ai AI Video has arrived - It looks incredible

sfoster

Staff Member
Moderator
Click this link to watch some of the videos

Really incredible stuff from OpenAI
There was also a wired article, but the article was so terribly formatted with ads I'm not posting it

Instead here are the different videos from the wired article with a direct link.





 
Every time I have to deal with an AI chat bot, I end up wanting to tear out my hair and never being able to get the help I need.

Same is true with trying to report abusive people on FB, X/Twitter, etc. - all I get is a canned response that does nothing to resolve the issue but allows the platform to say it's been "handled." Same is true about banking or credit card AI bots, plane reservations, etc. Not one time have I been able to resolve the issue that I called about without hanging on for god knows how long to get a real person, even though I'd love to get done in 5 minutes by talking to the AI bot. It never actually works, except that I'm sure they get plenty of callers to simply hang up in frustration - which save employees' time, right?

100% agree with you there, it's often a facade intended to be useless, so of course it doesn't solve your problem!
And yeah it's really frustrating when businesses are disingenuous, totally agree.
 
I believe we'll see radical, unprecedented change in our lives throughout the 2030s

You know I'm a big believer in this stuff, but even *I* would not be surprised if the next 50-100 years or more continue to be limited to "phoney emotions"

<snip>

AI is about problem solving, you give it any problem (e.g. 'draw me a picture of a pitbull in a suit') it gives you a solution.

And yet, as Mara points out, "AI" is incredibly bad at giving us solutions to real problems - the kind caused by irrational humans doing stuff that wasn't built into the taught responses. If it draws a picture of a pitbull in a suit, chances are it's plagiarising an image from an 1840s Punch cartoon. In any case, these aren't problems in need of solutions, they're amusing challenges that may or may not eventually be put to good use in some other context - which is what we humans have been doing since the first Neanderthal picked up a stick and poked something with it.

I'd hazard a guess that most of here have already seen "radical, unprecedented change" in our lifetimes. I learnt to type on an old manual typewriter weighing about 10kg and where every letter had to be battered out individually onto a sheet of actual paper before it could be shared with anyone else; now I have a device that fits in my pocket, can guess what my next word will be and send my composition to a million people in a fraction of a second. But it's still damnably difficult to find any computer-assisted problem solver that can look at a mess and say "ah, I see what you did there ..."
 
And yet, as Mara points out, "AI" is incredibly bad at giving us solutions to real problems

Having a universal translator on your phone is "incredibly bad" eh

It's so funny you say AI is incredibly bad, i think its incredibly amazing.
I use photoshop now, I just click and objects disappear, i just drag and suddenly my whole scene expands with generative fill

But to you thats all incredibly bad. man we live in different worlds.
 
Last edited:
Having a universal translator on your phone is "incredibly bad" eh

It's so funny you say AI is incredibly bad, i think its incredibly amazing.

But to you thats all incredibly bad. man we live in different worlds.

Yeah ... except that's not what I said. Read it again:

"AI" is incredibly bad at giving us solutions to real problems

Maybe it is because we live in different worlds that I'm not amazed by a piece of technology doing ... well, nothing special. It's not fixing a problem and it's not doing anything new. In my world, it's perfectly normal for people from different countries to speak different languages; and in my social (and working) life, it's perfectly normal for those people to turn up in the same place and find ways of communicating with each other.

Waaaaaaaay back in the last century we had the Collins Gem dictionaries to help us out when faced with any apparent obstacle to communication; in more recent times we've had Google Translate. Now we may or may not have this GPT tool, but forgive me if I hold off on the amazement until I see it work in the real world, not in a staged promotional video. I think it highlights how useless the platform is in it's current state that you have to post the same clip twice! :D

I'll tell you what I do find amazing, though: real humans doing simultaneous translations in chaotic environments - treat yourself to a visit to the European Parliament building some day (the parliament chamber in Strasbourg is quite mind-blowing - almost like a real world version of the Star Wars senate chamber, but the pods are for the interpreters). And even better are the humans who can "translate" something in a creative - but entirely appropriate - way.

