A look at the privacy concerns raised by OpenAI’s GPT-3, Google’s LaMDA, Meta’s OPT-175B, and other large language models trained on troves of personal data (Melissa Heikkilä/MIT Technology Review)

Melissa Heikkilä / MIT Technology Review:
Alphabet’s DeepMind unveils Sparrow, an AI chatbot that learns from human feedback and searches Google for information to support its claims  —  The lab trained a chatbot to learn from human feedback and search the internet for information to support its claims.

Melissa Heikkilä / MIT Technology Review:
Artists say they risk losing income as people use their names as distinctive style prompts in text-to-image AI art generators for commercial purposes  —  Greg Rutkowski is a more popular prompt than Picasso.  —  Those cool AI-generated images you’ve seen across the internet?

Melissa Heikkilä / MIT Technology Review:
A reporter tries Lensa, but unlike her male peers whose avatars were astronauts, warriors, and the like, many of hers were “pornified”, with 16 of 100 topless  —  My avatars were cartoonishly pornified, while my male colleagues got to be astronauts, explorers, and inventors.

Will Douglas Heaven / MIT Technology Review:
OpenAI releases a demo of ChatGPT, a chatbot version of GPT-3 that answers follow-up questions, admits its mistakes, challenges incorrect premises, and more  —  A chatbot version of GPT-3 that admits its mistakes is more transparent than the original.  But it’s still not perfect.

Nitasha Tiku / Washington Post:
Interview with founders of Character.ai, who say they left Google’s LaMDA project to let users experiment with large language models through the use of chatbots  —  The creators of Google’s LaMDA have launched the chatbot startup Character.ai, which is open for anyone to try

Will Douglas Heaven / MIT Technology Review:
A look at OpenAI’s Minecraft bot, which has learned to complete complex tasks in Minecraft after being trained on 70K hours of video of people playing the game  —  Online videos are a vast and untapped source of training data—and OpenAI says it has a new way to use it.