#1. Are you really talking about AI? (part 1)

AI hype distorts the way in which ‘AI’ is attributed. A lot of tech developers like to claim that their products are ‘AI’ when no AI technology is actually being used. Conversely, a lot of non-experts understandably presume that most digital technologies are ‘AI’ nowadays. It is really important to be specific about what AI we are talking about, and exactly how this AI is being used.

#2. Are you really talking about AI? (part 2)

Many important talking points that the societal implementation of AI throws up are not really about the technology at all. A systematically-racist police force using AI in a systematically-racist way really needs a full-scale conversation about systematically-racist policing … not quibbling over the accuracy of algorithms or the benefits of algorithmic transparency

#3. There is a lot more to AI than GenAI

A lot of long-term AI developers and computer scientists are now increasingly wary of talking about ‘AI’ – something they see as little more than a marketing term. People working in AI development tend to be very specific about which form of AI they are referring to. There are many very different forms of ‘AI’:  LLMs, natural language processing, machine learning, deep learning, computer vision, robotics, predictive analytics, neural networks, expert systems and so on. ChatGPT is just the shitty tip of a very shitty AI iceberg.

#4.  AI has a 70-year history … and a 70-year history of critique

This is not new technology. The field of artificial intelligence stretches back to the 1956 Dartmouth workshop (or earlier). Similarly, this is not new technology critique. Many of the critical points now being made about AI were first raised decades ago by the likes of Joseph Weizenbaum, Herb Dreyfrus, Steve Woolgar and others. AI criticism in the 2020s does not need to simply repeat these points – instead it needs to be building upon them. Above all, AI criticism in the mid-2020s needs to be led by voices and viewpoints that were not part of this first wave of AI criticism – arguments about Indigenous AI, Black AI, Feminist AI, and so on.

#5. We need to talk about ‘actually existing’ AI

Well before the current GenAI bubble, Divya Siddarth and colleagues were stressing the need to stick to talking about ‘actually existing’ AI – i.e. the actual computational, material and meta-physical limits of what this technology is capable of doing. There is nothing to be gained from loose talk of ‘super-intelligence’ or speculations around ‘AGI’ and ‘the singularity’.

#6. Your own experiences of using AI do not give you any particular expertise to talk about how other people experience and/or should be using AI

AI power-users and those who feel that they are doing cool things with AI often feel aggrieved when criticisms are levelled at AI. Just because AI ‘works for you’ does not mean that it works for other people in the same way (or that other people could experience the same things as you by simply using the tech ‘properly’).

#7. One does not have to have ‘built AI’ in order to have opinions about the societal application of AI

A regular bad faith response to any criticisms of how AI is being rolled out into society is that ‘You’ve never built any AI’. In fact, those involved in the design and development of AI often have very little experience of how ‘lay-people’ encounter AI in their everyday lives.

#8. If we’re going to be critical then we need to be constructively critical

There is growing public, political and professional awareness of the big-ticket problems around AI – algorithmic discrimination, deskilling, environmental harms of AI and so on. Most people are now well-aware of these problems … yet things carry on regardless. We are fast reaching a point where we don’t really need more critique of AI per se. What we do need are critical conversations that mobilise people to ask: what can be done instead? 

#9. Other AI is possible!

In fact, the most important thing for AI critics to now do is lean hard into the ‘AI can be otherwise’ agenda … if we don’t like the ‘AI that we have’ then what is the ‘AI that we want’?