@CopyRebeldia So we can now generate infinite proper 3D environments for robots to train on from all images and videos easily scrapeable online
English
Laszlo L. Mari 🇪🇺
925 posts
Laszlo L. Mari 🇪🇺
@laszlolm
Every story has two sides. Leading an extremely talented team at Dakai, past & current clients: Google, Spotify, Binance, Solana & many others
Abu Dhabi Katılım Nisan 2014
909 Takip Edilen526 Takipçiler
Hoy una industria entera dejó de tener sentido.
Un tío publicó en GitHub un repo que convierte cualquier foto en un mundo 3D explorable: meshes con físicas, splat del fondo, audio ambiente. Todo.
Una imagen entra. Un mundo sale. Cinco minutos.
La gente que se pasó diez años aprendiendo Blender lleva todo el día mirando esto en silencio.
Se llama image-blaster.
Español
Laszlo L. Mari 🇪🇺 retweetledi
wtf did i just read
NATO is equipping live cockroaches with AI models to spy on enemies, steering their movement with electric shocks to the nervous system!
the tech is fucking insane:
- each cockroach is wired with cameras, microphones and microscopic AI chips that process data locally.
- swarm algorithms coordinate MULTIPLE cockroaches at once
- these cyborg cockroaches are sent on military scouting missions moving through tight spaces, rubble undetected.
- german military has already paid for and deployed these AI cockroaches
i’ve heard of ai drones but never ai-powered fucking combat cockroaches lmaoo
Rowan Cheung@rowancheung
NATO is testing live cockroaches as AI-powered spy drones. Incredible AI engineering, but also something I kinda wish I hadn't learned about: > Swarm Bio-tactics wired real cockroaches with electronic backpacks containing AI hardware, radios, cameras, and microphones. > Cockroaches are steered by sending electrical signals directly into the insect's nervous system > They can crawl through rubble, tunnels, and spaces where drones can't fly, and troops shouldn't go, transmitting data back the entire time. > Within one year, they went from concept to field-validated systems with paying NATO customers, including the German military. The qualities that make them useful for military recon (small, silent, nearly undetectable) are exactly what make them creepy. ...International laws weren't written with cyborg insects in mind.
English
@dronedoommetal Focus on why the other parts don't hurt instead of the one that hurts gurus would say
English
Laszlo L. Mari 🇪🇺 retweetledi
Humans and octopuses evolved eyes independently, but we got ours with optical nerves attaching to the front of the retina and octopuses through the back so we have a random blindspot and they don’t.
Ryan Moulton@moultano
Anatomy is full of cables that were connected 3 hundred million years ago and are tied in knots because they can never be disconnected.
English
Laszlo L. Mari 🇪🇺 retweetledi
Did you know that when they say stuff like "The A18 uses TSMC's 3nm process" or "announced the 2nm node"
The 3nm, 2nm actually doesn't mean anything?! It's just like a version number. They make it up. Literally nothing measures 2nm or 3nm.
I certainly didn't know.
English
Laszlo L. Mari 🇪🇺 retweetledi
@ivanhzhao @garrytan @NotionHQ Can't import my Apple Notes images when importing notes which are sometimes more important than the text itself -- also doesn't have proper Apple Shortcuts support (at least when I last tried)
English
Is there a “Perplexity for Notetaking” yet?
I realize I take a lot of Apple Notes but per Apple being not very good at software anymore, the chance of it becoming smarter anytime in the next 24 months is zero
Eg if I paste a link I want the important or interesting parts inline expanded, and I want it to be a part of my RAG queries
Arguably this should be a part of memory in Claude or ChatGPT
English
@MostlySpecial @youkhna I got mine recently and yes, can confirm it works perfectly with my tesla
English
another Vibe Draw demo! we are LAUNCHING SOON, comment "DRAW" for a link to our waitlist!!!
martin@martinsit
we built Cursor for 3D modeling.
English
things will get truly weird when European companies will start stealing from Chinese companies
Linus ✦ Ekenstam@LinusEkenstam
This active suspension could be the best in the world. This experiment clearly demonstrates how smooth this ride is. This is an executive vehicle. Period.
English
After all this effort coding becomes writing tests for the AI 😅🙈
Georgi Gerganov@ggerganov
pack it up boys, it's over
English
@andreasklinger 4D chess is 4D chess. We'll not grasp incentives while he is in power, we will probably be able to dissect it later.
English
On one hand Elon Musk says
- "multi-culturalism dilluting" everything
- to protect Germans from immigration
- encourages smaller countries
- and not listen to the global elite.
On the other hand
- he is an immigrant
- fathered multiple kids with other immigrants
- pushes for more immigration into the US
- would never want the US to split up
- and is as global elite as it gets.
Trying to be as open-minded here as possible - but wtf?
Can this be explained outside of "some races are bad" mindset?
English
@shl the middle ages hanging pictures in front of every store would like a word
English
@youkhna I had a Volvo, the smaller key can get you stranded somewhere, it discharges without notice. Dealership told me to only use it when I go into water / beach or in risky tourist areas where thiefs probe cars with the buttons on the normal opener - Minimal phone pre-orderer here :)
English
Laszlo L. Mari 🇪🇺 retweetledi
This is the most important paper in a long time . It shows with strong evidence we are reaching the limits of quantization. The paper says this: the more tokens you train on, the more precision you need. This has broad implications for the entire field and the future of GPUs🧵
Tanishq Kumar@tanishqkumar07
[1/7] New paper alert! Heard about the BitNet hype or that Llama-3 is harder to quantize? Our new work studies both! We formulate scaling laws for precision, across both pre and post-training arxiv.org/pdf/2411.04330. TLDR; - Models become harder to post-train quantize as they are overtrained on lots of data, so that eventually more pretraining data can be actively harmful if quantizing post-training! - The effects of putting weights, activations, or attention in varying precisions during pretraining are consistent and predictable, and fitting a scaling law suggests that pretraining at high (BF16) and next-generation (FP4) precisions may both be suboptimal design choices! Joint work with @ZackAnkner @bfspector @blake__bordelon @Muennighoff @mansiege @CPehlevan @HazyResearch @AdtRaghunathan.
English
