The Evolution of AI: From Models to Systems
The true power of AI lies not in models, but in systems. Recent advancements by DeepMind and OpenAI demonstrate this shift towards intelligent loops of testing and feedback. These systems can run thousands of options, learn from what works, and compound their intelligence.
Iterative Intelligence
DeepMind’s AlphaEvolve represents a significant milestone in AI development. It moves beyond one-shot model outputs to self-improving systems that deliver provably correct, real-world algorithms. This “algorithm breeder” spawns thousands of code variants, benchmarks them, and breeds the winners, repeating the process to achieve remarkable results.
- Reclaims 0.7% of Google’s global compute
- Shaves 1% off training Google’s core Gemini AI models
- Beats a 56-year matrix-math record
The focus has shifted from having the biggest model to running the fastest closed-loop experimentation stack. This approach is more agile and confers a new source of competitive advantage.
OpenAI’s Codex agent follows a similar strategy, automating software development through sandboxed tasks, real-world feedback, and continual self-improvement. Progress is now compounded in validated loops, not just raw model weight.
Rethinking HR and Technology Integration
Moderna’s decision to merge its HR and tech functions into a single command highlights the need for a new approach. This change reflects the understanding that task architecture is now the main driver of cost, speed, and compliance.
The role of future executives will require fluency in both organisational design and AI system engineering. Companies that fail to build this hybrid competence will struggle to leverage AI effectively.
The Geographies of AI Adoption
There’s a significant divide in AI optimism between societies. An IPSOS survey revealed that eight in ten Chinese citizens are hopeful about AI, compared to barely a third in the United States. Optimism translates into action, with the UAE investing heavily in AI infrastructure.
Belief in the potential of AI drives sustained action, which is crucial for abundance. Without this optimism, progress is hindered.
Recent Advancements and Applications
- Researchers have made breakthroughs in quantum computing, storing three-state and four-state quantum bits in a single superconducting device.
- Neuralink’s third patient with ALS can now control a computer by thought, improving their quality of life.
- Mastercard has launched payment cards that allow users to pay in stablecoins.
- An octopus-inspired soft robot works without wires or processors, using suction and fluid flow.
These developments showcase the rapid progress in AI and related technologies, transforming various industries and aspects of life.
Conclusion
The future of AI is not just about advancements in models, but about creating systems that can learn, adapt, and improve continuously. As industries and societies navigate this new landscape, optimism and investment will be key drivers of progress.
Today’s newsletter is sponsored by ElevenLabs, a platform for building low-latency AI voice agents that sound human and can handle unpredictable conversations.
