We’re entering one of the most exciting—and possibly world-shifting—moments in human history. That might sound grand, but I don’t say it lightly. Over the past few months, AI researchers like Demis Hassabis (DeepMind CEO) have started talking about something that just a couple years ago was considered speculative: self-improving AI systems that might not just learn from humans, but improve themselves—exponentially.
If that holds true, it won’t just revolutionize coding or science—it could change nearly every aspect of our lives including how long we live.
What’s Happening Right Now in AI?
Let’s simplify it into three powerful trends:
1. Beyond Large Language Models (LLMs)
LLMs like ChatGPT can generate human-like text and solve many types of problems, but they’re still largely reliant on human training data. They “learn” from what we give them.
But now, researchers are testing a different approach: letting AIs train themselves.
2. The Self-Improving Loop
Back in 2017, DeepMind’s AlphaGo Zero shocked the world. Without any data from human games, it taught itself to master the game of Go—beating the previous best AI 100 to zero. No hints. No teachers. Just pure self-play and feedback.
This same technique—self-play and reinforcement learning—is now being applied to more complex tasks, like software coding and even reasoning. For example, in a recent research project, two AI models played the roles of “problem proposer” and “problem solver,” continuously challenging and learning from one another. As they improved at coding, they also got better at math—without being explicitly trained on it.
It’s a feedback loop: the more they learn, the better they get at learning.
3. Scaling Up: The Hardware Push
AI companies like Google, OpenAI, and NVIDIA are now throwing enormous computing power into scaling this self-learning process. Instead of just pre-training on data, the next big wave will involve massive reinforcement learning—essentially, letting AIs experiment, learn, fail, and get better in simulation over and over again.
Think of it like building not just a smart assistant, but a curious child who never sleeps, never forgets, and never stops learning.
Self-Improvement Meets Reasoning: The New Frontier
Demis Hassabis recently introduced AlphaEvolve, DeepMind’s new project combining reinforcement learning (RL), evolutionary programming, and large foundation models. In simple terms, they’re creating AI systems that don’t just absorb data—they experiment, refine their own algorithms, and get better without us telling them how.
The holy grail? A continuous learning loop. A model proposes challenges. Another model solves them. They both learn, iterate, and push each other forward. And early experiments suggest this loop not only helps with coding—it actually boosts math and reasoning too.
Think about that: improving at one skill (like coding) without supervision also makes the system better at entirely different skills (like solving math problems). That’s a level of generalization we’ve never seen before.
It’s not hype. It’s already happening.
Why This Matters for Your Health—and Your Future
Here’s where it gets personal.
Hassabis has said that these next-gen AI systems—once combined with enough computing power—could help cure all major diseases within a single decade. Yes, cure.
That’s not wishful thinking. That’s based on what’s already happening:
- AlphaFold (another DeepMind project) cracked the structure of over 200 million proteins, accelerating drug discovery at a speed no human lab could match.
- AI is now designing molecules, simulating entire biological systems, and optimizing treatments in ways that used to take years—if they happened at all.
- And now, with reasoning-based self-improvement, we’re talking about AIs that don’t just process known facts but generate new hypotheses and test them.
We’re no longer just asking “What disease does this person have?”—we’re asking, “How can we eliminate this class of disease altogether?”
Lifestyle Implications: More Time, More Possibility
If you’re over 50, there’s a good chance you grew up thinking about retirement as a 20-year stretch of slowing down. But if we truly are at the start of an AI-driven longevity revolution… what if you get 40 or more healthy years?
That’s not a theoretical question. It’s a deeply personal one.
What would you do if you had more time, more vitality, and more tools at your fingertips than any generation before?
Would you:
- Reinvent your career around purpose?
- Travel, teach, or mentor?
- Explore your curiosity and passions with the energy of someone decades younger?
In my view, this is the real opportunity—and the challenge. The old model of “retire and wind down” is being replaced by one where you clearly define your purpose, stay engaged, learning, contributing, and growing.
But we’ll need a new playbook. A new vision for what a super long life should look like. Mind you, while it’s possible medical advances solve all diseases of aging, it will likely still be many years of approving and making these advances widely available. At the same time, in my opinion, the public would demand access and governments and corporations would have to oblige.
Final Thoughts: What’s Coming Next?
We’re watching the convergence of two AI “trees”: narrow superhuman systems (like AlphaGo), and more generalist systems (like ChatGPT). Now, they’re merging—combining self-improvement, reasoning, and real-world adaptability.
Demis Hassabis believes AGI—Artificial General Intelligence—could arrive just after 2030. Others say sooner. But what matters more than the date is this:
The foundation is being laid now. The loop is already forming. And the implications are nothing short of transformational.
So here’s my invitation to you:
Don’t sleepwalk into the future. Pay attention. Stay curious. And start imagining how you want to spend your extra time—because odds are, you’re going to get a lot more of it.
Live long, live well and prosper!
Michael
[see links in the above text for references]