One real-world example, from the festival at which I work in July, was our public announcer a few years ago: the (native) French version was something along the lines of "Attention, Mesdames et Messieurs. Would the owner of the white Renault, registration XX-60000-YY, please move it immediately as it's blocking the disabled access" Then he switched to English and gave his message as "Ladies and Gentlemen, I don't need to translate the previous message into English because I know that no English person would be so inconsiderate as to park their white Renault in front of the disabled access. I hope the arrogant French person responsible will move it immediately"
 
Last edited:
Yeah ... except that's not what I said. Read it again:

Thats absolutely a real problem.
Are you being deliberately obtuse? A planet of 8 billion people and you can't imagine a scenario where people speaking different languages would want a free app to translate to each other?

I had a girlfriend that was indian and her grandmother had dementia and required full time care but they couldnt find anyone that could ever help out because no one local for hire could speak punjabi/communicate with the grandma (given the salary they were able to offer). if something like this existed while her grandmother was still alive, they would have been able to hire someone to help, use the translator, and it would have eased an enormous burden off her family.
 
Last edited:
Are you being deliberately obtuse? A planet of 8 billion people and you can't imagine a scenario where people speaking different languages would want a free app to translate to each other?
Nope, not being obtuse, and I don't need to imagine any such scenario because it happens all the time in my daily/weekly life - at work and at play. And that's been the case for the last twenty years.

Would a free app be of help? Well, yeah, of course. But guess what: we've had that kind of free app available in one form or another for the last twenty years. I have literally been at the bedside of someone using one in a hospital to communicate with the doctors treating his workplace injury. Years ago.

The scenario you described is exactly why AI will never deliver on its promise - because the situation (any real life situation) is too complex. Picking it apart, the real problems were:
  • the grandmother had dementia, which makes communication in any language a problem; AI can't solve that.
  • the family appears to have (had) financial difficulties; AI wouldn't make them go away.
  • the family chose to live in a country with one of the worst public health care services; AI definitely won't fix that.
  • despite there being several hundreds of thousands of Punjabi speakers in the US, none of them were apparently able/willing to assist in this situation; AI wouldn't be correct that lack of community spirit.
  • there were free translation tools available (unless this was more than 20 years ago) but for some reason, no-one used them; so why would they be any more likely to use an AI tool?
Sometimes people find themselves in difficult circumstances, and no matter what tools you give them, they can't/won't get themselves out.

A tech bro staging an interview and to show how his app can make jokes about whales and linear equations doesn't mean it's going to revolutionise human-to-human communication, assuming it can cope with real world factors like background noise, strong regional accents, slang, pidgin languages, two or three people speaking at the same time, comprenhension difficulties on the part of the listener, etc. So far, all we've seen is a marketing gimmick; and for now, all we can say is that it might be a very, very, very small improvement on Google Translate. What's there to be impressed about?
 
Last edited:
This youtube channel with Sora 'shykids' released the first mix of sora/irl video composite to make a short film


It was a boring video, I couldn't finish watching, and seems like a really uninspired use of Sora.
'Yeah lets take an expressive human face and get rid of all expression, that's sure to make our video more engaging.'

there were free translation tools available (unless this was more than 20 years ago) but for some reason, no-one used them; so why would they be any more likely to use an AI tool?
For "some reason" 😄 😄 😄 too funny.
I'll tell you why no one used real-time audio translation software in 2004... cause it didn't exist.

The first I ever heard of it being attempted was 2014 and it wasn't REAL TIME and didn't WORK CORRECTLY
Two vital components.. lol.


I remember how awful it was when you would type english into bablefish, convert it to spanish, then convert it back to english, and you would get this mangled garbage pile of crap that was a puzzle to decode.

It's like you have absolutely no barometer for how far the technology has come.
 
It's like you have absolutely no barometer for how far the technology has come.

