Is artificial intelligence the way forward

March 18, 2017

In December 2015 I wrote about the technical predictions for 2016 and Artificial Intelligence, or AI, was on that list. In the last three months it now seems that every other tech story is about AI.  The tech industries are all racing to make things smarter and smarter; but does that mean better?

What is intelligence?

Before understanding artificial intelligence I think it’s a good idea to understand human intelligence.  The problem is that everyone has a different interpretation of intelligence. It was only in the 20th century that theories around intelligence began to emerge from the philosophical ideals that preceded. The IQ (Intelligence Quotient) test is probably the most recognised method people understand to measure intelligence. The IQ test was developed by psychologist Alfred Binet who was commissioned by the French Government to identify children that needed educational assistant at school.  Binet, together with his colleague Theodore Simon, began developing questions that focused on topics not explicitly taught in schools such as attention, memory and problem-solving skills. The early tests revealed that some younger children were able to answer more advanced questions than older children so Binet developed the concept of mental age. This is a measure of intelligence based on the average abilities of children in a given age group. Ironically, Binet identified the limitations of the test as intelligence is far too broad to quantify with a single number.

Whilst the average IQ has improved with each generation, it is still a limited measurement because overall intelligence is built from many types of intelligence. For example, deductive reasoning, spatial intelligence, long-term and short-term memory, emotional intelligence – the list goes on. Some say that IQ tests only measure linguistic and logical-mathematics but obviously we have to also consider “ability” and “aptitude” as part of intelligence.

When it comes to computer intelligence we have to recognise that many things we perceive to be intelligence have only been designed to perform finite tasks. A calculator is smarter than you at adding up.  Google maps is better than you at finding the best possible route. Search engines demonstrate excellent long term memory but are they really intelligent.

We already have many specific forms of artificial intelligence. Airplanes have had autopilots for years and quite often the pilot only files the plane during take-off and landing. Whilst planes are capable for complete autonomous flight, some planes won’t allow autopilot to be engaged below 1000 feet.  Artificial intelligence is also used by last companies such as Netflix and Amazon that predict what you want to watch or buy.

Teaching vs Learning

However, computers don’t think like us. Most of the tasks undertaken derive from a known set of parameters or instructions. For instance, the autopilot only knows how to fly a plane by a given set of tasks with in confines of externals influences. It knows how to fly on a straight course to a specific destination by adapting the heading. Should the air pressure change or the wind blow the airplane off course the autopilot knows what adjustments to make to maintain the altitude and heading. This may seem intelligent to us but we have to understand that the flight computer is not distracted by external emotions or mundane thoughts like “did I leave the oven on?”

Likewise, self-driving cars are able to make very quick decisions on the likely outcome of a situation free from any emotional distractions. Regardless of how impressed we are, all these system remain conscious free.

True AI is where things become a bit muddy.  In order to develop true artificial intelligence, the system has to take into account more factors than are presented at face value.  The reality is that current systems are more artificial than intelligent.

There’s a growing fear that any system that exhibits real intelligence would create a cascade known as the “intelligence explosion”.  This is the understanding that such systems would determine that they have surpassed human intelligence and the best way to improve would be to design and build a better machine creating a “Terminator” style future.  In 2014, Stephen Hawking said “The development of full artificial intelligence could spell the end of the human race.”

Working with AI

Rather than fear AI, the human race would find huge improvements from working with AI.  As mentioned above, a computer is already far superior to any human at adding up a large list of numbers and this is a strength we need to exploit.

Facebook already using such interactions to detect and prevent suicides. The AI scans posts and comments and compares them to others that have warranted intervention. Google is already helping doctors with medical diagnoses as it’s able to trawl through millions of patient records in a fraction of the time a doctor could. The strength in AI is harnessing this power with the doctor and AI working together.

Put simply think of a car.  It has the power for 250 horses not one. AI is really masses of human intelligence at your fingertips.