Become a crab

“Do not mimic real human beings” may be what saves us from Terminator-esque future destruction. Real human beings kill other real human beings all the time. Maybe “don’t be like us” is how we make them better.

​@@NWRefundNot only not being who we are but not hating us for who we are. We still do that the best

Yes, best to mimic our idealized human beings verses the real ones.

path of fire, that smith makes some DAMN nice stuff, often from iron dug out of the stream on the property...

Difference is making music, painting, acting, etc does more than just create things. There are psychological and sociological benefits to those skills. You can argue that smithing, while not a necessity anymore, is now an art form. Increased artistic abilities have shown to improve math comprehension, scientific compression, reading compression, better personal psychology, and the fact that so much of it is collaborative is good for communities and therefore society as a whole. That's the worry about AI becoming the main tool for making art. Smithing as a necessity became obsolete due to the ability to mass produce goods that we needed but now it's an art form that is greatly appreciated so the lack of smithing didn't add to any kind of societal decline.

*DESTROY AI*

It's perfect for content creators, individuals or even small businesses that don't have hundreds or thousands to spend on a single B-roll or getting stock footage from shutterstock. Edit: I might start making a crap ton of B-roll and putting them up on shutterstock when SORA is released, lol.

I definitely think this is good enough for pitch-reels... and I actually think also good for pre-viz (in the right hands). The question is - at the moment, a lot of the SORA video's are taking a long time to prompt-engineer and fix in post, at some point, those two actions will be combined. I suspect it's more months than years (and certainly not decades).

That says as much about prompt engineering as it does about the underlying model. There's a whole new class of learning about how to get the results you want from the prompts you use.

Idk about visualizing your dreams 😬... it's all kinds of possibilities

Don't worry, you'll soon be able to subscribe to an flawless AI representation as part of YT (or whoever steps up through the smoking rubble of YT) that will be better than IRL. And you will be happy to pay for it... because it makes you feel complete. We have reached the End Days for reality (ok... I'm not that negative on this... it was for dramatic purposes, but the point is, if it can make money, the easiest and most profitable option will be successful. It's not dream diaries btw.)

This skill erosion has already been seen in other industries - take a look at things like blacksmithing today vs back 100 or 200 years ago. However, blacksmiths still exists today, and i think the skill while still niche, is going to remain, because some people still prefer that handcrafted feel for their objects (of course, they have to pay thru the nose for it).

User friendly as in low in requirement cost friendly.

I'd put it more like - imagine your dreams turned into a Hitchcockian movie but in a Michael Bay universe.... wouldn't that be fun to share on your socials (not that I do social media, but I can see the attraction for both stream-content and self-aggrandisement).

Until there are no studios other than IT companies able to pump out AI generated blockbusters in 1% of the budget.

If there's no one to hear it fall?.....

Yes and no. Main story-line quests will still be scripted and voiced by humans, or at least, an AI model trained specifically on a single person's voice, however, generic NPC's will have environmentally aware and general AI-generated voices and responses. Basically, it'll come down to whether or not that NPC needs to follow a specific script. Just look at the NVIDIA demo that was released a few weeks ago.

That would be like denying artists into art museums out of fear that they may use what they see to create their own work. There is legal precedence to the training data being legal, a term called "transformative." Around 10 years ago, Google 1:1 copied books into its search algorithm without the author's or publisher's legal permission and were thusly sued. The courts ruled in favor of Google, citing that the way the data was used was not a derivative of the original author's work. There was no way for the Google search engine to produce a 1:1 copy of the book, or anywhere close to 1:1 on demand, and the data was used in a manner different from the intention of the original work. Because one cannot recreate specific art pieces using AI training data simply through prompts and can only mimic the original pieces, it proves the data as transformative over derivative. The exception, I would say, is when Img2Img generation is used, where a base image is used to aid in creating something the user would like to "recreate." If a copyrighted piece is used as the base or starting image without an author's permission, that would fall under the legal definition of derivative, as the final piece cannot exist without altering the original piece. If we get too engrossed on whether AI as a whole is legal or illegal, the important arguments will get ignored. The best example I can give here is when prosecutor in a murder trial refuses to pursue any outcome other than murder in the 1st degree. If there's enough evidence to support manslaughter, but not murger, if manslaughter is not an option, the defendant will go free. Fight your battles smartly where they matter, don't let bias sway one from making logical decisions.

On paper the 4090 remains superior but more expensive. The only advantages of the 4080 Ti are price and lower tdp.

@@Gonzora Thank you. Your thoughts are dually noted! 🤟🏻

Someone will make it eventually.