True Positive Weekly #51

The most important artificial intelligence and machine learning links of the week

Hey, in this issue: machine learning becomes a mathematical collaborator; humans find AI-generated faces more trustworthy than the real thing; how Uber predicts arrival times using deep learning; artist uses AI to perfectly fake 70s science fiction pulp covers; the unified ML framework, enabling framework-agnostic functions, layers, and libraries; and more.

This issue is sponsored by AICamp.

• Machine learning becomes a mathematical collaborator

• [SciAm] Humans find AI-generated faces more trustworthy than the real thing

• Deep reinforcement learning algorithm completing Tekken Tag Tournament at highest difficulty level (video, code, docs)

• DeepETA: how Uber predicts arrival times using deep learning

• [Sponsored]  A two-day ML community event for AI developers and data scientists

Join thousands of developers, engineers, data scientists, applied researchers at the ML Summit. Learn from 40+ AI/ML tech leads from DeepMind, Meta AI, Nvidia, OpenAI, Spotify, Twitter, HuggingFace, LinkedIn, AstraZeneca and more about building real-world NLP, GNN, and ML systems in production.

• EleutherAI releases weights for GPT-NeoX 20B and a tech report (tech report, code, slim weights, full weights)

• ML training compute has been doubling every 6 months since 2010 (paper)

• Artist uses AI to perfectly fake 70s science fiction pulp covers – artwork and titles

• [Project] The unified machine learning framework, enabling framework-agnostic functions, layers, and libraries

Enjoy the newsletter? Help us make it bigger and better by sharing it with your colleagues and friends.

Have a nice weekend! See you next week. — Andriy