243: The AI Gear Shift
AI's value to speeding up discovery is more than its value in performing tasks and jobs of today.
It’s taken me a while to write a post - some weeks (months) I get bogged down with a feeling of insignificance - who cares what I think? And then I remind myself that a part of the reason why I write is for my own clarity. Writing should be an act of humility, not arrogance…
The Slow Hand of Evolution
One of the XKCD cartoons that has always stayed with me is this one below, which takes a metaphysical view of time and tries to outline how given enough time, you can pretty much create the universe, so to say.
It’s a reminder that today’s chips are compressing billions of operations into a second to do the amazing things we take for granted such as pulling up a phone and checking to see where we are on a map.
It’s also a reminder of the layers and layers of innovation that have taken place to get us to where we are today.
The Acceleration
I’m reading a fascinating book: The History of Intelligence, by Max Bennet. (An aside - we were on a walk one recent holiday and I mentioned to Karuna that it would be really interesting to write a book about the history and evolution of intelligence. Then we walked into a book shop, and there it was, the book I wanted to write! Only it was probably much better than what I could have done…)
In the book, Bennet talks about the evolution of the early brains and how simply having 2 functions - ‘move forward’, and ‘turn’ - were the foundations of brain development - as the first ‘moving’ organisms looked to either find food or avoid danger. This simple evolution from the radial anemone like organisms possibly took a hundred million years. Remember that life on earth dates back to some 3.4 billion years. Nature was exactly like the XKCD cartoon above. There were no time constraints. No pressure to get to the industrial revolution within 3.4 billion years, or humans within 3.2 billion years. So evolution almost worked like that comic strip. Rock by rock, only not in such a directed manner. But driven by random chance, the selfish instincts of every living thing, and the survival of the fittest species.
And here we are today, locked in a frantic race between nations, businesses, and people to be the first to get to AGI, or super intelligence, or whatever definition we assign to the next level of intelligence. On the one hand there’s a massive artificiality to the timeline itself as it’s completely human driven. On the other hand, it’s a natural evolution in and of itself to get to this point of acceleration.
This slide from Mary Meeker’s 2025 presentation shows the number of new features released by the AI majors in one week. Yes, admittedly it was a specific week when Google had their big event but there will probably be many such weeks in 2025, and there seems to be something or the other being released most weeks.
People still often ask me whether the AI revolution is a real thing. Is it significant? Is it really going to transform everything? The honest answer is only history will tell. But every sign we see today points to it being incredibly significant. It is very much a general purpose technology compared to the Printing Press, Electricity, Computing, and the Internet.
The AI Gear Shift
When you look back on evolution, you see 2 phenomena playing out. One is the constant evolution that takes place every single time a living thing reproduces and introduces a bit of genetic mutation.
The other is the occasional gear shift when one of the millions of such mutations leads to something special, that not only provides a discontinuous change, but in fact, it makes everything else go faster from that point. The evolution of the brain, and the evolution of the pre-frontal cortex are examples of this gear shift.
Within the range of human innovation, general purpose technologies, as they are called, are not just significant in and of themselves, but they also provide acceleration to everything else that follows. All significant communication technologies are responsible to some extent for some level of gear shift.
According to some historians, the reason Homo Sapiens won out over their neanderthal cousins, was their ability to communicate and organise better. Language was a huge accelerator because it allowed ideas to go from the mind of one human to another. Especially abstract ideas like ‘the sky gods are angry’. The printing press allowed ideas to move from one human to another without them being in contact with each other, and across time. The internet allowed these ideas to spread at lightning speed across the world, marking the death of distance as far as ideas were concerned. Computing allowed us to outsource an unprecedented amount of cognitive functioning to machines. And AI is combining reasoning, calculation, and language to enable superhuman cognitive workloads. The real gear shift of AI isn’t the amount of human work AI can do, but the ways in which it can accelerate research, discovery, insights, and future breakthroughs. As this list below shows:
Even as we look around you can see the impact of AI on other technologies - whether it’s material sciences, or mixed reality, or cybersecurity, or robotics - especially robotics. It’s improvement on steroids.
In the XKCD cartoon above, imagine what it would be like if he could arrange a billion rocks and rearrange them a billion times in every second. And then train the rocks to rearrange themselves in newer and faster ways.
Recent Reading
AI for Care: in the UK AI is being deployed at scale to drive better social care, from fall tracking, to pain sensing. (The Economist)
China dominates research: In news that may not surprise anybody, according to Nature magazine, Chinese institutions lead research in a majority of categories. (Nature)
Possibility Chains: From my colleague Frank Diana, the use of possibility chains to track and deal with disruption. (frankdiana.net)
Apple’s AI Critique: Apple recently published a paper titled ‘The Illusion of Thinking’ where they argued that the AI models in play currently weren’t as clever as is being touted. There’s a kernel of truth, but this detailed piece will explain why it’s not quite on the money. (Medium)
Silicon Army: The US army has formed a new innovation core called Detachment 201 populated by AI experts from Palantir, OpenAI, and other tech giants, with a view to driving technology innovation into military operations. (WSJ)
Thanks for reading. And see you soon.
Ved