Bold claims for AI are hard to compute for economists

Bold claims for AI are hard to compute for economists

A new ‘virtual workforce’ is enhancing the productivity of human intelligence

FEBRUARY 20, 2017 by: John Thornhill

Talk to a bunch of economists and they will doubtless tell you that poor productivity
growth is the scourge of our age.

Lounge in the back of a limo with some chief executives, on the other hand, and they will enthuse about how new technologies are transforming corporate productivity.

Track down some experts in artificial intelligence and they may well babble on about standing on the brink of a productivity revolution. If we ever reach the point of technological singularity — when computers outsmart humans — productivity growth will accelerate exponentially.

From that moment, a computer superintelligence will rapidly discover everything left to discover. This Master Algorithm, as the author — a computer science professor at the University of Washington — Pedro Domingos calls it, will be the last invention that man makes. It will be able to derive all knowledge in the world — past, present, and future — from data.

There does appear to be, to put it mildly, something of a “productivity paradox”. Can all three stories be true? Quite possibly, yes.

Hype, of course, is not an alien phenomenon in the tech industry. At present, we are a very, very long way from technological singularity and opinion is divided about whether we will ever reach it. It is worth noting, though, that some (younger) researchers in the field are convinced they will achieve it in their lifetimes.

Yet even the application of narrow, domain-specific AI that exists today is producing startling results as the big tech companies — Google, Microsoft and IBM — pour money into the field. For a glimpse of what is possible, it is worth checking in with BenevolentAI, a London start-up attempting to revolutionise medical research.

Kenneth Mulvany, Benevolent’s founder, argues that drug discovery is in large part an information and data challenge that can be effectively addressed by AI. PubMed, the online medical research site, holds 26m citations and is adding about 1m new publications a year. That is clearly more than any team of researchers could ingest in a lifetime.

Benevolent has built a computer “engine” capable of reading and mapping such data and extracting relevant information, highlighting “conceptual hypotheses” in one field that can be applied to another. “You can look at things on a scale that was unimaginable before,” Mr Mulvany says. “This AI-assessed component can augment human intelligence.”

Benevolent is working with researchers at Sheffield university to investigate new pathways to treat motor neurone disease and amyotrophic lateral sclerosis (ALS). Early results are promising.

Richard Mead, lecturer in neuroscience, says that Benevolent has already validated one pathway for drug discovery and opened up a surprising new one. “What their engine can do is look across vast swaths of information to pick novel ideas to repurpose.”

It can also help personalise solutions for individuals according to their genetic make up. “We are really excited about it. The potential is incredible,” says Laura Ferraiuolo, lecturer in translational neurobiology.

Some economists argue this combination of fast-expanding data sets, machine learning and ever-increasing computing power should be classified as an entirely new factor of production, alongside capital and labour.

AI is creating a new “virtual workforce”, enhancing the productivity of human intelligence and driving new innovation. Moreover, unlike other factors of production, AI does not degrade over time. Rather, it benefits from network and scale effects. Every self-driving car can “learn” from every other such vehicle, for example.

A recent report from Accenture and Frontier Economics made the bold claim that the widespread adoption of AI-enabled technologies could double the economic growth rates of many advanced countries by 2035.

It estimated that AI had the potential to raise the annual growth rate of gross value added (a close approximation of GDP) to 4.6 per cent in the US, 3.9 per cent in the UK and 2.7 per cent in Japan.

Such studies are educated guesswork. Advances in technology are unpredictable. But some AI pioneers are convinced it could “change everything”, from material science to energy. “We are at the dawn of a new age of innovation,” says Mr Mulvany. “We already have human-augmented innovation. We will eventually have machine innovation.”

Even the most gimlet-eyed of economists may soon have to accept that AI is affecting productivity in profound and possibly extraordinary ways.

Copyright The Financial Times Limited 2017. All rights reserved.


Popular posts from this blog

High-speed Hyperloop track ready for first trial run in Nevada

How the Fed went from lender of last resort to destroyer of American wealth

Apple leaps into AI research with improved simulated + unsupervised learning