Lior⚡

Anthropic released one of the best series of tutorials on prompt engineering. It's literally everything you need to know. https://t.co/NSQG4F58Yt

https://t.co/bMc1JTguok

Turns out, LLMs represent numbers on a helix and use trigonometry to do addition. A new paper reverse engineers addition in models like GPT-J-6B and finds a “Clock” algorithm. Numbers are encoded using sine and cosine terms, then added like angles. https://t.co/Ru4jkYNddl

Paper: https://t.co/d04qbE165M Code: https://t.co/RRRH4Yb1pT

One of the most impressive AI demo I've seen. This is the future of customer service. Agents that can understand text, speech, images and even live video. Soon to be all open-source. https://t.co/kKoNTgTJ1T

"The _exact_ demo is not OSS because we have deep integrations with our CES product that are not free to use, but you can find a similar OSS demo here: https://t.co/9w2wfIDHRl You can think of it as a subset of the capabilities you saw in the keynote demo and the scaffolding to be able to build something like what you saw."

Ilya Sutskever (OpenAI cofounder) top 30 must-read research papers. "If you really learn all of these, you’ll know 90% of what matters today" https://t.co/VprtIapiFF

https://t.co/60x2dU5gZK

The secret to success can be found in AI’s central algorithm: gradient descent. Take a step. Course correct. Repeat.

RT @LiorOnAI: You can now run 100B parameter models on your local CPU without GPUs. Microsoft finally open-sourced their 1-bit LLM inferen…

You can now run 100B parameter models on your local CPU without GPUs. Microsoft finally open-sourced their 1-bit LLM inference framework called bitnet.cpp: > 6.17x faster inference > 82.2% less energy on CPUs > Supports Llama3, Falcon3, and BitNet models https://t.co/AGPOsUjlyB

Intel i7-13700H https://t.co/dKEcC521uq

Apple M2 Ultra https://t.co/aNBKf8EAWp

OpenAI just released a 34-page masterclass on building AI agents. It's a must read. https://t.co/fyRwABZ9T6

Read it here: https://t.co/FqpdIpjClp

New survey paper on Deep Dive into Reasoning LLMs is on ArXiv. Summarizes the progress made in post-training and reasoning LLMs. Explores how fine-tuning, reinforcement learning, and test-time strategies boost LLMs' reasoning, factual accuracy, and alignment. Read it here: https://t.co/qk3dOxkQDZ

I’ve said this for a decade: The best AI models will be trained on stolen, hacked, and pirated data. There’s no way around it. That’s why China is winning. They don’t have privacy laws slowing them down—and they have data from 1B+ people. https://t.co/By1EwOLW2k

Google released a list of 321 real-world gen AI use cases from the world's leading organizations. Basically, unlimited business ideas. https://t.co/Hk3v3iUtfF

Source: https://t.co/ca7cd0qAu9