By Dr Chris Needham-Bennett, Managing Director, and Garth Banks, Consultant, Needhams 1834 Ltd
A coherent argument can be made that we simply do not know the risks presented by artificial intelligence (AI). Additionally, there exists considerable conflation between AI, machine learning (ML) – a branch of AI – deep learning, and robotics etc., but let us stick with the term AI as an umbrella term. Despite the opening cautionary argument, several commentators insist that they do know the risks. Whilst some are well-informed and data driven, several are made by ill-informed self-publicists aiming to scare organisations into possibly nugatory and costly actions. It is worth recalling that when railways were in their infancy, responsible journals and qualified medical doctors supposed that travelling at such high speeds with the rhythmic jarring and noise of going over the points could induce temporary lunacy and aggressive behaviour, let alone the inability to breathe at such speeds (the Northern Line in some respects is indeed little changed).
Despite being a relatively young field, AI-generated developments and associated risks are likely to be accelerated at a rate beyond societal and government responses and regulations. (It is worth noting that the recent EU Artificial Intelligence Safety Bill has just been diluted by US lobbying and the need for EU companies to remain competitive). We are currently familiar with AI’s GPT authorship capabilities, its marketing power behind sending you all those ads when you only accidentally clicked on ‘golf clubs’; the generator of deepfake material, for ‘dis, mal, and misinformation’ purposes; the rigging of elections; designing fake/revenge porn; and the Bond villain developing the ultimate ‘Bio Virus’. One can only speculate as to how else the power of AI could be marshalled antagonistically, for instance in stock market manipulation and control, fake actuarial/insurance/medical data, weapons systems hacking, genomic sequencing corruption, miss-directing banking payments, and design errors in engineering and drug developments.
Whilst all of these ‘Dr Evil’ style options are possible not every organisation is even a target for AI and aside from a ‘blanket attack’ which could affect the whole of society, (in which case mere organisational responses would be of limited use) an organisation would have to be pretty specialist or important in some way for someone to mount an AI attack against it. Currently it seems that the spectre (forgive the pun) of AI is not so much overstated rather it is misidentified.
In the authors’ experience few if any ‘organisations’ are preparing to be AI resilient; this is at least partly understandable. The threat is imprecise and paradoxically remarkably similar to the threat presented by human agents; it is just made more efficient, speedy, and effective by AI. In response to AI developments parts of the population will be neophobic, sometimes with good reason (if not simple self-interest like the Luddites), and this is the likely, obvious risk posed to organisations by those whose jobs can be done relatively easily with AI. This has the capacity to be a serious economic consequence of AI.
Unfortunately, this is only half the story, and it has all been pretty negative as we seldom focus attention on AI’s benign elements. If AI has the power, sheer ability, and compulsion, plus the ability to learn from its errors, it has the capacity to be an agent for change upon society on a scale that is difficult to imagine, let alone comprehend. The domestication of the horse, the industrial revolution, the effect of the Black Death on the demise of feudalism, would appear as minor historical blips compared to the breadth and depth of AI-generated change. As one can never put the genie back in the bottle the real risk of AI is how we adapt to a world in which many work roles, perhaps 50%, are suddenly redundant, authorship and art mimicked and devalued, design, engineering and architecture no longer even studied at universities, and massively complex mathematical problems solved in the blink of an eye.
There is no real prospect of a competition between Humans and AI, but we have a shrewd idea as to who would win. Nevertheless, it is said that ‘The winner gets to write the history books’ and it would be a shame if this was authored by Chat Bots.