[dead]
Sora 2
911–916 of 916 posts
Re: Sora 2
#912Earlier quoted context omitted.
At least now we know that AGI is definitely not happening.
Can you explain a bit more? I'm intrigued by your comment, but not seeing the connection
Someone else put it best, so I'll direct quote them:
>If you genuinely believed you were 2-4 years away from AGI, is Sora slop really the thing you'd release?
Re: Sora 2
#913Earlier quoted context omitted.
Myopic view. Just because your motives may be pure and your greed may be a non-issue doesn't mean that multinational corporations who have aligned themselves with fascist regimes (Meta, Alibiba, etc -- point is that they're demonstrably amoral/immoral) do not intend to exploit the tech to its fullest extent. However having said that, the intention/aim need not be to deliberately replace creatives. That is not the cla…
I like how you start with "myopic view" the tell me to use my "big boy brain" to see the ease of creation across the board going down as a negative thing. The purpose of the society was never to pay money to people, and if we figure out how to get grains of sand to replace skilled labor, there's no amount of greed that can outpace what it will do for humanity.
You are making a straw man argument. I am all for the benefits of AI; I can be pro-AI while still acknowledging the negative impacts.
Re: Sora 2
#914Earlier quoted context omitted.
[flagged]
>It’s fucking video made by a computer after you type a sentence Geez. Breathe a little bit. It's always weird to see somebody who had zero involvement in the creation/engineering/design of a product so disproportionately defensive about it. The concept of text to video since LTX, CogVideo, AnimateDiff is closing in on two years of development at this point so there's naturally going to be a little less breathless en…
> It's always weird to see somebody who had zero involvement in the creation/engineering/design of a product so disproportionately defensive about it.
Why is this weirder than the guy who goes out of his way to post negatively about them 90% of the time?
Re: Sora 2
#915Re: Sora 2
#916Earlier quoted context omitted.
I think you're overestimating how much power LLMs consume. Let's say one video pegs a top of the line Blackwell chip at 100% utilization for 10 minutes. I think a Blackwell chip (plus cooling and other data center overhead) is somewhere around 3000 watts when running 100%. So that's about 0.5 kilowatt-hours. I suspect this is a severe overestimate because there's probably a significant amount of batching that happens…
> Let's say one video pegs a top of the line Blackwell chip at 100% utilization for 10 minutes. Where do you get this from?
I was roughly using estimates for as single Nvidia GB300 NVL2 superchip, which has about 1.3 TB of memory (which I think is probably ballpark the size of the distilled models that are being used for inference).
Timing-wise, I'm making an extremely liberal estimate. (In reality, Sora 2 takes about 2 minutes to generate a video in my testing).
The thing that matters here is ultimately the product of time and memory (the latter of which I use as a stand-in for calculating number of chips being used), so I'm more or less asserting that the models being used to generate the free Sora 2 videos have less than 6 trillion parameters (since 2 minutes being 1/5 of 10 minutes give us a 5x multiplier we could apply to 1.3 TB), or less than 24 trillion parameters with aggressive quantization. I can explain why I think parameter count is less than 6 trillion if you'd like, but it's pretty hand-wavy intuition.