Number 1: The ‘Bicycle for the Mind’ gets an update

From Bicycle to Beyond: Computing's Evolution

In 1981 Steve Jobs talked about how a computer was like a ‘Bicycle for the Mind’, a tool to amplify our mental capabilities through elegant digital engineering. The computer, like the bicycle, isn't just a passive tool - it's an efficiency multiplier that dramatically extends human potential.

And that was 43 years ago. When the most powerful computer on the planet, the Cray 1 supercomputer, costing the equivalent of $39 million dollars, and the size of a large room, could process 160 million floating-point operations per second. Compared to todays iPhone, which can process 5 trillion floating-point operations per second.

Bicycle no longer feels adequate.

The Exponential Reality

And the speed of change is getting faster. The computational power of an Nvidia GPU, which are the processors powering much of the current AI boom, has increased 1000 times in just the last eight years. As Jenson Huang, their CEO, likes to say, they are currently improving at Moore’s Law .... Squared.

Moore’s Law itself (named after Gordon Moore, one of the co-founders of Intel) has held for nigh on 60 years, with the power of computers roughly doubling every 18 months. Which equates to an 100X increase in power every decade.

But at Nvidia’s current pace we might well be looking at 1000X over a decade.

And there’s more going on ...

Over the last decade we have increased our access to data by 10X, and that is likely to repeat itself over the next decade.

The Scale of Change

And then we have the scale of the models being developed for Artificial Intelligence (AI). This is what Mustafa Suleyman, co-founder of the pre- eminent AI research lab Deepmind, and now head of consumer AI at Microsoft has to say:

‘the scale of these models has grown by an order of magnitude
that is 10X every single year for the last 10 years. And we're on a trajectory over the next five years to increase by 10X every year going forward, and that's very, very predictable and very likely to happen’.

And finally we have Aaron Levie, founder of Box, pointing out that the context window of Large Language models (the technology behind ChatGPT) has increased some 500X in just two years. Which is unparalleled in technology and means that these programs can process ‘in memory’ some 1.75 million words at the same time. This ability to hold and process vast amounts of data in memory sets the stage for AI agents that can tackle increasingly complex tasks, which we discuss below.

The New Operating System

We are living in extraordinary times. Exponential times.

What’s more, our base is now so much higher. Double something even very small and within not that many years you have something very big.

So we are moving from 'Bicycle for the Mind' to an emphatically 'Human Plus Machine’ world. Where AI, in the form of LLMs, is going to become the kernel of our new operating system, as we move into an age of ‘Natural Language Computing’. Instead of programming computers we are going to be conversing with them, through text or speech. We will soon be able to say what we want, and a language model will be able to ‘compute’ what is required to achieve it, and go off and find us the tools we need to ‘get the job done’.

All of this means that computing becomes less about operating a tool (riding a bicycle) and more about having at our side an intelligent collaborator, that amplifies our intelligence, and enhances our cognition and creativity.

Technology is becoming an extension of human capabilities, rather than an external aid, that learns and adapts to individual users, and provides personalised support.

It’s also becoming a great enabler, where we can cognitively offload lesser value mental tasks and free ourselves for higher-level thinking.

In this world, ‘bicycle for the mind’ feels quaint.

The Age of AI Agents - Harnessing Exponential Growth

And next year, and for the next 10 years, we are going to see a huge increase in what is rather clumsily called ‘agentic’ computing. This is where your have small, discrete AI mediated programs (‘agents’) developed for very specific tasks, which can be bundled, unbundled and reassembled - think LEGO - to fulfil specific goals. Each ‘agent’ can complete a task, but collectively a group of ‘agents’ can complete and finalise a wider goal.

Critically this is based around a new methodology in AI, where language models are given the power to reason and think through an answer, step by step, in response to a question. Daniel Kahnemann, the Nobel prize winning behavioural economist, wrote about our brains having System 1 and System 2 thinking. With System 1 being about instinct and immediate response, and System 2 being about logic, reflection, deliberation and rationality. To date language models have been System 1 ‘thinkers’ but the ability for them to acquire System 2 capabilities is advancing, quite rapidly. Which would mean ‘a swarm of agents’ wouldn’t just answer questions—they would engage in reasoning, weighing pros and cons, and making thoughtful, multi-step recommendations.

Looking ahead

The 'bicycle for the mind' captured the essence of personal computing's first revolution - a simple tool that dramatically amplified human capability. But as we enter this new era of cognitive collaboration, with its exponential growth in processing power, data, and AI capabilities, we're not just changing our tools - we're transforming the very nature of human-machine interaction.

This shift from mechanical augmentation to intelligent collaboration isn't just the first great theme of the next decade - it's the foundation upon which many of the others will build. Because when our minds are amplified by artificial intelligence that can reason, learn, and grow alongside us, the possibilities ahead become truly extraordinary.