yanik

@varien relighting splats isn‘t an issue. result will depend on the source‘s lighting though, as shadows etc are baked in. but if captured under soft lighting it can work well

@Dayanch96 @comet

Perplexity’s @comet browser and Safari on my ancient iPhone

@tekbog yes, been saying the same thing

idk how to explain this but openclaw is for the same people who use notion

@gmoneyNFT i can understand why many people seem to draw this conclusion, but it’s a bad analogy. and the quoted numbers are based on api pricing, not actual cost.

This reminds me of when you could get a 30 min uber in nyc for like $10/$15, and now it costs $50 to get 5 blocks. We’re going to look book at this time and wish the vc’s would subsidize our compute again.

@scaling01 wasn‘t going to chime in, but this is the third or so post on this i‘ve seen; do none of you read anymore?? he explains it further down; single 4090 for 32B distill, dual 4090 for 70B distill… the tl;dr is clickbait

that's how much "research" "investors" do 1 trillion param models now fit into 48GB VRAM thank you, but where is the trash?

@charles_maddock the suggested next steps you show the user. if you put the user through extra steps in the onboarding to get accurate context, i feel like they should (and could) be more tailored to the user - getting that right is what takes most time for such features tho

@yanikschaffer How do you mean more specific? So the user understands, or more model context? Their is a hidden field with much more info for the model as it builds workflows

I made an onboarding that researches new users and generates personalised workflows You can confirm/deny info it found about you, then get 15 high-value Strawberry use-cases to try Worried it's too long/creepy though, thoughts?

@jacobrintamaki 👀

I could not find one that exists so I am making the robotics paper reading gc. Please comment or DM if you want to be added.

@asallen what i was thinking the second i saw it. seems so obvious and intuitive, but for some reason no one seems to have thought of it before

How did it take us this long to see this possibility?

@orenmeetsworld could be a black pro mist filter (or similar) on the lens if done in-camera. similar effect can also be achieved in post - pretty easily even

I need someone to explain what’s happening with this camera look this is so good

@alojoh @Tesla @Reuters only misleading if you don‘t know the difference between the EU and Europe

Here is a clear example of how Reuters mis****s their readers about Tesla. @Tesla @Reuters 1) Reuters reports that "BYD outsells Tesla" on latest data released this morning for August. 2) However, the information is cherry-picked. 3) It is true that BYD sold a few more units than Tesla in the EU area (9,130 vs 8,220, 11% more) but Reuters denies their readers the full picture. 4) The full picture is that in Europe which includes the EU, Iceland, Norway, Switzerland and the UK Tesla outsold BYD significantly: 14,831 vs only 11,455. BYD sold 23% less than Tesla. 5) This data was part of the same data release! Reuters had to have and saw it. This was intentional. 6) This means Reuters cherry-picked a smaller subset just to be able to fabricate a negative headline. This is one of their go-to methods to sm*** Tesla. This dece****ion doesn't end here. Reuters makes no mentioning of important context: 1) BYD's sales are driven by large fleet sales (unlike Tesla's). 2) Also, Reuters makes no mentioning of Tesla's delivery wave strategy: preliminary data for September (and data to which Reuters has access) imply Tesla is poised to grow European sales by around 100% month-over-month in September and thus will significantly more outsell BYD. 3) The reason why Reuters doesn't tell their audience is that it would not fit the "Tesla bad" narrative. This makes Reuters compl**** in the br***was***ng which led to horrific outcomes like C****ie K***'s assa***na***n. [Why is AJ not spelling out certain words and uses *: unfortunately, these words cause the X al** to flag a post and inflict a db on my account limiting a post's distribution. If you want to help get truth out like/share this post. Thanks]

@teknium i think we‘re almost at a point where you can generate a movie; but it would be terrible. in two years it‘ll be much faster, cheaper and easier - but i‘m not yet sure whether it‘ll be much better (not talking about visual quality, consistency etc)

When this says AI generates any movie on demand in just about 2 years, I think this paper is just wrong. How far have we actually come in video? We still generate like 4s videos, mostly without sound, with quality thats still pretty lacking, and thats after 3 years since image gen (which you can kind of just make into video gen with more and more frames). I'm not saying that progress is linear, but we should have seen a much larger exponential if we expect full 2 hour movies with audio and traditional movie quality in 2 years.. I dont know how difficult longer context is with video/diffusion models, but my guess is its not much easier than LLMs (and maybe is even harder/more expensive...)

@Mascobot @a16z x.com/iamgingertrash…

🚨 New: We built @a16z's personal GPU AI Workstation Founders Edition - 4x NVIDIA RTX 6000 PRO Blackwell Max-Q (384GB total VRAM) - 8TB of NVMe PCIe 5.0 storage - AMD Threadripper PRO 7975WX (32 cores, 64 threads) - 256GB ECC DDR5 RAM - 1650Watts at peak (runs on a standard 15Amp/120V circuit). For training, AI research, and deploying models locally. A datacenter-class AI rig you can keep under your desk. We are planning to make a limited number of these a16z AI Workstations. Build guide + how you can make your own 👇

@internetvin @sama big opportunity @sama

Hey @sama, I’m in SF this week, let’s link up. I’m the founder of Brampton, one of the most exciting AI models in the world, I think we can help OpenAI a lot amigo, like a lot, imagine all of the talent from Brampton working on your technology, just imagine it man.

@hud_zah huge alpha. been doing this for years

if ur not writing ur ideas down in the shower with a waterproof notebook you're leaving so much alpha on the table it’s my new ultimate life hack

@ns123abc x.com/iamgingertrash…

Elon was right. “They don't actually have the money”

@mbalint @khoomeik x.com/iamgingertrash…

@khoomeik Who is building the hazmat suit to traverse the future internet?

it’s so over

@khoomeik interesting how almost all the comments on the pika post are positive… there should be more people questioning stuff like this

@benhylak @Jaguar @a16z one of the strongest portfolios in the game at this point!

since apple, i've had the chance to work on designs for a few iconic brands (including @jaguar) today, i'm excited to announce the new look of @a16z