Understanding Intelligence: Humans vs Machines - Part 1

Like many who work in IT, I'm often asked questions by non-techies about 'Big Data' and 'Artificial Intelligence' and so on. People are very curious to know what these things are and how they work and if the future will see the world dominated by super-intelligent robots, like the Terminator movies' evil 'Skynet', or if computers will one day really think and maybe even feel, like Star Trek's Lieutenant Commander Data.

To understand Artificial Intelligence, it would help if we understood natural intelligence first. Very often computer scientists jump straight to building robotic processes without even thinking about the natural equivalent. In this article I will discuss aspects of natural intelligence outlined in a book I've been reading recently - 'Thinking, Fast and Slow' by Daniel Kahneman.

The book describes the research that Kahneman, winner of the 2002 Nobel Prize for Economics, and other behavioural scientists conducted over many decades into how people think. The parallels between their research conclusions about how the human brain works and how AI works are fascinating.

Kahneman describes two cognitive systems at work, a dominant always-on rapid-decision making mechanism called 'System 1', and a slower, more rigorous, decision making mechanism called 'System 2'.

System 1

System 1 relies heavily on memory and uses the ingrained results of experience and training that form memories. Its aspects are frequently named as intuition, instinct or unconscious behaviour, such as the ability to sense from a person's voice that they are upset, to recognise images or to instantly recall familiar facts. System 1 is 'always-on', and we usually cannot consciously prevent its actions. If someone says 2+2, we know the answer is 4 and cannot stop ourselves from knowing it. If someone says a more complex sum, then we must stop and work it out.

Evolution has apparently given most animals a very good 'System 1' and it is right most of the time. This makes sense - animals in the wild are rarely faced with very complex problems and are often required to make instant decisions to avoid predators, e.g. to run based on a smell or sudden movement. Intellectually understanding a problem and working out a solution is usually not necessary - your first 'gut' opinion is very often the right one.

Many animals are as good as people at repetitive tasks reliant on memory alone, even fairly complex tasks - as a famous experiment with chimpanzees and number sequences demonstrates.

Malcom Gladwell's best-selling books 'Blink', about the power of instant, intuitive judgements and 'Outliers', that with 10,000 hours of practice at something people will become unconsciously expert are based on the same idea.

The fact that most people are right most of the time, gives rise to phenomena like 'the wisdom of crowds'. However, System 1 is vulnerable to disruptive influences such as bias, framing, availability, over-confidence and other circumstances in which the conscious mind is misdirected to ignore available evidence that might, with better consideration, produce a different and more accurate conclusion.

Sometimes we meet a new situation that we have no memory of and encounter a class of problem that requires deeper thought, or are alerted in some other way to the possibility that System 1 might not produce the right answer. In this situation, the brain fires-up System 2.

System 2

System 2 uses concentration and intelligence to work out solutions from first principles. Intelligence is an obvious component of System 2, dictating people's ability to work out increasingly complex or abstract problems, but the effectiveness of intelligence is dependent on both relevant knowledge and on the ability to concentrate for the required period.

Knowledge is key - if you don't know a method for solving quadratic equations then you are unlikely to invent one on the spot. Equally if you don't have good recall of the relevant details then answering many questions will be little more than guesswork. Knowledge and conscious recall are rarely perfect in humans, but the research seems to suggest that we are over-confident in assuming the knowledge we have is complete and accurate, and this over-confidence is a frequent source of errors in System 2 thinking.

Memory is still important for recording intermediate results and tools can help in this. If you have a pencil and paper to make notes on you will be able to work out large and complex sums problems as reliably (if not as quickly) as any mental maths maestro.

Kahneman also describes experiments showing that concentration limits are finite and differ between individuals. System 2 requires energy - lots of energy. This is why we feel tired after carrying out demanding mental tasks, why the brain uses most of the calories the people consume, and why System 2 is not 'always on'.

When System 2 is engaged and people are concentrating there are distinct measurable differences in physiology - pupil dilation, heart-rate and so on.

As we tire, our concentration wavers and we begin to make mistakes. When concentration is lost, System 2 'crashes', and we revert to System 1, reacting instinctively instead of making considered decisions. Simple tactics like taking breaks and regular glucose intake can improve and prolong the ability to concentrate.

Studies have found that people with a strong ability to concentrate have much more productive and stable lives, and that the relevant traits are observable from early childhood. Apparently, people with weak concentration, regardless of intelligence or knowledge, are prone to instinctive 'System 1' thinking and frequently make decisions for short term gain or pleasure that will result in more negative long-term consequences.

Studies also suggest that the ability to concentrate can be trained and improved. For example, Kahneman cites research which found that the prolonged focus on dynamic and difficult problem solving in many video games is a good way to achieve this. This correlates other studies I have read about which suggest that people who play video games outperform their non-playing peers in subsequent education. Kahneman's book outlines a possible mechanism for this, which is worth remembering the next time you feel like giving out to your kids for spending all day on the games console.

Switching between Systems

As System 1 is normally right, unless there is a mental 'alarm' of some kind, people frequently proceed based on rapidly made assumptions that can be wrong.

Most of the poor decision making that happens in people's lives is when System 1 provides a quick and credible, but ultimately wrong answer.

If System 2 is not engaged and people carry on with an initial false assumption this can have negative consequences, sometimes profound or life-threatening. When neither System 1 nor System 2 can come up with a solution, then what happens? We enter a state of confusion and may well freeze, fluster or panic, depending on the circumstances.

In the next article I will discuss various modes of machine learning and decision making, and the similarities and differences between those and human modes.

Questions on Data Warehousing, Data Integration, Data Quality, Business Intelligence, Data Management or Data Governance? Click Here to begin a conversation.

John Thompson is a Managing Partner with Client Solutions Business Intelligence and Analytics Division. His primary focus for the past 15 years has been the effective design, management and optimal utilisation of large analytic data systems.