Capitalism never worked, if it worked out for you was because someone else was being explored for it to be possible. You and me very likely did not explore anyone personally but the system did it for us.

Yep. Capitalism is the problem that really needs to be solved in the coming decades, if not sooner.  

Then it should be limited to medical/scientific research. It shouldn't be a commercial product. We shouldn't be devoting this amount of money, energy, land, and resources into something used primarily for chatbots, industrial-scale plagiarism, and a generalized excuse for the capitalist class to automate as many jobs as possible, all while environmental degradation is worsened in the process. And if we need to do all those other things causing actual harm just so we can potentially get some medical advances out of it, then no, it's not a net benefit to society, and given the drawbacks society would be more than justified in banning it just as we've banned a great many other things we've invented but whose costs far outweighed any actual or potential benefits, like asbestos, leaded gasoline, and CFCs. If "A.I." could cure my muscular dystrophy, but the cost is water shortages, skyrocketing electricity costs, the death of art as we know it, and the final nail in the coffin of trust in each other, then I'd rather spend my remaining years suffering from the MD.

Just because something is some new technology doesn't make it inherently good. There's even less reason to trust that tech when profit is on the line and executives and shareholder benefit from trying to force it on everyone. In that post of mine I linked to I referenced Ian Malcolm's monologue in Jurassic Park. We all know the famous line: "Your scientists were so preoccupied with whether they could, they didn't stop to think if they should." A radical technological advancement (genetics, in that case) was used not for the betterment of society, but because someone saw a business opportunity and thought he stood to profit, and people needlessly died in the process.

While JP was a sci-fi film, it and others like it serve as allegory. They're cautionary tales. We as a species keep hurtling headlong into new technological advancements with little to no concern about the consequences of that technology.  In fact, so much of our technology was purpose-built not to save or improve lives, but explicitly to destroy lives. Much of human history has revolved around devising ever more efficient ways to kill as many people as possible, "because we needed to do so before our enemies did." We learned to split the atom, and the first thing we did was create weapons that could destroy entire cities, then we built enough of those weapons to render ourselves extinct. Mutually assured destruction. "MAD." Possibly the most apropos acronym ever.

Granted, that's obviously not all technology. We've invented a great many things that were or are more than a net benefit to society: the steam engine, aircraft, radio, telephones, television, antibiotics & vaccines, and of course the electric light bulb, among many others. But history does show that technology is just as often used for ill as it is for good, either deliberately or accidentally. Sometimes some new invention or gadget is just plain pointless and has no real benefit, but mostly harmless. Sometimes technologies that had actual utility are in many ways being actively made worse in some way or another for the sake of squeezing more profit from end-users. And sometimes we found out well after the fact that a particular technology, while beneficial in many ways, was also incredibly harmful.

To expand on that last one, when we did find out that a technology was causing serious harm we were previously unaware of, the businesses that profited off of that technology refused to do anything about it, because they felt they stood to lose money if they didn't maintain the status quo. Fossil fuels are the biggest example of that. Yes, the technologies surrounding fossil fuels did allow humanity to advance considerably. It made the Industrial Revolution possible, and it resulted in increases in living standards around the world. But once it was figured out that burning those fuels was causing damage to the environment and was harming public health, instead of investing in new forms of energy, these business leaders and their political allies have for decades been like "We should do nothing because the line must go up." They'd rather pretend climate change is some sort of hoax than do anything to alter their business model. Meanwhile, our continued inaction on the matter has resulted in a situation that's getting worse, to the point where even if we fixed the problem today we're not going to be unscathed.

