Windows Copilot Newsletter #9
Mistral enters the game; Copilot gets a bigger 'Notebook'; and Microsoft goes nuclear...
Welcome to the ninth edition of the Windows Copilot Newsletter, where we share some of the most interesting stories from the rapidly evolving field of AI chatbots. As we approach the end of the year, things are slowing down a bit from their fevered pitch. But there’s still plenty going on…
Top News
Everyone’s talking about Mistral: French AI startup Mistral secured a bucket of new funding, putting them at a valuation of US $2 billion (NY Times). Mistral released the very innovative and highly competitive model Mixtral 8x7B, a ‘Mixture of Experts’ that rivals OpenAI’s GPT-3.5 in accuracy. Read about that here. (And read a recent, sociopathic interaction with Mistral 7B below.)
Copilot gets a big Notebook: As part of its never-ending series of enhancements to Copilot, Microsoft announced Copilot Notebook, which gives its AI chatbot a far larger 16,000-character ‘context window’, meaning it can remember far more of the details of its interactions. Read about this upgrade here.
The Imitation Game: Research from the University of California seems to confirm the hypothesis that AI chatbots are great immitators and very poor innovators. Makes sense, given that the draw from their training data sets - a form of imitation. Read that research here.
FunSearch innovates: Counter to that argument, Google DeepMind announced FunSearch, which puts the ‘hallucinations’ endemic to AI chatbots to good work - creating new mathematical proofs. So perhaps - in the right circumstances - AI chat can be innovators? Read that here.
Microsoft goes nuclear: Microsoft has decided that only nuclear power can supply the energy they need to cover the planet with data centres running Copilot. Could this be due to Bill Gates’ substantial investments in nuclear power startups? Read all about this here.
Top Tips
Brainstorming with Bard: Now that Bard’s had a proper upgrade to Gemini Pro, use it to brainstorm new ideas. Read how to do that here.
Role Play with an AI: The wonderful Benjamin Breen is back with more observations about how AI chatbots can be powerful tools for the exploration of ideas. Read it here.
Safely and Wisely
Incorrect, confidently: Which AI has the greatest propensity to make up facts? ZDNet did a study, and report their findings here.
A crisis in science: Nature reports that the misuse and poor use of AI is leading to a ‘reproducibility crisis’ - drowning good science in bad data. Read about this growing problem here.
Longreads
Sam Altman, CEO: For reasons known only to them, TIME has designated OpenAI’s Sam Altman as CEO of the Year. Read the full profile, with a dramatic retelling of the recent unpleasantness at the world’s most important AI startup, here.
Book News
We still have about half a dozen autographed copies of my new book Getting Started with ChatGPT and AI Chatbots. Grab one at the button below - and if you order by Monday the 18th of December, we’ll (very likely) have to you before Christmas!
That’s all the news for this week - we’ll be back next week with more about AI chatbots in our final newsletter of 2023.
If you got something useful from this newsletter, please consider forwarding it to someone else who might benefit.
Until then, thanks!
Mark Pesce
www.markpesce.com // Need help with AI?