I hope you’re keeping well. Looking forward to some more content, when you have the time.
@ptrckqnln
Жыл бұрын
Hoping to see more from you!
@ametrinefirebird7125
8 ай бұрын
Nice front end to play around with....hehe
@PunmasterSTP
Жыл бұрын
I’ve been playing around with Stable Diffusion and even made a few videos of it, but when I saw the notification, I knew the king had arrived. As usual, you managed to succinctly and brilliantly explain a topic in astounding detail. The images in this one were just mind-blowing too. Thanks for uploading!
@Kram1032
Жыл бұрын
That's a neat AI
@tritoner1221
Жыл бұрын
i see what you did there
@uku4171
8 ай бұрын
When's the next video?
@anondude6361
Жыл бұрын
Please talk about evoloop cellular automata. I don’t really understand it
@flanhed
Жыл бұрын
New microphone? Glad to hear your voice again.
@shanruan2524
Жыл бұрын
Wow
@fmj.mytube8846
Жыл бұрын
Hi, I love your channel, I just found it trying to find some videos about traits. There is a section on the paper of stanley but the subject is not well elaborated. Do you know something about TRAITS or any article to share? Thank you and keep them coming!
@J1428753
Жыл бұрын
nobody: ai: *C O L D D U N T*
@ryanmapping7944
10 ай бұрын
0:52 from astronaut to duck to yoshi to frog, what did you change to it to do that?
@samevans4834
Жыл бұрын
“Nice frontend to play around with” comes up with a woman in bikini armor? That’s a bit sus
@ZardoDhieldor
Жыл бұрын
The AI has got humour.
@sebastianjost
Жыл бұрын
It would be great if you added some links to the installation you used in the description (and maybe some extra info)
@neatai6702
Жыл бұрын
Done; ... and loads of info on the reddit page www.reddit.com/r/StableDiffusion/wiki/index/
@nutzeeer
Жыл бұрын
Shouldnt running this backwards achieve computer vision?
@feedayeen
Жыл бұрын
Functions are usually not invertible. But the same types of algorithms are used for AI based computer vision and object recognition. Biggest problem is computer vision applications usually involve a high cost of failure, the hole lets generate 100 images and find the coolest ones really doesn't work if you're driving cars or diagnosing a patient. And humans vision also has other factors to help us out which we use heavily to filter our responses so while our vision is far from perfect we compensate in other ways that are hard to train for.
@nutzeeer
Жыл бұрын
@@feedayeen i just remembered that the first AI generated images came from running image recognition backwards
@ristopoho824
Жыл бұрын
@@nutzeeer Yea. It's weird how these work. And awesome. There was something in twominutepapers video on some AI being ran halfway backward and making weird images. I'm waiting for my morning coffee to finish so sorry for being incoherent. But, yay AI. Also useful but 100% neat
@revimfadli4666
Жыл бұрын
@@feedayeen even if the function's not invertible, couldn't doing backprop on the input still work(like deep dream)? Though performance issues might come
@feedayeen
Жыл бұрын
@@revimfadli4666 They're different goals, DeepDream isn't trying to recognize things in an image. We train it on large datasets and then it develops a set of patterns that match natural objects. It then amplifies them. Although image recognition at some percentage of accuracy is not difficult and there's a lot of even low effort examples out there. It's when you need many 9's of precision in different environments that it's a problem.
@asdf8asdf8asdf8asdf
Жыл бұрын
+1 for the frontend
@thehammurabichode7994
Жыл бұрын
Makes me sad. Amazing tech, _unbelievable_ imagry, but I feel cheated, somehow. Guess you could call me "human."
Пікірлер: 25