AI. Separating Fact from (Science) Fiction.
Published B&T – April 2017
This genesis of this article was a recent report “Be warned. Robots are coming for your job (Called Albert)”
I do not profess to be an IT expert – until recently I thought a gigabyte was a mark left by a male prostitute. My background is pure and statistical maths and we did not use computers. We worked out the formulas and then the coders “organised” for the computer to do the calculations.
Therefore, I have called upon the help of a close friend and colleague, Professor Stephen Holden, to jointly produce the following:
Stanley Kubrick’s 1968 classic “2001. A Space Odyssey” introduced the masses to AI, with HAL the computer, who takes on a mind of its own.
Science fiction has always had a huge following (Dr Who is still running 54 years after it first launched) and so much of what was once science fiction is now science fact. This has led to the general consensus that “nothing is impossible” and computers can solve complex problems.
This is patently false. AI is a fashionable term for powerful computers, but they cannot solve problems – they can calculate literally at the speed of light and hence can be programmed to use pre-determined formulas, to perform complex calculations that would take a person with a calculator years to do. (Plus, that don’t make the simple errors people are prone to do).
The Institute of Advance Studies at Princeton is where the brightest come from all around the world to work at the frontier of physics and mathematics. (Einstein came here when he settled in America). There is not a computer in sight (well, there is, but only to test the complex equations with real numbers). They work with blackboards and chalk/pen and paper. They do the thinking; the computers do the labour.
Genuine AI is the point at which computers overtake the human brain – referred to as Singularity. They take on a personality of their own and do not require “programming”. (HAL goes from a movie concept to real life and the impact on civilisation is profound and possibly even fatal).
We are still a long way from singularity and there is a credible school of thought that says because of the difference in the way computers and the brain operates, it will never happen.
(The following is courtesy of another friend, Professor Kevin Ho-Shon. A very rare person who is medical specialist and has a PhD in computer science and mathematics – he understands how both the brain and computers work).
There is a basic, but fundamental, difference in how a computer and a brain work. Computers are deterministic and linear. They are faster and smaller, but essentially unchanged since Alan Turing invented the famous Turing Machine. (A significant factor in the Allies winning World War2, when his machine cracked the code used by the Germans.)
Electronic circuits used to build computers are logical gates and the output of a logical gate is deterministic. All programmes written are dependent upon this fact.
Neurons do not behave like circuits. Given a certain stimulus they can produce a different output. (Unlike a computer which produces the same result each time) - referred to as a Stochastic Process (series of random events) whilst a computer is deterministic. This is where the brain has it over computers in imagination and creativity. Machine learning is a powerful tool, but it is a linear process and lacks the ability to shift sideways and create.
So, what has this to do with advertising, marketing or even life in general:
Do you want a machine to fly a plane? Manage finding a partner for you (one you are stuck with), etc? All of these can be programmed, but most of us would be reluctant to hand over to a computer to that level of decision making. It may be because it cannot be reduced to a deterministic equation, or that it is more complex, more nuanced than the program can capture -e.g, the film “Sully” highlights the capacity of a human problem solver and the limitations of programs/AI/models. The pilots could land the plane back at La Guardia airport if and only if the human decision to return to base was already made. The on- board computer could not decide what to do, only to execute the decision made.
A common business failing which most MBA programs are at pains to eliminate is the notion that there is some fairly simple solution that can be determined with complete confidence as the solution. When asked a business question, the answer is always "It depends". Anyone that declares emphatically that there is only one right solution is full of shit. Can our program capture all the dependencies, all the contingencies? If the answer is yes, then programmatic solutions are possible - and we're ALL out of a job. For now, at least, the answer is a pretty clear no, and we still have jobs.
Marketing has not been noticeably impressive in their development of and implementation of even the simplest "algorithms". Massive companies implement CRM programs to try and emulate the corner-shop which knew your name and your preferences. And of course, the massive CRM implementations miss their mark horribly. Right now, online advertising is notorious for offering products/services that you have already bought - if it's so damn smart, how come it is offering me a trip to Italy when I already bought one online two days ago? (And then there is the little problem of programmatic buying placing ads in pornographic and terrorist/extremist websites.)
A neat story that Stephen’s father likes to tell illustrates the problem of AI. The first completely automated plane flight is about to take off. Everyone is on board and belted in. The emergency drill has been done, the plane is pushing back, and a Stephen Hawking computer voice comes on the PA announcing "This is your computer captain speaking, welcome to the maiden flight of a completely automated plane. Please sit back, relax and be assured that nothing can go wrong... can go wrong... can go wrong... can go wrong..."
The moral of this story? If you hand over all of your creative work to Albert, HAL or any other machine, you really will be f….ked!
