OpenAI's SORA - April 2024 NEW VIDEOS!

It has been some time since we've seen some new SORA examples and here we go finally! A few of these are really really impressive, especially the rally car or the Ted Talks intro (unfortunately after many tries I found out TED is constantly blocking anyone from reacting to their intro. So I needed to cut that segment to a bare minimum - so I recommend to check it here: • What TED Will Look Lik...
Do you think this tech is here to take the jobs of artists, or is it a bubble about to burst? Let me know in the comments!

[@danzilthard.7248]

Anyone railing against TED for expanding awareness on a topic is an anti-intellectual pile of hot garbage.

[@vanisjakub]

exactly, I find it really bizarre that there is such an uproar..

[@legoenginemechanic101]

Wow you can create some very unique visuals. Some new types of illusions

[@vanisjakub]

indeed, some of them look really mind-blowing! I hope the rumour is right and they will release to public this summer! Can't wait to play around with this myself!

[@SoraTheBests]

better than all

[@olmoslara]

Why the hate? People don't see the opportunity this brings to everyone

[@R6ex]

Wow ... some weird ... some mesmerizing visuals.

[@vanisjakub]

exactly, it can be a mixed bag, but when it works, it REALLY WORKS! :-O

[@raaghavgr1990]

All this has just increased my anticipation for Sora. Hollywood producers, directors, content creators, and actors have already made millions or even billions of dollars producing films thus far. Therefore, there is no need to worry about them losing their jobs due to Sora. "Quickly release Sora and Long Live Sora".

[@Astra2]

Maybe the top 0.1%, all the rest are probably struggling to make ends meet. But this tech is inevitable and amazing, those people will need to find new jobs.

[@vanisjakub]

I guess we will see, we have already seen quite some heavy Hollywood strikes recently and this will only make it worse, depends how unionised the artists will become. But I still see it as a tool, the best can learn how to use it in their advantage and create better things for less money, and that is always a good thing (and quite literally - it is progress)

[@user-gz9wt8vj7f]

AI can also make music but hasn't replaced any artists yet

[@JMVenteremail]

Yet.

[@vanisjakub]

@@JMVenteremail this right there!

[@dsmoke1972]

The rally car morphing it's direction was straight up bad ass. Not weird, stop your whining.

[@vanisjakub]

haha you are right, it might be just that I was looking for the "most realistic looking portrayal" and even thought that it is a unintentional error by the model. But you are totally right, that if you look at this from a "artistic" POV, it is cool AF! :D

[@CoryRayGordonMusic]

Allow us our creative freedoms, corporations shouldn't be the only ones with the key ring.

[@vasipro]

The problem with AI is that they can only evolve so much because most of their progress is based on what WE did. It can't reinvent the entire world, create new worlds infinitely. At some point it will need to learn from us again. We can't become lazy. We just need to progress with AI.

[@hardikpopli]

You're mostly right. However, its only a matter of time before these shortcomings are addressed as well.

[@christian-schubert]

Not a problem with AI - rather a problem with OPEN AI (and all companies currently pursuing similar paradigms). Suggest checking out Ben Goertzel - more robust AI systems are well underway (well, at least according to him)

[@Astra2]

The world produces an absolutely mind blowing amount of data. Once we create more generalized models, they'll be able to self-improve without human intervention using that data. Right now most models use supervised learning, but I don't see any reason why that will continue to be necessary.

[@vanisjakub]

Exactly right! Seeing the progress OpenAI made in just a couple months makes me feel like the "super-intelligence" is really just around the corner and that might be the last thing we ever need to develop/create as humanity ever. Also - if these videos show something it is that we really likely live in a simulation :D

[@christian-schubert]

@@vanisjakub Not quite. So far, OpenAI's main strategy seems to be tweaking parameters, refining training data sets and throwing more processing power at their models with every new iteration. What their products in principle do is a fancy form of multidimensional "nearest neighbour matching" (grossly oversimplified). This approach can and will (...probably...) never lead to "super-intelligence", as these model do not inherently "understand" their actions and are not able to contrive truly novel ideas either. Let's assume color film had never been invented. In this thought experiment, OpenAI's SORA would still create impressive video footage - black and white video footage. HOWEVER, and this is where it gets weird, it COULD possibly conjure up a color motion picture once you prompted it with the idea of combining [color] paintings with black and white videos. Still, the process of coming up with this idea in the first place requires a human in the loop (with its successful implementation being far from an easy feat). OpenAI's models remain narrow AI models, however, since their training data is literally ALL of current human knowledge, they APPEAR to be intelligent. Also, the fact that this has the potential to replace 99% of the human workforce (99% of all jobs do not require the generation of truly revolutionary, disruptive new ideas) is truly mind-boggling.