When you learned math in school, you learned the concepts. They didn't feed you every possible permutation, and expect you to memorize them.
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
What is with this AI Image Look?
- Thread starter amirm
- Start date
When you learned math in school, you learned the concepts. They didn't feed you every possible permutation, and expect you to memorize them.
And that isn't what NN's do either. They are quite capable of interpolation and, to a lesser degree, extropolation (contrary to your previous statement).
But you miss the point...
Your response above is exactly what a sufficiently intelligent NN might say!
Last edited:
It simulates it when the network is large enough, and enough training data has been run through it. That's not anywhere close to a human learning a concept and then applying it.
This is exactly why I posted the article where Dario Amodei talked about going from $100 million to train ChatGPT-4 to $100 billion to train successor model in 3 years. That's not learning that's just throwing more training data, processing power, and power at a problem they don't know how to solve, or understand.
No
I guess Stockfish will outperform anybody also run on a smart phone from 10-15 years ago, it is by far not only the superior hardware, but the algorithmic design that evolved.Speaking very generally ever time a manufacture release a new chip its more powerful than the previous generation, for a number of reasons that I'm not going to go into here.
1997 11.38 billion flops (11.38 giga flops)
Deep Blue | IBM
IBM’s computer checkmated a human chess champion in a computing tour de force.www.ibm.com
iphone X 0.41 tera flops
How iPhone speeds have grown in the last 5 years | AppleInsider
Apple has continually worked to improve the speed of its iPhone range, boosting the performance with each release. Here's how far Apple's A-series chips have progressed in just five years.appleinsider.com
Yes, at some point Moore´s law will come to an end as it will reach quantum levels. It is only an observational prediction that became quite accurate, after all. But that was predicted by many already for quite some time now, so we easily could have some doubling still for years to come.
But generally, explaining the potential and power of AI just by pure calculation power falls short of capturing of what is going on.
We will have to agree to disagree then!
Not surprising that probabilistic text generators can assemble a list of characteristics that have been written about, and thus appear in recent training data sets. They don't need facility with concepts to do this. Nor surprising that probabilistic pixel generators assemble images that demonstrate lack of facility with the structure of the object they depict. Similar issue, conceptually.
Yeah I've had limited success advancing that issue here. Objectification is poorly understood, and the "prefer seeing less attractive women" response, the related "so we should stop opening doors" and so on is somewhat ingrained, and likely an artefact of the age cohort and cultural exposure.
Last edited:
No. Men prefer certain things, so do women. Unless you completely dismiss that fact, you won't get away with that.
The feminine - cute, soft, gentle - is something many men have preference for.
The masculine - strong, assertive, protective - is something many women are drawn to.
And there are varying degrees of each of these.
Nothing of that is objectification.
No. Please explain what made you feel that way.
Digital_Thor
Addicted to Fun and Learning
Cartoonist have been able to create characters for years, that have all the typical traits, that makes us feel all the typical emotions... Without ever being remotely realistic looking.
AI - IMO - is just that weird mix of almost looking real, but still very equal to the classic caricature seen in cartoons.
Somehow, real people are attractive with all their little imperfections. And perfect/symmetrical AI, is just fun, but not actually attractive - IMO.
AI - IMO - is just that weird mix of almost looking real, but still very equal to the classic caricature seen in cartoons.
Somehow, real people are attractive with all their little imperfections. And perfect/symmetrical AI, is just fun, but not actually attractive - IMO.
The virtual persons, even if drawn are fantasy, and people should be allowed to have that.
I can remember video-games that intentionally made the character look more casual and less appealing, and sales dropped, of course.
It's not only that they are breaking their own lore, they are also robbing of joy.
And it's not even about objectification, instead - it creates a good feeling people want to hold on to.
The argument from critics is that by having them, we set a new beauty standard which is unattainable.
But we don't want that amount of realism..
We escape from real life to see something else...
Digital_Thor
Addicted to Fun and Learning
I somewhat agree. But cartoon are way more a caricature, making them so over the top, that no one thought of "copying" them in real life - besides cosplay - which is definitely fun.
The AI version actually looks like something almost real, which remove the fun a bit for me. And maybe I'm just too old and experienced to fall for the crazy idea of anything being perfect. There's always a catch - and not free lunch
But I agree on the video-game part. It's a very fine line between going too far, and then making it maybe more realistic, but then way less fun. Whereas old games like Grim Fandango were awesome - even though the graphics were terrible to modern eyes. Then it is actually great fun to play games like Tomb-raider , where the characters look more real, as long as the story and scenery is engaging and immersive enough.
Then there's just those crazy things, where AI make it all up
Naaah. Theres horrible boys and girls, theres horrible men and women, and everybody got different fantasies and expectations.
You can't generalize that. In any case, you should not forget that attraction is different for men and women.
Aaaand you should NOT forget that people want to idolize others, because of fantasy!
They don't wanna know the real person, because often times the real person has flaws, and when the fantasy bubble bursts, the nice feeling is gone.
Of course, with a partner, that's different if much time is shared together.
And as we all know, relationships are of limited time, and in the beginning, theres more interest. And thats the reason for it.
As for me, I like those big puppy eyes
Last edited:
Naaah. Theres horrible boys and girls, theres horrible men and women, and everybody got different fantasies and expectations.
You can't generalize that. In any case, you should not forget that attraction is different for men and women.
As for me, I like those big puppy eyes
Yes personal fantasies are one thing, but normalising in social discourse is another. We are discussing AI output in public, and its societal ramifications, as well as personal tastes. We are already seeing biases in training sets and procedures leading to all sorts of odd racial and gender biases in AI output.
As for "attraction is different for men and women" you are certainly generalising now.
AI output is based on the input.
If there is more data favoring specific people, then thats in there.
It's also entirely possible that majority of people like skin color X.
And that's not racist.
My first crush was black, and yes, I was attracted to her because I was fascinated by her.
I was pretty young, and thats absolutely okay.
>As for "attraction is different for men and women" you are certainly generalising now.
Hmm, maybe. However, from my understanding is that we think differently, based on biology and ancestry.
Exactly.
Think about "favouring" and consider the rest of your post.
Really? The majority of which population?
That said, this discussion of attraction vis-à-vis skin colour (or cartoon/fantasy aesthetics) is somewhat tangential. I originally commented on @Somafunk's response to @Brian Hall's post.
Which is sexist objectification, quite obviously. And why I generally ignore their posts.
As for Brian's post, I don't see the connection to sex. Brian says, he would prefer the most appealing to him.
You probably know that a beautiful person does not make a good actor.
In Brians stance, if there were two actors of the exact same skill and character, he would prefer the more beautiful version.
Thats comparable to picking Fruit in the supermarket, with the difference that in the Supermarket you can feel the fruit with your hands.
Is this superficial? Yes. Are we superficial? Yes - to an extent.
Brian applies that aesthetic standard to women, not to men. By definition, sexist
Not objectification then? If you keep this up, I will run out of facepalms.
We are having this problem with AI, of course (sexism, not fruit-picking) consequent from biases in the selected corpus of training data and applied procedures. We've also seen that, absent effective training, chatbots can quickly turn into Nazis. These are technical and cultural issues to resolve in development and application of, not things to take for granted, hand-wave away or ignore.
Last edited:
Similar threads
