General > General Technical Chat
Mind blowind OpenAI's text to video SORA.
RAPo:
Is it really, or are we blinded by the beatifull images?
BrianHG:
--- Quote from: ebastler on February 17, 2024, 07:51:18 am ---
--- Quote from: BrianHG on February 17, 2024, 01:43:47 am ---On the other hand, now novelists may release new books and they might create a real life looking film simultaneously.
--- End quote ---
What makes you so optimistic regarding the future need for human novelists? ???
At least as far as Hollywood-style film scripts are involved, AI should be able to produce those nicely.
--- End quote ---
Someone still has to come up with the prompts, for now anyways. However, soon everything will be almost all AI generated and the human reaction will become the want to have some story from an authentic human. Having human actors, actual locations and sounds will be reserved for live theatrical performances.
berke:
All of this is going too fast.
What is the likelihood that the big players have training algorithms and/or hardware that is much more efficient than what has been revealed publicly? I have a hard time believing that gradient descent-like algorithms (ADAM etc.), which are theoretically inefficient (much slower convergence than second derivative Newton methods) but fit in memory, are a good use of resources to train these humongous models. Do they really just throw obscene amounts of computing power at these kinds of algorithms with no or very little secret sauce? Or did they figure out how to run something with a higher rate of convergence on these big models? Are they using some kind of analog hardware (not even talking quantum here)?
Also I wouldn't be surprised if TLAs already had 10 to 15 years ago technology comparable to what we have today, specifically something at least like OpenWhisper and for images maybe something similar to the first versions of Stable Diffusion. After all we know they had voice keyword spotting systems in the 90s.
How good was the accuracy of classical, publicly known "non-AI" voice recognition algorithms back then? It has to have been be pretty damn good to be usable for mass surveillance without triggering a deluge of false alarms that have to be cleared by tens of thousands of human operators.
In other words, I think the AI we're seeing now has been recently technology-transferred to the public, however some of the secret sauce (for fast training) might still be held as secrets. Heck, maybe OpenAI subcontracts the NSA to train their models on their massive farms and/or non-classical computers. Someone call the WTO, I sense unfair government subsidies!
KE5FX:
--- Quote from: berke on February 21, 2024, 05:41:00 pm ---In other words, I think the AI we're seeing now has been recently technology-transferred to the public, however some of the secret sauce (for fast training) might still be held as secrets. Heck, maybe OpenAI subcontracts the NSA to train their models on their massive farms and/or non-classical computers. Someone call the WTO, I sense unfair government subsidies!
--- End quote ---
That's not realistic, given that the silicon needed to train these models takes up entire warehouses and didn't exist only a few years ago.
It's been a long time since the three-letter agencies had access to better tech than the public.
berke:
--- Quote from: KE5FX on February 21, 2024, 05:48:49 pm ---That's not realistic, given that the silicon needed to train these models takes up entire warehouses and didn't exist only a few years ago.
--- End quote ---
What I'm saying is that they might have better algorithms or custom hardware.
--- Quote ---It's been a long time since the three-letter agencies had access to better tech than the public.
--- End quote ---
If you're in a position to know you shouldn't be revealing such secrets!
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version