Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Karpathy’s Pelican

twitter.com

Post reply on HN
Print view View original

181–190 of 461 posts

Re: Karpathy’s Pelican

#181

Post by sinaatalay »

On consumer devices, AI communicates with us through speakers and screens. Screens are the richer medium, so most consumer AI innovation will happen there.

Computer graphics will have enormous applications because they are directly controllable by LLM-generated code. Video models are probabilistic and less suitable when precision matters. In education, for example, we need exact visuals. If an AI wants to plot y = sin(x), it should generate the precise graph through computer graphics rather than approximate it with a video model.

Re: Karpathy’s Pelican

#182

Post by try-working »

this is not a good benchmark for models, but it's great if you're optimizing for attention on twitter because video content and 3d animations perform best on social media.

a real benchmark is instead running evals on your own traces, and building a cost/quality/speed profile for models based on real workloads. but it doesn't get you a shiny video you can post on twitter.

Re: Karpathy’s Pelican

#183

Post by techblueberry »

fzeindl wrote: post #18

Regarding the argument about LLMs having difficulties auditing their work: I wonder whether we are entering the era of throwaway software. Just like cheap plastics and improved processes has enabled us to rapidly manufacture anything we want for a very low price, maybe LLMs give us the same for software. Produce it cheaply and if it breaks throws it away and reproduce it.

To a certain extent to what end though,

On the one hand yes, almost every task I work at now is one off one of scripts I throw away.

The question is - where does the software the spec or the “code”.

A really complex game will probably always be token heavy. At least for the next few years code is still not free.

But certain software is just iterative by design. If we mean we regenerate all the for loops of a game from scratch, sure but I think “code” Is really more spec then implementation, and we’ll want to continue building things through iteration.

And even on the for loop point - Do you really want to spend millions of tokens rewriting a game every time you need to make balance changes?

Re: Karpathy’s Pelican

#184

Post by garganzol »

trvz wrote: post #173
garganzol wrote: post #162

Earlier quoted context omitted.

The product you've built the animation for is only for macOS, which is a pity cause it has a way wider appeal than macOS market share is. You have low-hanging fruits right there, don't miss them.

macOS covers 90% of people who would ever pay him.

This is a bizarre claim to make in this case, AI tech is used on all platforms universally.

Re: Karpathy’s Pelican

#185

Post by duxup »

sinaatalay wrote: post #181

On consumer devices, AI communicates with us through speakers and screens. Screens are the richer medium, so most consumer AI innovation will happen there. Computer graphics will have enormous applications because they are directly controllable by LLM-generated code. Video models are probabilistic and less suitable when precision matters. In education, for example, we need exact visuals. If an AI wants to plot y = si…

Coding, graphics, all seem to have very defined background data for an LLM to base their decisions on, test, and even when I correct it we're all on the same page. Seems like there's lots of room for AI productivity there.

What I find funny is that approximate to computer graphics are video games. When I ask AI about a decision available to me in a video game AI completely fails, OFTEN. I assume all the forums and changes made to a game over time might be quite confusing for AI. But I've also seen it completely make up characters and decisions and weapons and so on about some very clearly defined games and paths. It's an interesting dynamic.

Re: Karpathy’s Pelican

#186

Post by emp17344 »

Supermancho wrote: post #178
emp17344 wrote: post #138

Earlier quoted context omitted.

Better than the mean human? The mean human creates far better stories while daydreaming. AI enthusiasts have such a distorted view of human capability, it’s bizarre.

The qualification "better" is doing some hefty lifting in your assumption. What you think is better is not what I think is better. Imagination is not storytelling. You ask the "mean human" to write, it's going to be worse than an LLM in spelling and grammar, if you get anything at all.

Daydreaming is absolutely storytelling. The average person is able to craft elaborate stories on a whim. Writing is used, in part, for conveying stories, but writing is a separate skill. The good news is the average person can learn to write well, but LLMs, so far, have not demonstrated the ability to craft interesting stories and it’s pretty unlikely they’ll be able to learn, as this is an unverifiable domain.

Re: Karpathy’s Pelican

#188

Post by altmanaltman »

charcircuit wrote: post #164
jmugan wrote: post #93

A lot of people are posting here about how bad the end product is, but that is kind of the point. Models have moved beyond generating images to a new kind of benchmark that better exposes understanding of the physical world, and we can use benchmarks like this to measure future progress. (Of course, it will have to be a qualitative/subjective measurement.)

Bad? It has a charming style. I would watch the whole book if it was made like this.

Very few things are universally hated. One can love something truly that is hated by most. But it doesn't change the fact that it's still hated by most. An objective and a subjective opinion can exist at the same time on this.

Re: Karpathy’s Pelican

#189

Post by thejazzman »

garganzol wrote: post #184
trvz wrote: post #173

Earlier quoted context omitted.

macOS covers 90% of people who would ever pay him.

This is a bizarre claim to make in this case, AI tech is used on all platforms universally.

iOS users spend dramatically more on e-commerce. I worked in e-commerce. Maybe it’s changed in the last 5 years. But that’s where the ops sentiment comes from.

Re: Karpathy’s Pelican

#190

Post by fasterik »

beepbooptheory wrote: post #171
fasterik wrote: post #148

Earlier quoted context omitted.

Taking a single paragraph of literary text, which is abstract and ambiguous, and converting it into a 3D animation requires an enormous amount of implicit knowledge about spatial relationships, intuitive physics, everyday objects, and so forth. Not to mention the mathematics of 3D transformations and computer graphics more generally. Saying that it's indicative of no more than three.js coding ability is absurd.

Why does it require knowledge about spatial relationships?

Because the degree of realism in a video is determined by implicit knowledge of concepts like "in front of", "behind", "next to", "inside", "outside", "between", "occluded by", etc. as well as distances, angles, and the relative size of objects when viewed from a perspective.
Post reply on HN