I see in a not so distant future, companies like Netflix producing an increasing share of their content through generative AI. This would shift budgets from production into AI spend. Massive data centers, running sophisticated AI software managing enormous amounts of data. Cost/benefit could be an important decision factor. So the cheapest to train and to run to generate content the most profitable these companies would be. The same could be said about music, writing, etc.
On a first note, I wonder what the impact on the different industries like entertainment this would have.
Then on a more immediate way, I wonder what companies would benefit from this. Thinking on NVDA, AMD, and others. There are also other players that could have an edge here by designing cost efficient ASICs targeted at specific AI workloads. My question would be if they have the resources, the vision and ability to execute to capture this opportunity.
Have a look at Adobe Firefly recently announced, or the explosion of projects around stable diffusion which is open source, or what Meta is doing. They could start with cartoons, but they would easily move onto real characters. Mix it with unreal engine, or add a layer of generative AI to unreal engine. AI has a compounding effect. It’s just a matter of time.
Granted, realism for human like characters we are not yet there. However for anything else we are doing it already. The technology being developed by Nnvidia and others, massively parallel architectures like H100, are generating a quantum leap in compute power, and AI itself will help in find ways to improve algorithms for achieving realism with human like characters. It won’t happen overnight, but we are already on track and it won’t take decades.
The images are the least of it. The story telling is a much bigger problem. AI does not tell stories worth anything.
BTW on a related note. What are the odds call centers in corporate America are going up in cost with AI. I avoid the chat options now. Because the AI even calls itself an 'agent" and wont turn me over to a human being. Pay for AI and no one usually gets a result from it. I have to call the call center or write a human being an email instead. There possibly is hiring going on across this country for call centers.
Right now there is technology that transcribe speech to text and run AI on it to extract metadata (keywords), extract sentiment and suggest potential answers as well as information that the caller could potentially benefit from (e.g. the manual for a device, terms and conditions, nearest branch or pickup location, etc). There are studies that show that the younger generations prefer to chat to a bot vs with a human, as it is available everywhere and they don’t need to move to go to a quiet place to “talk”.
All this data is then analyzed to understand why people are calling, what are the top reasons, what product/service features are failing that generate the most problems, and then use that to prioritize future launches, new products or service enhancements.
I see all this really helping to scale. It would never replace a well-educated human interaction, granted, but it allows to deliver products and services at a much lower price point, as the after sales costs are now lower.
Creativity needs editing. That is where AI gets lost. Wrong answers need editing. Of course, there is more than one type of AI. Much of it is not creative.
Just over a year ago none of this was possible. The pace of improvement is astonishing. Why some people cannot see the train coming at them is beyond me. Looking forward to what you and others like you can bring next.
Oddly, those CLOSEST to the technology are often the ones who cannot see it developing.
… well, maybe NOT so ODDLY. It’s a natural feature of experts since the framework and tables of understanding are most cemented on past experiences within the subject matter expert group.
This is why visionary leaders are so annoying to the “engineers and techs”.
This morning spoke with a friend who is an aeronautical engineer Ph.D. He sees FSD as happening in the US eventually. Talked of putting a perceptual level in the process and it will work. So far not done that way.
No need to argue with me on this. While I am on the fence it is not so much about the tech as the image qualities and more importantly the economics of selling such art. It is not at all easy.
I don’t really see the relevance of needing a human hand for storytelling and editing as far as being a massive winner in the AI space. Jeffrey Katzenberg said AI will take 90% of animators’ jobs on animated movies within three years. That’s a significant chunk of the cost. I’m sure Netflix & Disney will happily spring for a writer, editor and director when most of the cost has already been stripped.
CGI plus some AI is how it is done now. It is denied now.
The jobs are mostly computer jobs in the movie editing etc…and not human work on movie sets. The jump from CGI to AI in many crazy scene instances will happen. But even that might be unnecessary as it is now.
