Countless Wheels, Where Are the Cars?

Cambrian Explosion combined with the concept of developing wheels

As I posted last week, I made the analogy that we are witnessing an AI Cambrian explosion. It’s a sudden and rapid diversification of artificial intelligence models. Unprecedented burst of AI offerings.

Cambrian?

Brittanica.com

It speaks to the imagination but you have to see it at a relative scale. The Cambrian explosion happened over a period of about 13 to 25 million years. While animals diversified fairly quickly, the major divisions of plants, the equivalent of phyla in animals are spread out over an almost 400 million year stretch. Angiosperms, for example don’t appear until the late Jurassic, hundreds of millions of years after ferns. So with those kind of time frames, 25 million years is the blink of an eye.

Re-invention of the wheel?

Last year, the narrative was simple: ChatGPT, Claude, and Gemini. Now, Deepseek, Kimi, Metas AI, Perplexity, Grok, and others that are not directly top of mind. This is amazing, but I keep wondering. Are we seeing innovation, or just a dizzying array of technical variations on a theme?

The Cambrian explosion was a pivotal moment in Earths history. It was the emergence of diverse body plans and their ecological niches. Similarly, the AI landscape is overflowing with wheels – chatbots, language models, and AI tools that, while impressive, often feel like re-inventions of existing concepts.

I see a lot of incremental improvements in performance, tweaks to architecture, and fine-tuning for specific tasks. But where are the cars? Where are the truly transformative applications that leverage the immense potential of these foundational technologies?

Our AI Cambrian explosion is producing a multitude of specialized models. But a truly revolutionary leap would be the creation or orchestration of integrated systems. Systems that combine the strengths of multiple AI agents and or models to solve real-world problems.

I want more than just faster chatbots

I was promised flying cars in the future! No more wheels!

The wheels without cars analogy stems from a few observations.

The focus on language models as the primary driver of AI innovation has led to a proliferation of chatbots and text-based tools. While these models are powerful, they are only one piece of the puzzle. The emphasis on incremental improvements. ChatGPT 4 vs ChatGPT 4o or even o1-preview. Radical innovation comes from other companies. Helix for example. The way we organise companies can lead to optimizing existing models rather than exploring entirely new paradigms. The lack of robust integration and interoperability between different AI models has on the one hand hindered and on the other hand helped the development of complex applications. I love the idea of orchestrating and delegating different task to different Agents.

To loop back to my first analogy. Many of early life forms became evolutionary dead ends. Their unique adaptations (value propositions) failed to provide a long-term advantage.

Similarly, many of todays AI models will become fossils. Relics of a brief period of intense experimentation. The organisations that are likely to survive and thrive will be those that can experiment, adapt, evolve, and integrate with these new technologies. Not just focus on

We need to move beyond the wheels and start thinking about building the cars. We need to focus on thinking about AI systems and where they can solve complex problems, automate critical tasks, and enhance human capabilities. This requires a shift from the focus on individual models to a a view on integrated capabilities.

The future

Simple Google Search on gameplan

The future isn’t just about having LLMs or AI agents. It’s about orchestrating thousands of them across your organization. As a CTO you need to have an AI game plan that supports the strategy of the organisation. Where AI initiatives are prioritized objectively. Where you measure the impact. To both the business, to the people and the processes.

It’s not just a Tech Play

Let’s not make it a tech focus play and start without a plan. First perform a good discovery on where AI can support people and the processes they use.