We saw similar things with other technologies. Take refrigeration. Amazing technology that qualitatively made life better by greatly limiting food spoilage. However, the gases used for early refrigerants were incredibly toxic. To their credit, after some fatalities from exposure to toxic refrigerants in the 1920s, chemical companies did make a good-faith effort to find something non-toxic, which led to the development of CFCs in the 1930s. As far as anyone could tell, it was the perfect solution, as CFCs were non-toxic and seemingly inert. But in the 1970s we started to find out that CFCs were in fact not inert, and were destroying the ozone layer, which we kinda need. But by this point, companies like DuPont no longer cared. They weren't about to invest more money into finding another new refrigerant just because some scientists discovered CFCs destroyed ozone. They eventually had to be forced to phase out CFCs by the Montreal Protocol, one of the few times world governments actually got together to do something to actually protect the environment. We actually fixed a problem! Had that not come to pass, those companies probably would have been more than content to continue producing CFCs and not giving a shit about the effects to the environment and public health that would have resulted from worsening ozone depletion.

But what real benefit is there to society to be had from text-to-image/video models or LLMs? All we've seen actually manifest has been invariably negative. A deluge of slop that threatens to strike at the heart of something so central to humanity that we've been doing many millennia before the first civilizations arose, all so companies in many different sectors, most notably but not limited to entertainment, can find some sort of justification for firing as many people as possible (personally, I think the fact that things created by "A.I." cannot be copyrighted is a good thing, and I hope that extends to anything using "A.I." in any part of the process). A deluge of fake news further undermining trust in a society that already has a trust deficit, something actively being encouraged by the current fascistic American administration. A deluge of unreliable chatbots, which have literally driven some people mad, yet provide nothing that couldn't be answered by a quick jaunt to Wikipedia. And all of it is produced by soulless machines trained on the collective creative history of mankind, taken without permission, of course, which last I checked was theft. Machines that are some of the worst resource hogs of anything ever invented. Yet all some people can see out of Plagiarism Bot 3000 is "More money! Lower costs!" or "It saved me a bit of time!" Generative "A.I." is not only a solution in search of a problem like so much of the pointless tech being foisted on consumers these days that exists purely to bilk more money out of gullible, easily-impressed consumers, like coffee pods or smart toasters (or 90%+ of any "smart"/IoT gadgets). It's also a problem in and of itself.

Technology for technology's sake is not some good or noble thing, and the profit motive necessarily motivates businesses to ignore any negative externalities they produce. The greed, covetousness, complacency, apathy, and wickedness of man knows no limits, and we should always be wary of those with great wealth who seek to commodify all things and who view the world solely through the lens of profit. While I'm not a religious man, this basic observation of human nature has been well understood since ancient times and well understood by a great many faiths, just as it's understood by the philosophers, historians, authors, and activists of today, whether it was Saint Paul, who said "for the love of money is the root of all evil," or Upton Sinclair, who said "It is difficult to get a man to understand something, when his salary depends upon his not understanding it." We shouldn't let the starry eyes some of us get from the latest gee-whiz gadget distract us from this truth. This isn't some creature comfort or toy we're dealing with here, like a microwave oven or a DVD player or a Furby (I'm old; sue me). This is something that is already having a negative impact on lives and society in general, and could make things exponentially worse in the future, with seemingly no real benefits emerging. And in the face of that, well, as far as I'm concerned tech companies and their bottom lines can go to hell if they want to force this shit on us.

Anyway, I think I'm done here, and I've said all I care to say on the matter. Arguing about things involving existential threats to mankind is not helping my anxiety any. If any of you want to get all excited about this crap, that's your prerogative. Have at it. The only reason I came in this thread in the first place was to concur with one user and correct another.

To steer the topic back to DLSS5, I guess the big question I have right now is currently Nvidia only showed examples of an AI creating images that are relatively close to the source graphics, my question would be is there some limit on that? For example if the developer wants Grace in RE9 to look like a photoreal version of the actress Jennifer Lawrence (for example) instead ... can they give the DLSS model photo data of Jennifer Lawrence that it would spit out a final image that looks just like Jennifer Lawrence?

My guess is yes it can do that already, they just did not want to anger game artists and throw employees of a studio into a rage, but that's where this will go. 

Once that happens that means effectively you can probably change just about any kind of source input image for the model to output something based on whatever photo reference/training data you give the AI algorithm. And what that means is basically you can change the graphics to any kind of photographic reference you give it. 

