AI spiritual bliss, future of government, analyst clones, agentic half-life, and more

May 26, 2025

“The point is not AI becoming conscious. The point is humans realising they already are—and have forgotten.” - James Breeze

AI driving the future of government

Over the years I have done quite a bit on the future of government. Interest is stepping up at the moment, with two keynotes on the topic in the space of a few weeks. As I share below, governments must be more adaptive and agile, but not at the expense of public accountability.

The new Claude 4 models consistently gravitate to spiritual bliss. Hopefully we can learn from them.

Be well!

Ross

 

📖In this issue

  • Framing: Creating the Future of Government

  • UBS rolls out AI avatar clones of its analysts

  • AI agent half life as a metric for AI performance improvement

  • A thorough analysis and framing of AI agents vs Agentic AI

  • Podcast: Bruce Randall on energy healing and AI, embedding AI in humans, and the implications of brain-computer interfaces

💡Framing: Creating the Future of Government

On Thursday I gave the opening keynote at the BiiG - Business Improvement and Innovation in Government - Conference to 700 public servants on Creating the Future of Government.

The theme through my session was around how government can be both accountable and stable as well as agile and evolutionary. It is a steep challenge but there is no other option - we must do this as well as we can. Governments of today and tomorrow are not fit for governing the accelerating world of today. And there are many wonderful examples from around the world of government institutions rising to the challenged, some of which I shared in my keynote. One notable one is how the UAE’s Regulatory Intelligence Office is using AI in regulation and policy development.

There is work to be done.

🧠🤖Humans + AI

UBS rolls out AI avatar clones of its analysts

UBS has created avatar clones of 36 of its equity analysts to deliver client briefings and updates. Currently the analysts approve all scripts and videos, but it's easy to imagine this going a next step to where it's not just video generation, but emulating the analysts' expertise.

AI agent half life as a metric for AI performance improvement

ecent research showed that every 7 months AI doubles the length (in human time taken) of the task they can solve. AI researcher Toby Ord has built on the original study to show that AI success probability declines exponentially with task length, defining model capabilities with a ‘half-life.’

A thorough analysis and framing of AI agents vs Agentic AI

AI Agents and Agentic AI are not the same thing. A useful review paper analyzes the literature to offer a structured conceptual taxonomy and application mapping to clarify the distinctions, use cases, and challenges

🎙️This week’s podcast episode

Bruce Randall on energy healing and AI, embedding AI in humans, and the implications of brain-computer interfaces  

Why you should listen

This was a bit diffferent from usual in the guest being a Reiki Master and practitioner of energy healing as well as AI and BCI advocate, it was an interesting conversation.

Thanks for reading!

Ross Dawson and team