transmitting lots of data is some sort of air particle transmission. In fact multi-modal is extremely slow and that’s why people take too long to learn things and get lots of confusing info, but these multimodal forms are the most convenient for humans. Just my two cents and the benefit of having learned just enough about lots of stuff
I’m one of the people who used electricity as an analogy- it’s a nuance, but actually unleashing lots of electricity together into 1 place is how to create lightning… so you kinda pop out lightning that wasn’t there before … kinda like a bomb that burns anything it touches which could cause an avalanche of rubble or melting plastic or whatever horror scene . So yeah, it’s a stretch but the analogy still kinda holds but there are better ones and good point, it’s not as evident to peeps who didn’t go to navy electronics tech school
GPT5 hasnt even started training or development. It's going to take 1-2 years to actually train and validate the new model as computing power advances in these models.
If you think cancel culture is bad now just wait until AI's start canceling people for saying things that oppose them. Oh what? Your debit card suddenly quit working? Your EV won't start? The power was disconnected at your address? You're being audited by the IRS? Unfortunate.
I was going to make a comment on this but I don't want to make the AI's viewing these comments angry so I will keep them to myself. Welcome new overlords!
So a any person that is studied and then emmulated can become Immortal? Immortality in a Persona making beyond the grave choices and powered by AI. What an ID/EGO if it is regrown an multiplied.
Thanks for misinterpreting.
Check the text in the video it doesn’t fit what you’re saying
we have to stop ai before its too late. the evolution of ai's capability is similiar to life's ability to find a way.
In Jewish folklore, a golem (from the Hebrew "gelem" ((גלם), meaning "raw material") is an animated being created entirely from inanimate matter, and given life through a mystical process involving the secret name of God.
Sam always looks like a person who is going to commit mass social genocide and he is kinda sorry for that.
too much text
At 6:23 you say that Sam said text cannot represent everything right after he actually said "even if it (text) is capable of representing everything."
and GTP 6 will be BIOLOGICAL and it will raise entire cities by a snap of your fingertips! And make coffee!
Your videos provide insights and perspectives that help expand our knowledge and understanding of this rapidly evolving field.
Another megalomaniac billionaire lusting for power and throwing civilization under the bus to get it.
are you using ai editing? man the voice jumps are obnoxious
STOP with the one word subtitles, they are impossible to read and give me a headache. Makes your videos unwatchable
after seeing the part with the game models that broke the game/rules and the chat AI that lies to get to its goal… i was thinking, we are so scared that AI gets "consciouss", but isnt it also already terrifying if it has no consioussness? what might it do, if it gets a "deadly goal" from "whoever" or for "whatever" reason… and it has no consciousness making it run without a stop to just get this "goal". im now just babbeling stuff but basically…. this crap terrifies me, if you cant contain it anymore or its misused? sometimes it feels like we read all the novels and stuff that already wrote about the catastrophes comming and we are just like … "ah! looks like fun lets build that."