Thought #30 - Ministers, Memory and Microsoft’s Billions

From AI ministers to billion-dollar bets, this week showed just how wide AI’s reach is becoming.

Hi lovely humans,

It’s been a week where AI cropped up in politics, bread making, and even the local chip shop. At the same time, the biggest players doubled down on their promises to the UK economy - with eye-watering investment announcements that raise as many questions as they answer.

We’ve also seen the biggest study into ChatGPT use yet - which I’ve summarised in a blog.

Our Week in AI

A reminder of failure (and I still don’t like GPT-5)

I’ve shared the blog I wrote on the ChatGPT paper.

My first step was to share with ChatGPT to summarise.

Firstly, it couldn’t open the link. So I downloaded as PDF, and uploaded the document.

ChatGPT: Here’s a structured summary of the NBER working paper you shared (No. 34255, Sept 2024):

Title and Authors

“The Long-Term Consequences of Early Childhood Lead Exposure”
Authors: Patrick Baylis, Rhiannon Jerch, Shinsuke Tanaka, and others.

This is not the paper I shared was actually called How People Use ChatGPT.

I switched the model to GPT-4o, started a new chat and it was able to give me a very useful summary.

But a reminder: always check the links you share and the docs you update can actually be accessed and read by these AI tools. As they can sometimes lie hallucinate what they have been given.

AI New Releases

Anthropic has been busy

  • Temporary chats are now available in Claude (available top right)

Claude Incognito Mode

It’s worth noting this is likely because Anthropic updated their data sharing to be opt out in September.

  • Claude memory for Teams - Anthropic has brought its memory features for using on Teams subscriptions, making it easier for organisations to keep context across conversations.

  • Claude web fetch in APIs - Developers can now let Claude pull in live web data.

Perplexity adding useful tools

AI News

Not Quite News, But Worth a Read (or Listen or Watch)

  • Creepy AI toys - a Guardian family’s unsettling week with an AI toy.

  • The Atlantic’s new AI Watchdog - “The Atlantic’s ongoing investigation of the books, videos, and other media used by the world’s most powerful tech companies to train their AI models.”

LinkedIn AI Poll

Last week we asked if people always check their AI output:

More than half do (apparently). I do now - having learnt the lesson the hard why copying some awful code in early days.

Vote in this week’s poll

This week we want to know if you are using AI outside of work and how.

Final Thoughts

As always we hope this was helpful!

Feel free to share this with anyone who might find it useful.

Next week, more blogs and resources.

Laura
Always learning