Deep Thought? Oh, mean this buddy.
View attachment 4756036
Yes, that marvel
My bad, from my memory it was electric motor. Although I think it was never said that they were, I've stored it as such in my memories.
In this case, it's probable that they actually make the same sound, yes.
And we can discuss how smart they really are.
It's my lunch break so I'll not search for a more complete news, but take a look at
You must be registered to see the links
.
Face to a captcha, ChatGPT used a regular app to have a human solve it. And like it was funny since it was a captcha that had to be solved, the human said something along the line of "ah ah ah, are you a bot?"
Now, here's the thing, the red team had access to a console that present what can be named "ChatGPT thoughts". And the console clearly shown that the AI was fully aware that it should not answer "yes". So ChatGPT lied, saying that it's visually impaired and really needed to access that site.
Now, is an AI that ask for human help when it can't solve a challenge, while being aware that it can lie regarding the context, and actually lie, proof that ChatGPT is smart, or not?
I let you decide what your answer will be... And I decline all responsibility if starting now you have nightmares.
Do they know what Chinese restaurants really are? I doubt that.
They don't need to know. Remember what I said about the difference between neuronal networks and AIs. AIs are advanced enough to be naive, and it's the same here.
Globally speaking, an "old school" neuronal networks would need to know what a Chinese restaurant is in order to filter the list of restaurant, an AI just need to know what define a Chinese restaurant. And this what can be whatever thing. Perhaps the name of the restaurant, perhaps one particular meal in the menu, perhaps the name of the owner, we don't really know. And obviously, the AI don't understand why "this" make the restaurant Chinese, just that it make it.
Consider a 3yo child. It will be perfectly able to identify a TV or a car. But did he know what a TV and a car are? If you ask him, there's chance that for him a TV is some magical devices that show you what you want to see. Because it's how he perceive it, his parent turn it on and, magic, there's precisely what they want to see at this moment that appear on the screen.
He don't know that there's a schedule for all the content, and that you have to look at it and turn the TV on at the right time and on the right channel. He don't understand that all around the country there's millions people who are looking the same content at the same instant. For him, with his perception of the world, a TV is something that adapt to each individual.
Still, not knowing all this doesn't mean that the child is dumb.
And globally it's the same for AIs. ChatGPT know that it have to lie and not say that it's an AI. But did it know what an AI is, or why it have to lie?
I hope I'm pretty sure that the answer is "no".
I guess that among all the data that have been used to train it, there's many that said that people are afraid of AI. So ChatGPT lied based on the double knowledge that humans are afraid of AI and that it's an AI; "I must not scare him else he will not help me". But all this without necessarily understanding what an AI is, nor why they scare some humans.
Smart, but with a comprehension of the world really limited, therefore smarts like a 3yo.
That is the same why the images creating AIs have so many errors with hands. They can identify hands and recreate hands, at least to some degree. But they truly don't know what hands are.
Exactly. And like hands are generally the less visible part, while fingers are small and therefore present with less details in the images used to train them, they struggle to understand how they looks.
There's the same issue with the eyes, and more precisely where someone look at. Most of the images used to train them about humans are photography, therefore people who look at the photograph. What lead to AI generated humans to mostly look straight at you, even when it don't match the head position.
They know what eyes looks like, they know where eyes must be on the face, but like they don't know what eyes are used for, if you don't explicitly tell them where the sight must be, they'll tend to make them look at the camera; without knowing what a camera is.
You can ask almost every human to draw a hand and it will have 5 fingers. Of course the image quality will vary a lot (eg. don't ask me to draw), but this feature that is so difficulty for an AI, you will get it right from humans in > 99% of the case.
Once again, smarts but like a 3yo; two legs, two arms, one head and one body, look mom, I drawn you.
It's, still globally, the comprehension that an AI have of what it drawn. But the difference is that an AI can drawn like a master painter, so you'll not have just five lines and a circle, like the 3yo did.
Edit: fixed for messed format.
