🔮 Sunday commentary: The next ten years
I spoke at a Google event this week celebrating ten years of the UK’s TechNation startup programme. The past decade has seen a fascinating deepening of the UK’s tech scene.
In 2012, the whole of the tech scene was valued at $53bn. A year later it weighed in at a smidgen under a trillion dollars, representing a compounded annual growth of in excess of 32%.
Examining this ten-year retrospective with Marta Krupinska got me thinking about what has been happening in the core exponential technology domains, and what might happen next.
A flywheel of falling prices (courtesy of learning) and growing demand (helped by falling prices) has transformed the availability of these technologies in our economy.
In 2012, a desktop computer packing an AMD Radeon 7970 system could deliver a gigaflop/s of performance for roughly 89c. By 2022, the latest Xbox Series X did that same number crunch for under 4c.
A decade ago the state-of-the art chip had about 5 billion transistors. By 2022, the Cerebras WSE2 packed 2.5 trillion.
The ultramodern deep neural network of 2012, AlexNet, had 65m parameters. Alibaba’s M5, unveiled in late 2021, has more than a trillion. AlexNet could classify images, M5 was a multimodal network.
In 2012, under 1 of 20 cars sold in Norway was an electric vehicle. In March 2022, 92% of all cars sold in Norway were battery electric vehicles or plug-in hybrids. (The vast majority the former).
The cost to write a DNA base pair for 90 cents in 2012. By this year it had fallen to a tenth of that.
And sequencing the human genome had fallen from nearly $7,000 to under $600.
Looking across this small dataset, we can see improvements in cost of 10- to 20- fold; in engineering complexity of 16,000-fold (looking at the most complex AI models). This has amazing consequences which we’ll all have noticed today (some trivial, some less so):
We can unlock our phones by looking at them,
Much better web search,
Affordable electric vehicles,
Faster Covid vaccines through the use of synthetic models of Sars-Cov-2.
So if that was the last ten years, what might the next ten hold?
Let’s consider three sets of questions.
