Windows Copilot Newsletter #14 - ChatGPT gets memory; MIT fixes memory; Gemini remembers too much
Windows Copilot Newsletter #14
OpenAI adds personal memory to ChatGPT, while MIT researchers fix the context window problem. Google warns you just how much it learns from your prompts to Gemini...
G'day and welcome to the fourteenth edition of the Windows Copilot Newsletter, where we curate all the week's most important stories about AI chabots. It's been another big week, so let's dive right in...
Top News
OpenAI adds 'memory' to ChatGPT: Tired of 'starting over' every time you start a conversation with ChatGPT? OpenAI now lets users instruct ChatGPT to remember specific details. The chatbot will also pick up personal details from its interactions with you. Privacy? Maybe not so much. Read about that here.
Copilot gains a context menu: Preview builds of Microsoft Windows 11 show a Copilot-based 'context' menu that will pop up to recommend Copilot-enabled actions, such as summarising a document. Microsoft wants everyone using Copilot, so it's adding it everywhere it can. Read about it here.
MIT fixes a memory 'bug': All chatbots have a 'context window' of several thousand characters; if you engage in a conversation longer than that window, the beginning of the conversation will 'disappear' from the chatbot's memory. Having lost its context, the chatbot will quickly become confused. Researchers at MIT have developed a technique that allows chatbots to hold the thread of the conversation for much, much longer. Read about it.
Gemini - pro, con and future: Reviews of Gemini 'Advanced' - a subscription-based chatbot from Google - place it somewhere between frustrating and fantastic. This week, Alphabet CEO Sundar Pichai let leak some powerful new features coming in Gemini 1.5, including a million-character context window, big enough to digest the entire text of 'The Lord of the Rings'.
Top Tips
3 tips to build an AI-positive work culture: Workers fear AI will make them redundant; the truth is far more nuanced. In an AI-fueled future, human wisdom is more important than ever. This feature explores how leaders can communicate this to their staff. (h/t James Hare)
Performance reviews with ChatGPT: Happily, it appears that one of the most onerous tasks within organisations - reviewing the performance of yourself and others - can be ably assisted by an AI chatbot. Read how to do that here.
Safely & Wisely
ChatGPT misdiagnoses pediatric cases: Using ChatGPT to diagnose your child's sickness? Don't. This report details research showing that the free version of ChatGPT (based on GPT-3.5) does a very poor job. Read about that here.
Don't tell Gemini anything private: ZDNet took a look at all the data Google collects when you use their AI chatbot: Pretty much everything. Data which can be used pretty much however Google wants. So be careful when you share! Read that here.
Longreads
The Birth of the Spreadsheet and the Age of AI: The Financial Times compares the current age of AI to a heady time, almost fifty years ago - and the birth of the spreadsheet. A history of the present. Read it here.
That moment you realise you've been AI-unemployed: How do you respond when a client says that unless you agree to their contract terms, they'll simply use an AI to get the job done? It happened this week - to me. Read about that here.
Understanding the Microsoft Copilot Pro Value Proposition
This week, Drew Smith and I released our first Wisely AI white paper - designed to help organisations evaluate whether the features in Copilot Pro justify handing Microsoft a hefty AUD $45 per month per person subscription fee. Read or download the white paper here.
That’s all the news for this week. We’ll be back at the end of next week with more AI chatbot news!
If you learned something useful in this newsletter, please consider forwarding it to someone else who might benefit.
Thanks!
Mark Pesce
www.markpesce.com // Need help with AI?