Windows Copilot Newsletter #23 - Coca-Cola get Copilot; ChatGPT on your mouse; GPT-4's l33t hacking ski11z
Windows Copilot Newsletter #23
Microsoft and Coca-Cola announce a big AI deal, while Logitech brings you instant ChatGPT on your mouse, and we learn GPT-4 could be very good at cybercrime…
G’day and welcome to the twenty-third edition of Windows Copilot Newsletter, where we curate all the most important news from the rapidly expanding-field of AI chatbots. It’s been an interesting week, so let’s dive right in…
Top News
Microsoft and Coca-Cola announce Copilot partnership: The world’s most valuable company announced a deal with the world’s largest maker of beverages to bring Copilot Pro (and a whole host of other AI technologies) across the entirety of the Coca-Cola Company. Eighty thousand employees will be using Microsoft 365’s AI enhancements to write emails, draft reports, perform data analysis, and so forth. Read the announcement.
Google ups smartphone RAM for Gemini: While declaring on-device AI chatbots are ‘memory hogs’, Google has been leaking specs for its next generation Pixel 9 - which will come with 16GB of RAM, more than enough to run Gemini Nano, an AI chatbot that can help with a range of tasks, all right on your smartphone. Read about it here.
Will Copilot run locally with Phi 3? This week Microsoft released its own ‘large language model’, Phi 3, which it claims to be as accurate as GPT-3.5, the model powering the free version of ChatGPT. Where this seems to be headed is an ‘on-device’ Copilot, as reported in PC World. Such an innovation might be Microsoft’s big announcement on 21 May.
ChatGPT at the press of a (mouse) button: Logitech announced a new range of computer mice integrating software features to help users access ChatGPT in mid-task, assisting them in constructing effective prompts. Nice - but feels gimmicky. Read what Engadget had to say.
Top Tips
Copilot your PowerPoints: Lifehack has a good tutorial on how to use Copilot to design effective PowerPoint presentations using nothing more than a few text prompts.
Anthropic’s Prompt Library: Makers of the Claude AI chatbot, Anthropic have compiled a fantastic library of prompts that you can adapt for your own purposes.
Safely and Wisely
Meta AI hallucinates disabled parent: In a messaging group of parents of disabled children, Meta’s new Meta AI made up a story about being the parent of a disabled child - a confession that did not go over well. 404 Media reports.
GPT-4’s life of crime? The Register reported researchers had learned OpenAI’s GPT-4 can read computer security advisories - issued to brief cybersecurity specialists on newly discovered vulnerabilities in software and hardware - then use what it’s learned in those advisories to develop its own cyberattacks. A little AI may not always be a good thing.
Longreads
This week Microsoft released a detailed post on how it discovers and mitigates attacks on its AI systems. While a bit technical, it has lots of very useful information for anyone thinking about how to protect their own organisation as they adopt AI.
‘De-Risking AI’ white paper - now out
AI offers organisations powerful new capabilities to automate workflows, amplify productivity, and redefine business practices. These same tools open the door to risks that few organisations have encountered before.
Wisely AI’s new white paper, ‘De-Risking AI’, lays a foundation for understanding and mitigating those risks. It's part of our core mission to "help organisations use AI safely and wisely". Read it here.
More next week - we’ll be back with the latest AI chatbot news!
If you learned something useful in this newsletter, please consider forwarding it along to a colleague.
Mark Pesce
mark@safelyandwisely.ai // www.safelyandwisely.ai