I’d first come across Andrew McAfee a few years back with his illuminating articles on the digital economy so when I saw that he was scheduled to speak at an NYU evening event, I had no hesitation in confirming by RSVP, even if it was on a Thursday night after a long week at work.
His Q&A commenced with the statement “We continue to underestimate the speed of advances in technology.” The world’s best chess player, Gark Kasparov, was beaten in 1997 by Deep Blue. But while the rules for good moves in chess are relatively limited, an ancient game like Go is far more sophisticated. There are far too many possibilities for each move so the brute force of computing power doesn’t work. Software programmers need to be more strategic and sophisticated. The problem arises that Go masters cannot easily explain why they make their moves – they don’t know what they know they know – they can’t explain their more intuitive and experienced responses.
And yet, a program called AlphaGo has just beaten the world Go champion. In 2015, the consensus amongst experts was that this would not occur until at least 2027. How did it happen? Through an extended pattern matching exercise based on machine learning. Programmers told AlphaGo to figure out how it works then test these theories by playing against yourself through adversarial play. Later versions of AlphaGo have taken a fundamentally different approach without the pattern matching component, just the rules. What previously took 21 days now only takes 24 hours for AlphaGo to reach super-human status.
What is important here is that AI’s advances can push human knowledge forward. We are now learning from the way that AlphaGo has developed new approaches to Go.
Andrew then went on to talk about disruption in the economy. When a profound set of new technologies come along, the currently powerful companies tend to be disrupted and do not remain on top after the technologies become established. It’s a maxim that the leadership of old companies tend to underestimate the impact of new technologies. It also rings true for countries as new technologies change the relative influence of power throughout the world between countries.
As technology races ahead, it leaves some people behind. The large stable middle class was created in the USA but is now under threat from new technology that tends to create lower middle class jobs.
In response, McAfee stated that he does not believe that government should restrict market dominance. In the past, people have said that IBM is too powerful, or Microsoft, or AOL, or Netscape, etc. Or Nokia and RIM for mobiles. The pattern tends to be that dominance is established but disruption happens that gives rise to another generation of dominant companies. A caveat though: in general, large concentrations of power require vigilance.
There will always be a role for humans. Is a robot going to manage a kids soccer team? That’s impossible for technology to do; it needs people to do that. We will still need leaders, managers, mentors and coaches when people are involved – we still need these jobs. Technology increases the pace of change and change is uncomfortable. We need leaders to mediate between the fast pace of technology and our desire to not change.
In response to questions from the audience, Andrew said that he believes that the notion of a Singularity regarding creating consciousness in technology is unrealistic and stupid. He believes that people read too much science fiction. There is no pathway to get there because the current programs are doing math through machine learning. We don’t understand how brain works. We are not creating artificial brains but AI. It’s like worrying about overpopulation on Mars. It’s so far out there in time it’s not a concern.
When the crazy wins the argument (like we have now with climate change in the US, GMO in the EU, etc) then we all lose. It’s the same with restricting IT and AI. If you want innovation to happen then you don’t want government to be involved.
Machine learning incredibly good at pattern matching. It’s great at crunching data, especially good for health and health outcomes like genetics. Let Machine Learning do what it does best; recognise the connections to accelerate our science.
He took a particularly American view of Universal Basic Income stating that it was not a good idea. For the US it would take 75% of total government revenue. The greater issue is when work is taken away from communities, then you get more boredom and vice – witness the opiod epidemic in stressed areas in the US at the moment. It would be better to try and restore the sense of purpose and dignity and community through work in the coming new machine age. Better policies would be to expand earned income tax credits.
He loves the idea of Blockchain. It’s immutable – the notion of a ledger. His view of Bitcoin is less positive and considers the current price to be based on speculation.
Finally, he would like to see more innovation and entrepreneurship with more breakthrough technologies. He would like to see more Silicon Valleys but to do that you need these four criteria (Great research universities, Tolerance for risk, Rule of law so that you know contracts will be honoured, and a cultural tolerance for failure, weirdos and outsiders). Once established, they are hard to dislodge, although the current US Government is doing its best! It’s also something for the UAE and Dubai to consider in establishing a similar regional innovation powerhouse.