“beyond the term artificial intelligence there are popular beliefs and fears that have long been conveyed by the film industry…. we must navigate between all of this and not stigmatise technology”
Thierry Breton, EU Commissioner, Internal Markets
In the week that the EU unveiled its plans for AI regulation it was great to join the Profusion Technical Reading Group to discuss the concept of the Technological Singularity – the ambition to achieve (human level) Artificial General Intelligence, and the related prospect of a Superintelligence that far exceeds human capabilities.
Heady stuff for a Spring Tuesday evening, as the news swelled with a remarkable display of Super Stupidity from some of European football’s biggest, and richest, clubs. Yet the two events may not be as far apart as they appear.
The football clubs, like so many AI evangelists, demonstrated a remarkable lack of understanding or appreciation of the context and heritage surrounding their ambitions. In their financially driven and enabled hubris they mirror the excesses of Big Tech, forever justifying their behaviour with reference to historic inevitability, rather than an unprecedented level of corporate greed.
These very same Big Tech companies have hoovered up the best and brightest brains from across the world to support their ambitions and to complement the most powerful computing architectures ever built. In the same way the self-styled Super Clubs hoover up the best talent to train in the best facilities and to play in the best stadiums. So far so capitalist you might say – a football oligopoly to match the Big Tech oligopoly.
But one of the key questions raised in our group was the misallocation of resources to support the vaulting ambition to achieve (human like) Artificial General Intelligence, rather than using those same resources to address urgent real world problems and challenges (not the least of which is sustainability) – as someone pointed out, AI hasn’t helped much with the pandemic, has it?
Similarly, how much good could be done with the billions promised by JP Morgan to the nascent European Super League, to support grassroots, youth and women’s football in local communities across Europe, fighting the scourge of obesity while addressing racism, misogyny and other forms of prejudice?
A reality check
Just as the European Super League plans faced a rapid reality check as soon as they were unveiled so now is the time for the AI industry to dial back its own hyperbole and face reality.
Our reading group proved to be a remarkably sceptical bunch. While there is a profound intellectual and academic interest in understanding the human brain and attempting to recreate its capabilities, this is a world away from the realities of today’s practice of Artificial Intelligence.
Key techniques deployed under the umbrella of AI include Machine Learning, Deep Learning and Reinforcement Learning (not mutually exclusive). All have made great progress in recent years, much of it driven by enhanced computing power (enabling the analysis of vast datasets).
Reinforcement Learning, behind the development of AlphaGo Zero and the understanding of protein folding, is slightly different and potentially more exciting as it does not involve vast training datasets and has the potential to display genuine creativity (e.g. when playing Go).
Yet, this exciting example bears little relation to the commercial realities of AI, which remain profoundly human, and subject to the many failings, foibles, prejudices and biases of human actors.
The hyperbole surrounding AI, part of the wider tech narrative we’ve been imbibing for the last 50 years, has dulled our critical faculties to this essential truth.
Language is of course a key part of the problem, with the use of the term Artificial intelligence directly related to a confrontational view of the relationship between human and machine intelligence. From the very earliest representations of AI in film there has always been a threat scenario, an existential danger, a fight to the death, the man (always a man) versus the machine. But why?
And why can’t it be different? Why must we pursue our God Delusion (nice turn of phrase from our group)? Why are we so determined to disembody intelligence, to strip it from its corporeal and environmental context? Why don’t we step back down to earth and look at how we actually live?
Augmenting our intelligence
Shifting our language and so our cognitive frame from Artificial Intelligence to Augmented Intelligence would be a significant step forward, it would be more honest, more modest and more realistic. It would speak to our historic use of technologies and (other) intelligence (e.g., animals), and it would immediately present a more collaborative approach to that garnered from sci-fi and film representations.
The mundanity of much so-called AI was exemplified in our group when the humble thermostat was posited as one of the simplest technologies that matches a working definition of an AI as “any device that perceives its environment and takes actions that maximise its chances of achieving its goals” (Poole, Mackworth, Goebel, 1998).
Thinking this was may also help us to think differently about the relationship between AI and employment. The western world seems to be obsessed with robots and AI coming to take our jobs. The situation is very different in Japan and South Korea for example. The reality is that AI will automate tasks not jobs and we have always used technology to mitigate the dull, dangerous and dirty work that us humans would prefer not to do.
The question we should be asking is ‘how can we invest in technology to help our people to do their jobs better?’ rather than ‘how do we invest in technology to replace people?’.
Equally we should be celebrating the differences between human and machine intelligence, it is remarkable what machine intelligence can achieve in so many fields. Indeed, we can reasonably argue that the singularity has already been achieved in myriad vertical fields – including image recognition, medical diagnostics, natural language processing, drone flying and even flying helicopters on Mars.
There is so much that we can and are doing with technology and machine intelligence that we should celebrate that we don’t need the distraction of worrying about a Super Intelligence. As (leading AI researcher and tech entrepreneur) Andrew Ng put it; worrying about a Super Intelligence is akin to worrying about overpopulation on Mars!
Let a thousand algorithms bloom
In attempting to wrap up a vast subject (and over ambitious blog post) the key message from our reading group was that we shouldn’t allow ourselves to be distracted by the search for the Holy Grail of Artificial General Intelligence, let alone the prospect of a Superintelligence.
By contrast we should accelerate our commitment to using the latest data science techniques to address real world issues and challenges in academia, business and society. We should champion social responsibility and inclusive purpose.
We should build on the successes we have already achieved in key fields like image recognition and language processing, we should celebrate successes in medical applications, robotics and more.
And we need to embed all of this work much more firmly back into our real world. We need to demystify the discipline and to be humble and honest about its human nature – with all of our strengths but also all of our weaknesses.
We need to be very clear that we are the masters, and that technology is our servant, we need to celebrate the rich diversity of humanity, and the complexity and mystery of the human brain, rather than dumbing ourselves down to meet spurious AI tests.
Whether the fantasy is about a malevolent or a benevolent Super Intelligence, it remains just that, a fantasy, nothing is going to save us from ourselves, we are the masters of our destiny and we must take (back) control of our futures!