If Machines had an IQ they wouldn’t be in MENSA

Why we don’t need to worry about machines taking over the world … (yet!)

Artificial Intelligence (AI) is finally coming of age; for those of us of a certain generation the actual application of AI today is the science fiction of our youth in front of our eyes. For mainstream media and general understanding it appears in the popular press as though the machines are set to take over; making us redundant from ‘work’.

The crushing defeat of Lee Sedol, World Champion at the board game Go by the Google designed and built AlphaGo artificial intelligence machine has attracted headlines globally. Far more complex than Chess in terms of possible moves AlphaGo won the series 4-1 based on its capability to ‘learn’ through playing itself numerous times; attracting headlines for its ‘creativity’ and ‘intuition’ in the moves it made on its way to victory. But how clever is it?

The Intelligent Quotient (IQ) is arguably the most recognised measure of human intelligence. IQ is a total score derived from a number of standardised tests to measure cognitive ability; the median score being 100 – and c 2/3rd of the general population fall between IQ 85 and IQ 115. Members of MENSA, the not for profit IQ society; need to achieve a supervised IQ score that sits within the 98th percentile or higher; which on one test is IQ 130+. Humankind has on average advanced its average IQ score by 3 points per decade, something known as the Flynn effect; so we are evolving and getting smarter. We are smart and AI only exists because we are; so how would today’s new machines rate if they were measured in IQ? Or do we need a new quotient for machines, a MQ?

In 2015 Stellan Ohlsson at the University of Illnois set out to measure the IQ equivalent of one of the world’s (at the time) most powerful artificial intelligence machines. They used a verbal IQ test designed for children to gauge the intelligence of ConceptNet 4; an AI machine that has been under development at MIT since the 1990s. The test measures children’s performance in 5 categories: vocabulary, information, similarities, comprehension, and word reasoning. The result achieved equated to the average score of a 4 year old child; but below average for a child who was 5. ConceptNet4 has been usurped by v5; and there are no doubt smarter systems such as AlphaGo that could achieve a better result but even with exponential advances in AI are we even close to a machine having a below average IQ of 85?

So we aren’t going to get too concerned about an army of machines with the average IQ of a 4 year old child taking us on in the job market? But I guess that is missing the point about how machines outsmart us? They don’t need to be smarter than us in IQ performance because where they win is their ability to understand at a 4 year old human’s level but then process that understanding at a speed we can’t come close to; and with unrelenting performance 24 hours a day. In essence AlphaGo has the computing power to crunch through endless amounts of permutations on the game allied to a 4 year old’s IQ (?) which makes it smart enough to beat the world champion – but its one specific task it was specifically programmed and taught to be brilliant at.

When their brilliance is in the execution of specific tasks; programmed to learn and understand the ‘data’ presented to them; to execute a filtering / classification process on the data machines outperform humans. It will require further substantial increases in cognitive computing ability to actually get to average let alone MENSA levels of human intelligence. When you take the popular hype out of the ‘artificial intelligence’ bandwagon we should all realize that the machines rely on human input to ‘understand’ context and the range of ‘cognitive’ ability is limited. The clever scientific approaches and methods contained with cognitive computing solutions facilitate an interpretation; the machine simply excels with brute computing strength; executing at scale and when combined with multi-lingual capability an impressive ability to solve real world business problems.

Artificial intelligence isn’t a standalone capability; it is the alignment of human intelligence with machine scale and speed – the key requirement is human input and interpretation.
AI allows for the uncovering of insights in data that would take humans too long to achieve to be useful, allowing humans to make decisions on the insight that delivers a benefit (or indeed allowing another machine to make the decision based on what human intelligence has taught it to do). Embedding the ‘cognitive’ capability of a machine into solving specific business issues provides a general ability to be more efficient – faster, better, cheaper..

At Chatterbox Labs we are working with many of the world’s leading technology and consulting organizations who are harnessing our Cognitive Engine to solve real world business problems. Use cases are emerging daily as our partners apply their intelligence to build out new cognitive products in days to harness our machine learning algorithms, our speed and scale, our multi-lingual capability to bring artificial intelligence to benefit their clients.

In the business world ‘artificial intelligence’ needs to deliver, sustain or create competitive advantage; that is through reducing cost or improving the ability to acquire revenue. Our solutions provide ‘cognitive’ insight to the data to allow for a superior and advantageous form of decision making – taking away mundane or largely manual processes to allow human capital to apply its intelligence and creativity in new ways. In reality ‘artificial intelligence’ is simply an embedded form of our own intelligence that is elevated to new heights of performance; artificial ‘smartness’ perhaps is a better description. With this ‘smartness’ at our disposal perhaps our own IQ Flynn effect will accelerate faster than 3 points a decade?

Chatterbox Cognitive Engine delivers artificial intelligence, machine learning and natural language processing capability in one to provide mid level business analysts with the ability to create new cognitive products in days. Over 6 years of research and development by our PhD educated data scientists is embedded within the process and methods available to our partners creating new cognitive products for their clients.

For more information please visit our web site www.chatterbox.co or contact Andrew Watson, VP Strategic Alliances, Andrew@chatterbox.co