Researchers' Zone:
Chance separates the strongest from the weakest
COMMENT: Studies show that an element of luck and contingency determines whether actions are considered intelligent – both in a computer's algorithms and in executive offices.
Published
Many years ago, as a young computer nerd, I tried to program a simple game of connect four.
At the time, I did not understand how to program the game logic needed for the computer to choose the most optimal moves.
So, when it was the computer's turn to play, its actions often seemed deeply unpredictable and alternated between being extremely forward-thinking and then being stupid and self-destructive.
Later, I was helped to develop the game according to correct mathematical principles, so the computer chose the most optimal action by thinking several moves ahead.
To my great surprise, it now turned out that the computer became easier to beat, as its moves were now fully predictable.
Not only that – the game also became incredibly boring.
It then made me think about how randomness can appear as intelligent behaviour.
I later found out that I was not the only one who had made that observation.
Easy to build a clever robot – hard to understand its behaviour
In 1984, the Italian brain scientist and director of the Max Planck Institute, Valentino Braitenberg, described a series of thought experiments that involved tests with simple robots that could drive around the floor looking for light.
One of the main points in his book ‘Vehicles: Experiments in Synthetic Psychology' was to show that even very simple robots can seem intelligent to observers.
Nevertheless, it is almost impossible to calculate how robots think and function by simply observing and analysing their behaviour.
Braitenberg concluded that it is significantly easier to come up with behaviour that looks intelligent than to observe and analyse what brought about this seemingly intelligent behaviour.
Good to incorporate randomness into a learning machine
That there is some link between intelligence and randomness was also evident to the famous British mathematician, codebreaker and father of computer science, Alan Turing.
As far back as 1948, he wondered if randomness was somehow the prerequisite for intelligence. Among other things, he wrote:
‘It is probably wise to include a random element in a learning machine ... A random element is rather useful when searching for the solution of some problem’.
Turing believed that randomness was a fundamental phenomenon of physics, and he also philosophised over whether computers that contained an element of randomness could be considered machines with a free will, even if he himself was not in favour of that line of thought.
Finding treasure in the desert
Today, randomness is used as an important element in modern artificial intelligence.
Many of the algorithms used, for example, in internet searches and for optimisation and modelling purposes use randomness as a way of finding the optimal solution as quickly as possible.
Artificial intelligence is a broad area covering a variety of technologies and methods.
But if you look at the subset of artificial intelligence called machine learning, it is essential to find the most optimal solution systematically out of millions (or billions) of possibilities.
Artificial intelligence is therefore often an exercise in getting the computer to find the famous needle in the haystack within a reasonable time frame.
Imagine you have to find treasure buried somewhere in a desert, but you do not have a map of where the treasure is buried.
You know there is some treasure somewhere, but not where to dig, how deep to dig down or when to give up and start digging somewhere else.
If you do not have any other information to support you, or someone you can ask, then there is only one option – to roll a dice and rely on chance.
This is a recognised issue in artificial intelligence, and many of the considerations behind the advanced algorithms in the field consist in weighing up how to make use of randomness.
Should you dig the hole deeper in the hope that the treasure is further down, or should you start digging a new hole?
The role of chance in your career
Randomness not only plays a role in artificial intelligence; it has proven to be crucial in how our careers develop.
In a study from the University of Catania in Italy, an attempt was made to understand the role of randomness in our lives. A computer model was built for human talent and how it is applied throughout life.
The model was then used to examine the extent of randomness in the individual’s level of success. The simulations showed an accurate reflection of the wealth distribution we see in the real world.
But to the researchers' surprise, it was not the most talented who became richest but very much the luckiest.
According to the researchers behind the study, in our cultural circle we have an excessive belief that success is due to personal qualities such as talent, intelligence, skill, effort or risk-taking. But we underestimate the importance of randomness.
No correlation between bonuses and performance – but who cares?
The researchers concluded that one should be careful about giving extra resources or powers to successful people as one cannot expect such investment to yield higher returns than by providing resources to people who have performed averagely.
A similar pattern has been observed in financial markets.
In his book Thinking, Fast and Slow, the Israeli-American psychologist and Nobel Laureate in Economics Daniel Kahneman describes how at one point he had the opportunity to analyse the connection between the bonuses of a group of skilled stock market speculators and their performance over time.
To his surprise, it turned out that there was no connection at all. Those who performed best had basically just been lucky.
When he presented these results to management (and, moreover, to the stock speculators themselves), he was met with indifference. Apparently, we seem rather uninterested in knowing that luck has a large impact on how we perform over time.
With a little luck, the mediocre do great
Does all this mean that you can be indifferent about developing your talent and, in principle, just sit back and wait for lady luck to smile on you?
No, of course not.
Stock market speculators would never have ended up in Kahneman's study if they were not all skilled. Talent and skill are prerequisites for success – but so is chance.
According to the study from the University of Catania, the most talented people would not have been able to do very well if they were unlucky, whereas people with an average degree of talent would actually be able to do incredibly well if they were lucky.
Disquieting facts
If we accept the premise that randomness has a major impact on how we do individually, then one can consider the trajectory of life in the same manner as the aforementioned exercise in finding treasure in the desert.
In reality, you do not know where to start, when to change your strategy and whether you have found the best possible solution. And in many cases, the result will be based on a number of coincidences over which you have no influence whatsoever.
I think it is thought-provoking – and actually a little disquieting – that although we know that randomness plays a huge role in whether a politician, chief executive or stock trader obtains good results, we largely choose to ignore it.
We must remember that success is not just about talent
Perhaps this is because we cannot reconcile ourselves with the thought of uncontrollability and thus always seek a rational causal explanation? However, as Braitenberg demonstrated in his robotic experiments, it is hard to rationalise what constitutes intelligent behaviour.
We try to tame the level of randomness when optimising our society and our organisations, but experience from computer modelling and artificial intelligence shows us that randomness can be a useful tool in achieving the best solutions and most significant results.
As a society, we must be mindful of not only investing our resources in the safe and well-tested but also take chances – even though we may waste our stakes.
Success is not just a matter of talent but a combination of talent and randomness.
Translated by Stuart Pethick, e-sp.dk translation services. Read this article in Danish at Videnskab.dk's Forskerzonen.
References
Søren Tranberg Hansen’s profile (AAU)
'Vehicles: Experiments in Synthetic Psychology'. MIT Press (1986).
