It is not just that AI’s are doing complex tasks, it’s that they increasingly doing complex evaluative tasks better than the best humans in the whole world. These AI’s can be written by a few (or just one) persons using off-the-shelf compute. This makes it available to almost anyone – at prices far below the price of an average yearly employee salary.

‘A pilot A.I. developed by a doctoral graduate from the University of Cincinnati has shown that it can not only beat other A.I.s, but also a professional fighter pilot with decades of experience. In a series of flight combat simulations, the A.I. successfully evaded retired U.S. Air Force Colonel Gene “Geno” Lee, and shot him down every time. In a statement, Lee called it “the most aggressive, responsive, dynamic and credible A.I. I’ve seen to date.”‘

There’s a lot going on here. A computer should be able to defeat a human pilot due to its lack of concern about excessive G forces and other ‘wet-ware’/human factors. It can take advantage of the full capabilities of our modern strike craft (such as sustained dozen+ G turns) without an concerns of blacking out or loss of cognitive powers.

But there are other serious considerations.

Ease of Creating Systems better than the Best

We see that a single doctoral graduate at a public university was able to create something from commercially available hardware using well published/studied fuzzy logic algorithms that is able to beat some of the world’s best pilots. Consistently. With a little work, in theory, this could be plugged into a real jet and it could take out a good portion of the US, or any other, Air Force. It’s not science fiction – it’s doable by anyone in the world today with the desire and a jet. That puts this easily in the realm of all the superpowers, and even into the hand of a number of 2nd world dictatorships. The compute power, the algorithms, and everything except for the jet are available to you and I today.

The logical response is to then develop AI’s that can fight the other AI’s. This leads to an arms race where humans are not even a part of outside of the tricky work of training and writing the statistical training for the AI’s. Now we have autonomous weapons of war, armed for combat, fighting each other in which people would no longer be able to compete against. The ramifications of this are somewhat staggering – all the way up to a SkyNet style apocalypse. I would recommend the book “Killing without Heart” by Shane Riza for more on this topic.

Social impact

In just the last few years, we’ve seen the rise of IBM’s Deep Blue ravage Jeopardy and defeat chess grand masters. Google AI defeated the Go world champion numerous times. On the commercial front, the rise of Siri/Google voice and other systems that use AI techniques to recognize speech and context get better and better each day. Self driving cars are already becoming a reality and may replace all taxi systems. The US Postal system already uses machines that operate on 30,000 letters per hour (8 letters/second) and have eliminated rooms full of people. What’s surprising is that a lot of these feats are done with machines that only cost about the salary of a single employee for a year.

The ever increasing pace shows us this is becoming widespread and is rapidly being adopted by wide sections of industry. Cloud companies are already in place selling the compute for pennies (Google I/O just announced it is releasing it’s AI systems for super-low prices of just $10/mo). As it replaces big pieces of our day to day lives, it is going to be a profound impact on our society – much like the industrial revolution. It’s not that we can (or should?) stop this development, but these technologies are going to have an even bigger impact than the internet. With that, there are a lot of things to considered with each new development if we wish to avoid to social disasters that plagued the late 1800’s and early 1900’s. It would really be good for us to start thinking about these issues before they become serious problems.

Machine Morals

To that end, we see people already working on imbuing machine learning with morals. We got our first view of how wrong things can go when Microsoft had to shut down it’s Twitter AI Tay after just 24 hours when it started spouting racism, denied the Holocaust, and worse. We also saw a AI controlled robot mysterious escape from a lab – twice. We cannot ignore that these systems will do unpredictable and unexpected things.

So, in the end, we have massively trans-formative technology entering our society. It’s probably a good idea to realize these things are not ‘if’, but ‘when’ realities that will come in our lifetimes. Maybe we should start talking about how we want to live with them in the new world it will create.