A week in Generative AI: Meeker, Tidbits & Robots

News for the week ending 1st June 2025

Jun 01, 2025

We’ve had an unsurprisingly quieter week this week after the blockbuster of a news week had last week. No major news from the frontier AI companies, but we have had the return of Mary Meeker with her first AI Trends report and I’ve also shared a few other smaller tidbits of interesting news for you all as well.

On the ethics front, there’s still lots of reporting around the copyright issues surrounding generative AI models and Anthropic’s CEO has warned that AI could eliminate half of all entry-level white-collar jobs in the next 2-5 years.

It’s also worth checking out Dan Sinker’s long read on The Who Cares Era and Stratechery’s article on AI agents.

Mary Meeker is back with her AI Trends Report

For those who haven’t heard of Mary Meeker before, she’s a venture capitalist who has been called the “Queen of the Internet”. From 1995 through to 2019 she published an annual Internet Trends Report, which was the most highly anticipated annual report for tech investors. She stopped in 2019 when she started her own VC firm, Bond, but is back this year with her first AI Trends report!

As is her style, this is incredibly in-depth and incredibly insightful. It comes in around 340 pages and covers everything from usage, to costs, to revenue and investments, and all the way through to the impact on work and the economy.

TL;DR - AI is evolving faster than any other technology we’ve ever seen - across users, usage, hardware, investment and global adoption. It’s reshaping everything from geopolitics to work and presents both huge opportunity and huge risk. Below are some highlights:

AI is now viewed as a lever of geopolitical leadership, akin to the space race.
The main drivers that are driving AI’s rapid evolution are compounding developments:
- Training data has grown in volume at an average of 260% per year for the last 15 years
- Available compute to process training data has grown at an average of 360% per year for the last 15 years
- Algorithmic efficiency has improved by an average of 200% per year for the last 10 years
We’re starting to see the mainstream adoption of generative AI models with ChatGPT growing its user base tenfold in the last 2 years. Users have more than doubled in the last 6 months.
ChatGPT hit 365 billion annual searches in 2 years. It took Google 11 years to reach this milestone.
75% of CMOs are currently using or testing AI tools
Data Centre electricity consumption has jumped from 2% to 4.5% of total US electricity consumption in the last 5 years.
Conversely, the energy required for a large language model to generate a token has fallen 105,00x (?!) over the last 10 years.
This has led to the costs of using large language models to drop by 99.7% in the last 2 years.

Unfortunately there is a bit of AI slop in here where they’ve used ChatGPT to predict what it will be able to do in 5 years and 10 years time. I’m not saying the suggestions are completely ridiculous, I’m just surprised they included speculation like this from a generative AI model in such an otherwise very comprehensive and solid report!

Source

Some Tidbits from the week

Whilst it hasn’t been a week of headline news, there have been some interesting tidbits dropped throughout the week I wanted to share:

Hugging Face unveils two new humanoid robots

Just over 12 months ago Remi Cadene moved from working on Tesla’s Optimus robot to head up open source robotics at Hugging Face. They also purchased Pollen Robotics back in April to help with that push.

This is a preview video of Hope Jr, the fruits of their labour - an open source humanoid robot that costs under $3000. They also showed off Reachy Mini - a desktop robot that can move its head, talk, and listen for testing with AI apps, which will cost around $300.

The whole ambition with Hugging Face’s push into open source robotics is to make the area cheaper and more accessible and they’re certainly doing it with these two models. I’d love to have a go at building my own robot and play around with integrating it with large language models - Hope Jr might be this year’s Christmas present!

Source

AI Ethics News

Long Reads

Dan Sinker - The Who Cares Era
Stratechery - Claude 4, Anthropic Agents, Human-AI Agents

“The future is already here, it’s just not evenly distributed.“
William Gibson