On the contrary: it's because I've watched it develop in incremental steps over the course of forty years that I'm (a) not surprised that we've made yet another incremental step; and (b) extremely sceptical of the claims that whatever's promoted as the latest "groundbreaking" incremental step is going to bring dramatic change. Incremental steps don't cause dramatic change, and none of this "AI" is groundbreaking - it's all building on work that's gone before ... and rarely does more than simplify/improve an existing task or challenge for which there is already a working solution.

I have a mini-excavator that does a great job of digging holes in my garden, but there are still plenty of jobs for which a regular man-powered shovel is the best tool.
 
Last edited:
On the contrary: it's because I've watched it develop in incremental steps over the course of forty years that I'm (a) not surprised that we've made yet another incremental step; and (b) extremely sceptical of the claims that whatever's promoted as the latest "groundbreaking" incremental step is going to bring dramatic change. Incremental steps don't cause dramatic change, and none of this "AI" is groundbreaking - it's all building on work that's gone before ... and rarely does more than simplify/improve an existing task or challenge for which there is already a working solution.

I have a mini-excavator that does a great job of digging holes in my garden, but there are still plenty of jobs for which a regular man-powered shovel is the best tool.
Cool stilll waiting for you to tell me what equivalent software was available in 2004
 
Last edited:
Well, you're going to have to tell me what you mean by "equivalent" - I suspect our frame of reference is going to be quite different. Besides, I don't make much effort to try to remember what software I was using (or aware of) twenty years ago.

What I do know, though, is that that's the year I moved to France, and the numerous "move to France" forums at the time were full of helpful pointers to free translation tools, and there was a flourishing cottage industry of part-time interpreters for when you arrived. I later became one of those for a while, but (e.g. see previously cited example of guy in hospital) realised that the market for casual translations would fade away as Google catered for most casual needs and I had no interest in becoming a professional interpreter.
 
Well, you're going to have to tell me what you mean by "equivalent" - I suspect our frame of reference is going to be quite different. Besides, I don't make much effort to try to remember what software I was using (or aware of) twenty years ago.

Same thing I've already been saying 2-3 times now.
Accurate voice-to-voice translation in real time.

I know babelfish existed in 2004, but it didn't translate accurately, hence wasn't worth a damn, and was only text-to-text
 
Last edited:
These are not good videos IMHO. Sure, there are plenty of awkward details to pick at, but the main flaw is the approach/mentality. Ever seen Spielberg or Lean film a 5 year old (Poltergeist in example)? You don't need a great, experienced actor if you have a competent director. Really what you need is talent at orchestrating tension through elements you can control such as cinematography and editing. That music video, jesus, did they do a national search for the least talented filmmaker (and musician) in America?

I was way more impressed by the dog on the yellow windowsill.

It's costing me about $16.50 per 4090 card per month. Last month in example, I produced around 860 songs, 22,000 4k photos, 410 1500x800 videos, and beat Lies of P twice in VR. That cost 30 bucks in electricity, at 15c per kw/h. Work hours are the main cost. I expect the 90 minute film i'm making now will cost about 1-200 dollars in overhead, from opening credits to end credits.
 
Last edited:
Luma's Dream Machine available for use now, but long queues mean long waits because a lot of people are trying it.

Exciting potential, but I don't think I'll get too excited about these things unless and until they incorporate extensive editing functionality, down to the fine details—allowing you to get precisely what you're looking for. Also: no censorship of legal content in the jurisdiction of the U.S.

DREAM MACHINE 2024-06-13_045606.png



Of concern, of course, is copyright. I like how Suno AI has a nice, clear statement in their FAQs about copyright. I do not see such a statement on the Luma site. Perhaps it is there somewhere. There are statements about it in their longish TOS. Not sure I like the looks of some of it, particularly this part:

Use of Input. Customer grants Luma a worldwide, non-exclusive, irrevocable, royalty-free, fully paid right and license (with the right to sublicense through multiple tiers) to host, store, transfer, publicly display, publicly perform...[yada yada]​


Not sure what that means. Do you know? Does it matter?
 
Last edited:
Back
Top