Thought #29 - New Chips, New Cash, and New Cracks

Hi lovely humans,

It’s been an odd week in the AI and data world - new money, new tools, and some strange new challenges for the companies building AI.

The big one for me is Anthropic admitting (on Reddit, no less) that users who’d said Claude felt degraded weren’t imagining it - two bugs have now been identified and fixed. Again, with the Reddit voices causing big change in the AI world.

Otherwise we have a few small but impactful new feature releases, lots of money being invested (including in AI movies by AI companies) and OpenAI trying to solve the AI job crisis.

Our Week in AI

Holiday Tips

I decided to follow some ChatGPT instructions this week (note: do not do this without sense checking, apparently mountain rescue in the Dolomites are having to rescue people doing ridiculous trails suggested by ChatGPT). I used it to find some food places in Montreal and end up at Schwartz Deli. The queue was enormous (unsurprising, given it was popular enough to appear from such a vague prompt).

However, there was a takeout option which took 2 minutes and the sandwich was yum. A mild win.

Where it really shone was a specific:

L: Where in La Mauricie National Park am I most likely to see bears?

To which it gave a really long answer, ending with a suggestion to go on a specific evening bear tour which I hadn’t found. And we saw a Black Bear:

A massive win - and unlikely we would have found this tour or seen a bear without ChatGPT.

AI New Releases

• Claude can now create and edit files - spreadsheets, presentations, and more. A big leap towards everyday productivity use.

• ChatGPT adds a GitHub connector - for easier coding workflows.

• ChatGPT’s SharePoint connector gets an upgrade.

AI News

• Anthropic admits bugs behind recent Claude glitches after weeks of user complaints.

• Anthropic raises $13bn at a $183bn valuation - while Mistral secures €1.7bn. The race for capital shows no sign of slowing.

• OpenAI launches a job board for AI-related roles.

• Microsoft in talks to license Anthropic’s tech for M365. Significant, given Microsoft’s existing investment in OpenAI.

• Perplexity extends secure access to US government agencies.

• OpenAI backs an AI-made animated feature film.

• Claude conversations are appearing in Google Search results - echoing similar moves with Gemini and ChatGPT.

• ARM unveils a new AI-focused chip design, promising efficiency gains for mobile and edge devices.

• Google launches an AI literacy hub - aimed at helping the public understand AI basics.

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

• AI relationships and intimacy - The Guardian explores the rise of people forming romantic connections with chatbots.

PS thank you to everyone who shared a podcast, Youtube or article with me (I’m currently on holidays but will share with everyone once I get a chance to read, watch or listen). Please do keep them coming.

LinkedIn AI Poll

A bunch of you are (like me) using AI to plan holidays:

Vote in this week’s poll

This week we want to know if you really, really check your AI output before using.

Final Thoughts

As always we hope this was helpful!

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

Next week, we’ll be back with our AI news, blogs and articles.

Laura
Always learning