Artificial Intelligence #104

Hey, in this issue: a self-supervised algorithm that works for speech, vision, and text; Google's new Transformer-based model; a new approach for a visual explanation of classifiers; a curated list of practical metric learning and its applications; a Q&A-focused rival to GPT3; when should someone trust an AI assistant’s predictions?; and more.

This issue is sponsored by Everynet and Valohai.

• [Meta] The first high-performance self-supervised algorithm that works for speech, vision, and text
• [Google] LaMDA: towards safe, grounded, and high-quality dialog models for everything
• [Google] StylEx: a new approach for visual explanation of classifiers
• [Project] ?A curated list of practical metric learning and its applications
• [Sponsored] Stop data starvation with low-cost sensor data acquisition for ML & AI

The lack of data is one of the main reasons why both machine learning and AI projects fail. Data starvation puts multimillion-dollar investments into these projects at risk. Learn how Everynet’s low-cost connectivity enables the use of up to 99% of sensor data that is currently being discarded.

• AI2 shows off an open, Q&A-focused rival to GPT3 (and more)
• [MIT] When should someone trust an AI assistant’s predictions?
• [Sponsored] Free MLOps eBook: Looking to learn the basic concepts of MLOps? This is the book to start with
• [Paper] LaMDA: Language Models for Dialog Applications
• [Tutorial] How vectorization speeds up your Python code

The newsletter is now read by more than 550,000 subscribers. If you are building an AI or a data product or service, you are invited to join the list of?these great companies?by becoming a sponsor of one of the future newsletter issues. Feel free to reach out to?[email protected]?for more details on sponsorships.

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