Four reasons why hyping AI is an ethical problem
Bias, discrimination, privacy violations, lack of accountability — AI entails a lot of ethical problems. Hyping AI creates additional ethical challenges on top of the existing ones. Here is how:
1. We do not need AI for everything
The AI hype epitomizes the belief that we need AI for everything, or at least that AI makes always sense. It does not question the very purpose of AI.
Yet, there is a lot of what I call ‘AI for nonsense’. ‘AI for nonsense’ is an antithesis to the ‘AI for Good’ rhetoric. AI for good strives to serve a meaningful purpose and is popular especially in the area of sustainability. ‘AI for nonsense’ is a term that comes to mind whenever I read about some AI ‘innovation’ where any evidence of real progress is missing. Such as when Walmart uses AI to ensure that products are in stock and fresh but where that AI pumps out “1.6 TB of data per second, or the equivalent of three years’ worth of music”. In that case that seems to be just AI for the sake for AI.
Another example for ‘AI for Nonsense’ is an AI application that discovers when someone lights a cigarette at a gas station. In that case, we need to remember the following saying about technological progress: every augmentation is an amputation. In fact, this feels like an amputation of common sense. If we start to forget some basic principles of survival and we delegate our survival to AI, we are doomed.
So, evidently, ‘AI for Nonsense’ distracts resources from real world problems.
2. AI cannot solve our most wicked problems
Not surprisingly, AI hype is also linked to misleading promises. This becomes evident when we hear claims like ‘AI will end poverty’, ‘AI resolves climate change’, etc… Unfortunately, such misleading promises can often be found in debates about ‘AI for Good’. They are often linked to a ‘technological solutionist’ mindset, that is, the belief that even the most wicked problems can be resolved with technology. But when reading such promises, we always need to carefully ask: “What is the problem, exactly?” It’s impossible to solve wicked problems like climate change with technology alone. AI often only solves “proxy problems”, for example, it can identify extreme weather events, which helps mitigate the impacts of climate change — but it cannot resolve climate change. More generally, AI cannot free us from our responsibility to change our lifestyle.
3. Hyping AI clouds its impact
Even though I’m critical of an excessive or exclusive ‘impact orientation’, which in itself is a hype, at least in the realm of sustainable investments, in the case of AI, some matter-of-fact impact orientation would be helpful. ‘AI hype’ is pretty much the opposite of such thinking, particularly in scenarios like those presented by Elon Musk with brain implants that merge human brains with AI. Such scenarios again direct energy and resources (and emissions) at something that is barely tangible, far away in the future and carries a significant number of ethical risks.
If we adapt an impact-oriented perspective by contrast we always need to ask: Who will be affected in what way by currently hyped AI? Who bears the costs? And what are the unintended consequences?
4. AI hype downplays human contribution
AI hype is also part of stories that exaggerate the capabilities of AI in the present when effectively humans are still doing most of the work — we have all heard about the thousands of ghost workers who are manually labeling data to feed algorithms under dire working conditions. So, presenting something as machine intelligence when it’s actually human intelligence, is also dishonest and it deprecates the humans doing the real work.
So, addressing the ‘classical’ ethical problems of AI is still paramount. On top of that, we also need to call out AI hype wherever we see it.
The true value of AI does not lie in outsmarting or re-configurating humankind. The true value of AI lies in increasing our understanding of real-world problems whose complexity overwhelms the human mind.
This is where we should direct our attention and resources, not to unfounded ideas of utopian thinkers or doomsday prophets.
Context: These thoughts are an extended version of an input speech I gave as part of an online panel with the title “AI hype — fake it till you make it”. The panel was part of The Algo 2020 conference and it was hosted by Jack Stilgoe from UCL.