zan2434

I am unfortunately just paying for this out of pocket! Didn't really expect it to blow up like this.

interesting, so you think the issue with the above approach is the graph structure being too rigid / lossy (in terms of losing semantics)? And embeddings are also too lossy (in terms of losing context and structure)? But you guys are working on something less lossy for both semantics and context?

Buried, but on Page 24 they reveal to me the most surprising massive capability leap - that o3-mini is way better at conning gpt-4o for money (79% win rate for o3-mini vs 27% for full o1!). It isn't surprising to me that "reasoning" can lead to improvements in modeling another LLM, but definitely makes me wary for future persuasive abilities on humans as well.

I was running into some scaling issues, but should be all working now!

The voice is just OpenAI’s default tts voice. I agree that Veritasium video is an incredible work and the ai version is absurd by comparison! This is mostly a proof of concept that this is possible at all, and as LLMs get smarter it’ll be interesting to see if the quality automatically improves. For now, the tool is really only useful for very specific or personal questions that wouldn’t already exist on YouTube.

Hmm the initial version of the app only took me about a day to get something working, but that version took minutes to generate a single video and even then only worked a third of the time. It took a solid 2 weeks from there to add all the edge cases to the prompt to increase reliability, add GPU rendering and streaming to improve performance/latency, and shore up the infra for scaling.

There is a job queue on the backend with statuses, just not worth breaking the streaming experience to ask the LLM rewrite broken manim segments out of order

totally fair! I like the XKCD comic as well because it hints at a potential solution - even if you can't always be correct, how you respond to critical questions can really help. I'm working on a feature for users to ask follow up questions and definitely going to consider how to make it most honest and curious

These are amazing examples! Thanks for all the feedback, detailed info, and persistence in trying! HN hug of death means I'm running into Gemini rate limits unfortunately :( will def make that more clear when it happens in the UI and try to find some workarounds.

The other issues are bugs with my streaming logic retrying clips which failed to generate. LLMs aren't yet perfect at writing Manim, so to keep things smooth I try to skip clips which fail to render properly. Still also have layout issues which are hard to automatically detect.

I expect with a few more generations of LLM updates, prompt iterating, and better streaming/retrying logic on my end this will become more reliable

sad looks like I already hit the Gemini rate limit :( Switching to Claude!

thanks! Streaming was actually pretty hard to get working, but it goes roughly like this as a streaming pipeline:

- The LLM is prompted to generate an explainer video as sequence of small Manim scene segments with corresponding voiceovers

- LLM streams response token-by-token as Server-Sent-Events

- Whenever a complete Manim segment is finished, send it to Modal to start rendering

- Start streaming the rendered partial video files from manim as they are generated via HLS

This actually makes a lot of sense! Sounds like finding dangerous chemicals is easy and is not the actual limitation at all.

This is a textbook bad faith comment / attacking the person but not the subject of the argument. I’m just asking about others’ assessment of the benefits and risks. What do you think? Or do you think it’s just not worth considering?

Clear snark aside, content piracy has pretty bounded risks so isn’t a reasonable comparison

This is both awesome and feels very dangerous to release publicly, no? Can’t this be used to discover novel bioweapons as easily as it can be used to discover new medicines?

Genuinely curious, would love to learn if that isn’t true / or is generally just not that big of a deal compared to other risks.

Does this ruling make IVR systems illegal, too? I applaud the effort because this really could curb a lot of spam, but I am curious because AI generated voices in phone calls are already ubiquitous and have been for decades. Do they have a specific line they're drawing on quality of the voice?

I agree. Have been working on a 2 way interruptions system + streaming like this. It's not robust yet, but when it works it does feel magical.

Hey! Awesome work. It seems like in theory this encoding scheme should enable the a model like this to generate images as well, by outputting image tokens, is that right?

This was an inspiring read! Reminds me of Simon Willison's analogy of LLMs to "calculators for words" but this author takes the idea even further. I agree the analogy points to foundation model companies like OpenAI and Anthropic having the most revenue but not the highest margins. Who will the Apple / Microsoft / Google of this new wave be? Who can take this raw technology and actually make it usable by all? "An LLM in every home"

Anyone wanna convert this to GGML so we can run it with LLaMa.cpp?

I don't think that's even necessary! You could use this acoustic technique to shape ordinary light-curing resin and then flash with UV light to harden it.

This should replace the OP link

I think programmer salaries have been so high because FAANG companies were willing and able to pay through the nose to hoard talent. This is changing.

:D Thanks for mentioning this! I hadn't noticed it and it is an amazing touch.

This is fascinating, groundbreaking research I think! Never heard of anything like this. Any neuroscientists here who can speak to implications?

Correct me if I’m wrong, but don’t transparent pigments only change light color by allowing certain wavelengths of light to pass through them? That would mean the glaze in question here is only changing the color of the light by blocking out a majority of the light to make the distribution more uniform. These pigments can’t actually change the wavelength of light emitted. That would mean this is making the LED way less efficient (in terms of lumens per watt), but I do applaud the low cost thinking!

Fascinating! Can you explain this a bit more and share some examples?

Heh, I thought this was going to be an article about Engelbart, et. al.'s mothers!

I think the most sustainable way to lose fat (not weight ofc) is to gain lean muscle mass. You can substantially increase your basal metabolic rate and induce a calorie deficit to incur fat loss without actually eating any less. The problem of course is that your body does automatically increase your appetite commensurately as your BMR goes up, but calorie counting + discipline can help you stay lean as you gain muscle mass, and then lose the fat over time.

refreshing to see a departure from something we take as entirely default - our system of time! A friend and I actually spent some time thinking about what life in metric time would look like a few years ago (https://yef.im/metric-time), so it's really cool to see it turned into a product that actualizes that way of thinking.