🤯 ComfyUI Advanced Live Portrait 🔥 github.com/PowerHouseMan/…

Should AI be aligned with human preferences, rewards, or utility functions? Excited to finally share a preprint that Micah Carroll Matija Franklin Hal Ashton & I have worked on for almost 2 years, arguing that AI alignment has to move beyond the preference-reward-utility nexus!

It is remarkable that anyone can now train a 124M parameter LLM in about real-time on a MacBook M3. So easy to experiment. This would have been the stuff of dreams when I was in school. I ❤️ training neural nets, but I really admire the people who build the hardware.

Will the deserts be full of solar panels powering gigawatt GPU clusters for AI

🎶 FLUX that Plays Music [FluxMusic] 🎧 Jupyter Notebook 🥳 ⚖ Model License: Apache-2.0 🔥 💾 Base Model: 14.5GB VRAM 💿 Large Model: 19.5GB VRAM 💽 Giant Model: 25.8GB VRAM Thanks to Zhengcong Fei ❤ Mingyuan Fan ❤ Changqian Yu ❤ Junshi Huang ❤ 📄paper:

!

My first collaboration on diffusion models - VQDM! yandex-research.github.io/vqdm/ With the right kind of quantization, SDXL and SDXL-Turbo compressed to 3 bit weights can offer the same quality as previous 4 bit methods. And, of course, 4 bit quality is better with VQDM🙂

Timelines more than a couple of years

★ Dynamicland's new website documents ten years of progress toward a humane dynamic medium. dynamicland.org

Who’s founder mode

Some extra context here: if things don’t work out & team Ilya build get acquihired the price tag >> $1bn and investors get paid first If dollars mostly go to compute as is likely acquirer can use that compute too So more downside protection than you may think

Idiot generated content will be replaced by artificial idiot generated content tbh

What is the closest equivalent of the -sref function in MidJourney in the open source image model world

Anil Kapoor representing AI = actually Indians

"crypto : not your keys not your crypto :: ai : not your weights not your brain" - Andrej Karpathy

Jhakaas.ai is available you know just sayin’

Big paper I have been waiting for: what are the real impacts of AI on programmer productivity? It is a randomized controlled trial using the older, less-powerful GPT-3.5 powered Github Copilot for 4,867 coders in Fortune 100 firms. It finds a 26.08% increase in completed tasks.

Apparently Flux can sort-of generate hearthstone cards :) Prompts in ALT text

Andrej Karpathy says as we expand our brains into an exocortex on a computing substrate, we will be renting our brains and open source will become more important because if it's "not your weights, not your brain"

Most criticisms of generative AI seem to focus on imagining it needs to be a chef coming up with recipes versus cooks following them. Most tasks needs cooks not chefs. Maybe a sous.