Jump to content

We are not going to make it -OpenAI's SORA


UTGrad98

Recommended Posts

We had a significant % of the population who believed in porn rings inside a pizza hut just by reading some texts and taking it for gospel. Now they will get first hand accounts from Joe Biden on exactly that. 

Edited by UTGrad98
...
Link to comment
Share on other sites

11 minutes ago, MissingInAction said:

I, for one, welcome our new AI overlords.

 

Yall better get on board before you get turned into a battery.

I manage the people who repair the robots that mine for Lithium. Got a video to prove it.

I'm a people person and AI says I'm good, so I'm good. 

  • Hook 'Em 1
Link to comment
Share on other sites

  • UTGrad98 changed the title to We are not going to make it -OpenAI's SORA

The "realistic" scenes have several inaccuracies and unnatural movements but this is still very impressive and if there is exponential progress, then many people could be fooled. My concern is for the younger generations who are immersed in digital realms that have supplanted and stunted their observations of the actual world. They may not be able to tell the difference because they don't have enough experience with reality. 

Link to comment
Share on other sites

3 minutes ago, Newdoc said:

The "realistic" scenes have several inaccuracies and unnatural movements but this is still very impressive and if there is exponential progress, then many people could be fooled. My concern is for the younger generations who are immersed in digital realms that have supplanted and stunted their observations of the actual world. They may not be able to tell the difference because they don't have enough experience with reality. 

And they may not care.

Link to comment
Share on other sites

9 hours ago, safe sex said:

Legislation needs to happen ASAP

you forgot to include the global implementation of said legislative restrictions; and the ability to enforce those who do not follow the rules. And a manual backdoor to pull the plug and inactivate AI from spreading.

When do we thing the real life August 29, 1997 will take place?

Link to comment
Share on other sites

39 minutes ago, Newdoc said:

The "realistic" scenes have several inaccuracies and unnatural movements but this is still very impressive and if there is exponential progress, then many people could be fooled. My concern is for the younger generations who are immersed in digital realms that have supplanted and stunted their observations of the actual world. They may not be able to tell the difference because they don't have enough experience with reality. 

This is exactly the issue. Our eyes trained quickly to notice AI fake photos after being in awe at first. And I think our eyes will identify this new iteration of video, at least for now. 

But young people? Fucked. 

Link to comment
Share on other sites

Limited to 1 minute (technical limits to on, not safety), probably takes even with a super cluster of GPUs at least 15 minutes or longer to make. Probably not worth freaking out about, it’s a showcase of technology. Have none of you seen video games with generated scenes? They look really good any you can interact with them and control them dynamically. This isn’t some far fetched weird take between what’s real and fake, its just a neat demo that makes a moving picture instead of a still like midjourney is doing.

 

Link to comment
Share on other sites

17 minutes ago, immamac said:

Limited to 1 minute (technical limits to on, not safety), probably takes even with a super cluster of GPUs at least 15 minutes or longer to make. Probably not worth freaking out about, it’s a showcase of technology. Have none of you seen video games with generated scenes? They look really good any you can interact with them and control them dynamically. This isn’t some far fetched weird take between what’s real and fake, its just a neat demo that makes a moving picture instead of a still like midjourney is doing.

 

Yea for the next few weeks you are probably right.  Now say the same thing 10 months from now when version 3.0 comes out, let alone 5 years down the road. I will say  I'm  happy with this tech for the new video game I'm making. Cut scenes and what not. 

Link to comment
Share on other sites

Just now, UTGrad98 said:

Yea for the next few weeks you are probably right.  Now say the same thing 10 months from now when version 3.0 comes out, let alone 5 years down the road. I will say  I'm  happy with this tech for the new video game I'm making. Cut scenes and what not. 

I don't think people realize how incredibly difficult it is to even cluster these GPU's together. It's a problem I'm working on as my day job and it's one that is years away from solving because the hardware doesn't even exist yet.

What is CXL Anyway? · Jackrabbit Labs Blog (jrlabs.io)

 

  • Hook 'Em 1
Link to comment
Share on other sites

1 hour ago, immamac said:

Limited to 1 minute (technical limits to on, not safety), probably takes even with a super cluster of GPUs at least 15 minutes or longer to make. Probably not worth freaking out about, it’s a showcase of technology. Have none of you seen video games with generated scenes? They look really good any you can interact with them and control them dynamically. This isn’t some far fetched weird take between what’s real and fake, its just a neat demo that makes a moving picture instead of a still like midjourney is doing.

The takeaway isn't that a 60 second generative clip is in itself scary/impressive.  The takeaway is within the space of 1 year, we went from midjourney's impressive single frame output... to an uncanny motion image.  15+ frames per second * x seconds.  All created from a single user prompt.

Yes it takes some compute to churn through.  But compute itself is also growing geometrically. 

It's the trajectory that is the big thing.  We have AI text, image, motion, and sound.  How many people and how long would it take for to make the equivalent of that in the past?  How much does it cost to commission a 30 second animation or vignette for an advertisement?

Link to comment
Share on other sites

3 minutes ago, 52-80 said:

The takeaway isn't that a 60 second generative clip is in itself scary/impressive.  The takeaway is within the space of 1 year, we went from midjourney's impressive single frame output... to an uncanny motion image.  15+ frames per second * x seconds.  All created from a single user prompt.

Yes it takes some compute to churn through.  But compute itself is also growing geometrically. 

It's the trajectory that is the big thing.  We have AI text, image, motion, and sound.  How many people and how long would it take for to make the equivalent of that in the past?  How much does it cost to commission a 30 second animation or vignette for an advertisement?

Most interested in the last line.  We’ve had great VFX in games and movies for a while now, and AI is making this even cheaper and easier. For a long time it’s been cost-prohibitive for all but major entertainment corporations. What happens when individuals or even low-funded intell services can do this? 