Right now it looks like they just gave it probably some kind of broad "make the image look more real" parameter set that's likely trained on a shit ton of images/faces/outdoor nature images/fabrics/materials, etc. 

The first and perhaps most obvious use of this will likely be sports games where it will be used to get player models to basically look just like their real life counterparts. 

I think the issue with "it should only be applied to x domain" is that these tools ultimately have multiple applications. 

For example, Alpha Proteo, which again can't be understated in how its changed research into drugs and other protein research, uses generative models to generate candidates which then are checked against physical ground truths. These sub-modules aren't very different in architecture from diffusion models that produce video. So obviously one could imagine a reality where the protein use-case came first and then it became apparent that these models could also be applied to video generation. 

Similar can be said of Alpha Geometry and language models. 

It is true, you're not going to solve social problems solely with technological solutions. Likewise, you're not going to prevent social ailments solely by blocking technology. 

Sooner or later you actually have to strike the root of the problem and solve the social problems directly. Everything else is beating around the bush and tinkering with the margins. 

There are only two ways they could do this.

1. They pre-trained the model on generalized image-to-image. This is unlikely for a few reasons. Good general image-to-image models are relatively huge. The open-source ones start at around 13 billion parameters. That is not feasible to inference in real time even on a single data center GPU, let alone gaming ones. An RTX 5090 inferences on these models about 2 images per second, just for context. The datasets used to train them are also huge. Nvidia doesn't have access to any buffer data on these data samples like they do with their regular DLSS training sets. Now Nvidia could train an (or more likely source an already trained) image to image model and use it as a teacher for their specialized gaming specific model. But there are two issues with that. The first is that it would very much skew the codomain so much that you are risking the efficacy of your gaming specific model. The second is that it is a very inefficient method given the target objective is so specific. 

2. They have invested heavily in model interpretation research and pulled off something like Claude's Golden Gate Bridge experiment but for image models rather than LLMs. If that were the case, they'd be able to allow much more control than you are talking about. You really don't need text or image inputs in this case, and can just directly control the model parameter weights. See: https://transformer-circuits.pub/2024/scaling-monosemanticity/index.html

This model is more likely something like what is described in this paper, 

https://arxiv.org/pdf/2105.04619

but without using the G-buffer at all (if we take Nvidia's press release at face-value that they only use color and velocity buffers) and probably using a vision transformer instead of a CNN. 

Those two things are not the same for one and for two that critique is correct. The quality of textile has fallen, for example. As for what AI is, I am just saying it as i see it. I don't run an AI marketing firm so I feel no need to try to market it for them. 

I wasn't disagreeing that it was correct. My point was that the situation we find ourselves in now is very much analogous to what the original luddites found themselves in. Textile production was no less intellectual and creative than any of the current jobs being outmoded. It's only thought to be so because the labor was devalued with the introduction of industry. 

I really don't believe in the concept of unskilled labor. The automation of physically involved labor and intellectual labor is the same in my opinion. There is a lot of creativity in physical labor just like there is in intellectual labor. And we're starting to discover that it is harder to automate the last few pieces of physical labor which haven't been automated than it is to automate a lot of intellectual labor, because embodied intelligence is lagging behind other fields of AI and automation. 

Gen AI farms are not that. They said Gen AI farms do nothing good. The quantitative science analyzung big data bots are not gen AI. Its just deep learning technology that has been a thing for a while and they don't need huge data farms. 

This is the issue with smooshing a lot of tech together. Its for some tech to provide cover for the bad tech like gen AI. 

Deep learning tech in research and medicine did not cause worldwide disruptions mor are what is advertised. What is advertised is based on shit like Grok, Chat gpt, sora, etc

Alpha Proteo is a generative model and biologists need to use TPU/GPU time to inference with it (if they don't have an HPC locally.)  These TPUs/GPUs exist in "AI Farms." 

A generative model doesn't suddenly become non-generative because it is used for protein generation and not video, text, or image.