Ever-more powerful tech is delivering unprecedented changes to us all, and artificial intelligence is only accelerating that. It’s going to be exciting but also uncomfortable, warns Azeem Azhar – who says we have to rewrite the rules on what’s coming down the line.
We’ve heard a lot about artificial intelligence bringing a fundamental change, whether it is in healthcare, education or a whole range of other areas. There’s a rapid change occurring across multiple dimensions – and Ai is the glue for it.
It is a change that strains institutions. It is not merely that industries are going to live and die and new ones are going to be created, but fundamentally the institutions that we live with are going to need to change.
We are going to need to rewrite the rules. We don’t know how long it’s going to take and we don’t know how painful it’s going to be, and we don’t know what the outcome is.
If you are in a western developed country, it’s been a pretty good run for the past couple of hundred years. We’ve transmogrified our artisanal rule of existence into something that is more sophisticated, healthier, with longer lifespans.
But the question we need to ask is, do we really believe that the ideas that have driven this were the absolute peak of thought on how we should organise our societies?
We know that everything seems to be changing. Business strategies are changing, welfare systems are changing, education systems are changing. The rules of war are changing. How we account for value is changing. Even our way of treating intellectual property is changing.
So how have we got here?
One of the things that has driven value in the technology industry has been this relationship with Moore’s Law, which has said that every couple of years the amount of computational power doubles for the same amount of money that we spend.
That has been phenomenal. It’s what’s driven Intel, Microsoft, Google and Apple to the scale that they are today. Although Moore’s Law is starting to peter out, the demands from Ai have redoubled efforts for the industry to increase the availability of computation. It is getting cheaper and cheaper.
The thing that drove value creation for the past 50 years in the microcomputer industry is continuing in the Ai industry.
But it’s not just within IT. There is a coming abundance of genomic information. In 2001 we first sequenced a human genome and it cost depending on who you ask $100m or $4bn. We did it once and it wasn’t very useful.
If the cost of sequencing a human genome had improved as well as a Moore’s Law silicon chip it would currently cost about $10m to sequence a human genome. In fact, it costs about $1000 today. So we have radically improved our ability to sequence the human genome, which fundamentally changes the way we think about health and wellness but also insurance.
That has created a tsunami of data that needs to be managed.
You can go into other areas. Look at an area as disparate as batteries. We need lithium-ion batteries because they are fundamental to the carbon transformation that we need in our economy. The price of lithium-ion batteries has been declining 15-20% every year for the past six or seven years, and it’s actually accelerating.
So we’re getting much better at doing this, and there are other technologies that are coming behind.
The question is, what happens when these continuous exponential compoundings take place? Well we know historically with the single instance in the computer industry. Over 30 years the fastest computer in the world, in 1985 was the Cray-2. It cost $35m and it about 2bn floating-point operations per second. Thirty years later the Apple Watch was basically twice as fast for 1/100,000th of the price.
That’s because of the input factors having their price performance change because of the Moore’s Law relationship. The past 30 years has seen inputs accelerate at a lower rate than we will see in the next 30. And we’ve got early evidence for this.
So the demand from Ai is driving the computer industry to come out with better and better chips. It’s no longer that Moore’s Law 50% improvement every year. Just in the last five years there has been a 300,000-fold increase in computational demand for the most advanced Ai algorithms. And that is a market pull that drives innovation further down in the chip industry.
So what’s different this time compared to the time that kicked off in 1971 with the Intel 4004 processor?
The first is, we have these independent exponentials, a whole bunch of different core technologies that are not related to each other that are improving at double-digit rates.
The second is that there are these combinatorial effects, because each one can be combined with another and normally glued by Ai to create a new application. The declining price of genomic data can be combined with better processing power to create new applications. Lithium-ion batteries with cars and Ai, to create self-driving cars.
And we’ve created a frictionless global deployment structure. Think about containerisation, supply chains, digital marketing funds, everyone having a super computer that can connect to the internet in the back of their pocket.
So it’s no surprise that in 2018 we sit here where eight of the top 10 companies are technologies. The critical thing here is that all of them have data network effects on top of which sits Ai at their heart.
But there are some downsides to all of this. One is the trend for inequality. California is a great example of what is happening globally, which is massive increase in inequality. It is a tale of two states.
On the one hand it’s cheery-eyed Silicon Valley types enjoying the highest GDP of any state in the US, with 55% of all of the VC dollars. But it’s also the state, as anyone who has been to Berkeley and Oakland will know, with the highest incidence of poverty rate of any state in the US
There are some other fundamental shifts. With the economy, what government ministers talk about is manufucturing industry. But the economy has moved away from factories and stuff you can kick and touch. Back in 1975, 85% of companies’ values were in what’s known as tangible assets like stock and plant and drills. The rest was in intangibles. Today, 80% of the value is in intangibles: it is culture, know-how, software, code and data.
We don’t really know how to account for this stuff and we don’t really know to govern it.
The other thing that has cropped up is the platforms that drive algorithmic control, and there are many of them and they dominate their verticals. Through their algorithmic systems and their hold on data, they manage our preferences, they manage what we can see, they manage what we can buy and they even impact the operations of the markets themselves.
Right now our regulation simply doesn’t know how to deal with it.
So when I look at what has to happen over the next, not five years but 15, 20, 25, years, it’s a wholesale rethink of many of our institutions. And I don’t think we have the answers. I actually think it’s going to be a pretty uncomfortable two decades alongside making lots of great technology advancements.
The one thing that I urge people to do is remain curious and critical and find ways of participating in that discussion.