Link to comment
Share on other sites

1 minute ago, 52-80 said:

The takeaway isn't that a 60 second generative clip is in itself scary/impressive.  The takeaway is within the space of 1 year, we went from midjourney's impressive single frame output... to an uncanny motion image.  15+ frames per second * x seconds.  All created from a single user prompt.

Yes it takes some compute to churn through.  But compute itself is also growing geometrically. 

It's the trajectory that is the big thing.  We have AI text, image, motion, and sound.  How many people and how long would it take for to make the equivalent of that in the past?  How much does it cost to commission a 30 second animation or vignette for an advertisement?

except its not and this isn't something new. Frame generation tech has been around for a while. Compute isn't growing geometrically, it's at best linear with a declining curve. The new GPU's aren't somehow miraculously shitting all over the old GPU's. Same with CPU's. also neural networks have been around a long time. This whole idea of boiling the entire internet and distilling a model that you can generate stuff from is also not at all new. 

I'll tell you this - no one that can afford real videos and shots is going to look to this as an alternative, it could be an alternative to those who didn't already have the means. Complete creative control and letting a machine do something are completely different and I'd argue the prompt engineer you would need to hire to "direct" the AI isn't just some cost you don't get to include. That person right now is extremely expensive and probably not at all worth it when you combine the true cost of training and inference + the person who actually knows how to get the exact thing that people want out of it. Everyone is seeing the subsidized cost of AI now, the true cost is obscene and completely and wildly absurd for the level of output we are getting. It's why there is a real threat to the mega GPU deployments when a home 3080 GPU can just give nearly the same experience as some huge inference farm. 

If you know how some of this works behind the scenes it feels a lot less like the magic everyone proclaims it to be. The inevitability of knowing what we were trying to build/aim to build and what already existed is too easy to see for the people not to go from will smith eating spaghetti to still having weird fingers and stupid physics in a world that looks a lot more like a AAA video game with ray tracing. 

 

Link to comment
Share on other sites

1 hour ago, immamac said:

I don't think people realize how incredibly difficult it is to even cluster these GPU's together. It's a problem I'm working on as my day job and it's one that is years away from solving because the hardware doesn't even exist yet.

What is CXL Anyway? · Jackrabbit Labs Blog (jrlabs.io)

 

The first post in this thread addresses the bigger problem with this I think.  It can be weaponized to spread disinformation to stupid people and further inflame global issues because of a made up reality.  Unless I'm misunderstanding (which is entirely possible), this is what we all feared deep fakes would be, but way worse.

Link to comment
Share on other sites

except that's a symptom of a problem. No one that thinks critically would be susceptible to this kind of stuff, the issue and real problem is that society as a whole is not thinking critically anymore about anything. 

Idiots who believe Q-Anon shit popped up way before deep-fakes. They are just idiots. The problem is idiocracy not technology advancing.

Link to comment
Share on other sites

1 minute ago, Biff Tannen said:

6a492190-ee69-495e-8adc-732ee4c0e9a0_tex

I'm not trying to say everyone is an idiot in here and take away from the cool factor, I'm trying to temper the hyperbolic OMFG GAMECHANGER nature of this a bit. It's not even close to there yet and watching this thread is like watching some analyst try to justify 80x forward earning on Nvidia. 

  • Hook 'Em 1
Link to comment
Share on other sites

11 minutes ago, Biff Tannen said:

t can be weaponized to spread disinformation to stupid people and further inflame global issues because of a made up reality. 

We already have Fox News.

  • Hook 'Em 1
  • Haha 2
Link to comment
Share on other sites

1 hour ago, immamac said:

Limited to 1 minute (technical limits to on, not safety), probably takes even with a super cluster of GPUs at least 15 minutes or longer to make. Probably not worth freaking out about, it’s a showcase of technology. Have none of you seen video games with generated scenes? They look really good any you can interact with them and control them dynamically. This isn’t some far fetched weird take between what’s real and fake, its just a neat demo that makes a moving picture instead of a still like midjourney is doing.

 

Because no technology ever advances or is made scalable.

My fear of AI has nothing to do with a singularity and sentience, that is very far off and/or unlikely.  My fear is the moment of convergence between quality and commoditization, which is coming sooner than we will be ready for.  It's the point in time when any asshole can prompt: make a video of my political opponent having sex with a minor and the AI spits out something that is virtually indistinguishable from reality.

Reality is already obscured by the relatively slow stream of human generated misinformation.  AI is going to increase the amount and quality of the bullshit by orders of magnitude and make it incredibly difficult to understand what is happening unless you are there to personally witness it.

  • Like 2
Link to comment
Share on other sites

My daughter is majoring in New Media (inclusive of graphic design, film, & coding stuff that I don't fully understand). This shit worries me. She says she's not worried about AI affecting her job prospects. The subject is a hot topic in her program for obvious reasons, but they're being taught that it'll be an aid and not replace her and her fellow students. I do admit to having my doubts.

Link to comment
Share on other sites

8 minutes ago, bolverk said:

My daughter is majoring in New Media (inclusive of graphic design, film, & coding stuff that I don't fully understand). This shit worries me. She says she's not worried about AI affecting her job prospects. The subject is a hot topic in her program for obvious reasons, but they're being taught that it'll be an aid and not replace her and her fellow students. I do admit to having my doubts.

Serious question, what's considered "new media"?

Link to comment
Share on other sites

11 minutes ago, FirstTimeCaller said:

Serious question, what's considered "new media"?

 

9 minutes ago, immamac said:

The Internet GIF by DrSquatch

 

Yeah, pretty much. Like I said, coding, graphic design, interactive shit, film/editing, multimedia technologies, and other stuff I don't understand.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.



×
×
  • Create New...