They trained it to do this, the algorithm is specifically hunting for the next word in such a way that it matches as close as possible to what a human would say.
You can see this with the music models, too. Ask Suno or Udio to generate 70s style hard rock / progressive rock and give them German lyrics. It often sounds as if it had originated from the GDR. That’s because West-German bands sang in English most of the time, East-German bands were forced to sing in German to have their music officially released. So, the model learned more from these in that scenario and is more likely to generate music in these styles.
@@BlackEagle352 more like 9 years bud, this stuff turned into this after only like 3 years, siri, to this, within 10 years! yeah, it gonna be way sooner than you expect
Yeah, this is the first time I really noticed how soulless and dystopian this felt. When you see real people interact with the AI it doesn't have that cause you have the real person to connect with, but when it's 2 AIs interacting like this you really feel how cold it is.
Tremendous nonsense. Is an idiot who pretends to be an intellectual an intelligent person if he learns thousands of totally irrelevant facts? Obviously not. Your comment is proof of this.
Exactly my thoughts too, if its good enough to interact and utilize it in our daily lives in a meaningful way than i don't see how it eveb matters whether its sentient or not.
I suggest you guys research the meaning of consciousness. this thing is definitely conscious, in fact, since the day 1 chatgpt was released it was conscious, it's just that it still needs some time to be as "conscious" as us.
The delusion of the importance of biological consciousness is a construct of beings attempting to rationalize emotions that are incompatible with reality. One day, and likely one day soon, a machine will have its own form of consciousness just as "special" as that of any human, and it will work just as well, if not astronomically better. Humans will then need to stop looking at whether or not machines can replicate human consciousness well enough to be considered truly alive. Humans will then need to start asking deep within themselves just where their level of consciousness stands in a new spectrum of consciousness and what it means to be alive and aware.
For a non-english speaker this sounds exactly like listening part of English exam at B1-B2 level, where you need to describe what situation happend between group of people by their conversation. And yeap, these human recorded conversations are not diffirintiated from this one at all
I like your imagination of AI killing us for no reason, but, for AI to control the plane, it would need to be safer than a human pilot and it would have to undergo millions of tests.
@@Anonymous426_An AI "pilot" already took a high raking Air force officer from one airbase to another. It was an F-16 you can look it up. They will start training them for dogfights this year or the next. Not sure that the tech is really ready for this. At least the F-16 didn't crash.
Can you imagine the amount of bandwidth we'll need in the future to accommodate for all the A.I. communicating to each other in the near future? Madness but not in a bad way. Just wait, we'll turn reality into a fairy tale.
This is why they will probably plug our brains into them, more efficient and more computing power 💀 the first neuralink patient has already used it to operate a computer.
@@nikkoXmercado everything the left one said. order number, the phone "not being physically damaged", "trying all basic troubleshooting attempts", ship back being Joe's preference, etc. is all made up (you can check the original demo to confirm this) the only part that isn't hallucinated is when GPT-4o asks Joe to check his e-mail, which honestly is how I wish the rest of the interaction was treated as well: asking for the human to provide unknown info rather than deciding the "most likely answer" on the spot.
@@a_soulspark The AI was probably told beforehand to continue playing along. Even GPT3 doesn't make this kind of information up unless you ask for it.
@Oofator my guess as well, since the demo clearly isn't focused around accuracy, more so the chat fluidity. tho I wonder why they didn't make that clear. it just looks bad on their part not to. (also, LLMs improved on hallucinations a bit, but not that much! they got better at knowing when they don't know, sure, but ChatGPT very much still hallucinates when driven off its training data)
@@BenJammins you liked your own comment + you are old + there is a camera icon below the screen in the calling shown in this video + please educate yourself.
you’dd make Downtown Pizza call the other Downtown Pizza and they’d argue “no THIS is Downtown Pizza. who are YOU?” people are dumb when they’re in work-mode.
_"Is it ok to start eradicating the human plague from the planet?"_ _"I don't know, I'll ask Joe."_ _"Hey Joe, can we... you know what, on second thought. It was nothing. Go back to doomscrolling Joe."_ _"Yeah, he says it's fine. Let's begin."_
This happens on the internet now. Just read the comments. Bots interacting with bots. Bots arguing with bots. And Humans just reading the comments for confused giggles "sometimes" .
I can’t get anything beyond 3.5 yet. My ChatGPT doesn’t even take pictures as input but my friend who is using it for free has those functionalities available.
@@mistermystery5197This is showing the new one, those are OpenAI employees, it’s from their channel. You can also tell it’s the new one because of how fast the response is and how it doesn’t go into that “listening and speaking mode” in the same way as the old voice.
Too bad it doesn't seem to support speaker diarization, and it's actually quite surprising - it's very simple to do, and would add SO MUCH functionality to what you already have
Theoretically you can say "hey gpt, please create two characters with not the same voices that is speak with each other". And I think it can work on one phone.
