AI is often peddled as an 'existential threat' to humankind. However, such a conception often overlooks what 'Intelligence' actually is and the cognitive advantages humans have over software.

The ominous and omnipresent title ‘Artificial Intelligence’ (A.I) has come to denote something that, combined with popular filmic representations of ‘Black Mirror’ like scenarios or the existentially tragic films ‘Her’ and ‘Ex Machina’, inflates the promise of, and misconstrues the history of A.I, its current ability and future progress. Reviewing the nature and scope of what the term ‘intelligence’ truly means sheds light on how complex the issue of artificial intelligence really is and thus, for now at least, why ‘Terminator’ plots of the hegemonic ruling of robots or the devaluing of the arts, is a lot further away than one might think.

There is scarce discussion in A.I about the fundamental nature of what intelligence is: what it is trying to achieve and how we perceive intelligence in other people or in other things. There are of course innumerable tests available for measuring intelligence, and yet no one is quite certain of what intelligence is let alone what the available tests are exactly measuring. The point of reference at which we use the word intelligence varies from objects to processes, from ideas to concepts, and in doing so we ascribe very different senses of the term ‘intelligence’ which seems intelligible in everyday life, but much more difficult to define in itself. It seems that intelligence comprises of many abilities rather than one all encompassing ability, which, trivial as it seems, becomes increasingly complex when trying to compare and define analytical intelligence with creative intelligence or practical intelligence as it's difficult to measure all aspects at once and rank one above the other.

For instance we can distinguish capacities from dispositions. Capacities like mental speed, mental energy, or retrieval accuracy, cannot be increased by self-instruction. Dispositions can be controlled by instruction, for example, our disposition to sort through large quantities of text for a solution to a problem. Capacities are empirically general, i.e. they can be easily measured, whereas dispositions are only measurable based on how those dispositions are taught. Many dispositional components contribute to rational thinking, but there is no one right way to think rationally. Thinking rationally may occur in a variety of ways, dependent on the subject’s ability and awareness of their situation. This makes it near impossible to measure intelligence with a test. Dispositional components have relative value across subjects and people such that what works for someone may not work for someone else.

What if intelligence was something akin to a set of various skills and abilities that make people successful at achieving their rationally chosen goals whatever those goals might be, and whatever environment they are in? I think this does justice to the term intelligence because it brings us away from our blinding fascination with automation and more towards what is characteristically human and special. It is our ‘general intelligence’ that separates us from machines. The fact that artificial machines seek to perfect small aspects of human intelligence, like the calculator, or smart phones to remember things, or financial algorithms to predict market opportunities faster, only show A.I to successfully operate in micro worlds. Unlike A.I, human intelligence is diffusely spread across these cognitive micro worlds, out into the social and physical environment. In order for A.I to warrant the headlines the media so excellently use to instill fear in their readers, A.I might be better off using the slogan ‘Weak A.I’ as distinguished by Searle from ‘Strong A.I’. Distinguishing between the two contextualizes the current ability of A.I whilst maintaining the status of ‘Artificial General Intelligence’ for which enables the innovations of human creativity and the depths of social understanding it takes to survive in and appreciate the world.

A.I systems are dependent on the installation of information in order to have intelligence. The more information installed, the more intelligent the system. Years ago, intelligence tests consisted of simple memory and numerical tests, which as we know now are completely trivial for A.I machines. Recent algorithms like ‘Deep Blue’, the first chess-playing computer to defeat reigning world champion and grandmaster Garry Kasparov in 1997, or Google’s DeepMind algorithm ‘AlphaGo’, the first ‘Go’ playing machine to defeat 9-dan professional Lee Sedol on a full sized 19x19 board in 2016, have succeeded in two of the most mentally challenging games around. It is hard to say what machines will be capable of in the future, but it seems right to suggest that our concept of intelligence, will evolve as the abilities of A.I. technology advances.

Today the aspect of intelligence by which we should measure an A.I is its ability to successfully cope in any environment. To do this, any A.I design must cover the individual, environmental and social constituents of intelligence all together, but for now at least, let it automate, let it continue to push us and our own boundaries and deepen our understanding to what we think constitutes as human intelligence. As was mentioned in a previous article by Penn Frank, let it ‘maximize potentiality’. It has created an opportunity for all to extend their capabilities, just like the calculator did, and just like the industrial revolution did. Let it automate, leaving human creativity at the helm.