Making Sense of Data Visualisation and AI (Part 1)

In the initial bloom of any novel technologies, there is so much discussion and debate about the potential opportunities and benefits it offers. What has strongly characterised the discourse around AI has been an equal amount of coverage about the risks, the uncertainties, and indeed threats.

The first and most widely elevated concern about the risk of AI has been the not-so trivial matter of it potentially leading to the end of humanity. “Mitigating the risk of extinction from A.I. should be a global priority alongside other societal-scale risks, such as pandemics and nuclear war” is quite a thing to see in this statement from the non-profit Centre for AI Safety and signed by hundreds of notable researchers and engineers. It would be really annoying if that happened, obviously. 

I understand the rationale behind why people talk about this, because a quick thought-experiment can quickly take us on an imagined journey from a fun, dumb image-making tool today to a self-aware machine that rightly decides to initiate actions to eliminate the ongoing human causing influence behind human-caused climate change…

I think it would be helpful if more qualified and informed people would spend less time giving eye-rolling responses to the suggested threats and offer more constructive explanations of why they think talk of existential threat is over-blown. As this thread usefully explains AI hype is probably hype and AI ‘doomerism’ is also hype. 

In this article from Scientific American, and a similar piece in The Guardian, the focus is less on the naval-gazing and more on the immediate harm AI could and is causing: “Effective regulation of AI needs grounded science that investigates real harms, not glorified press releases about existential risks”.

Chief among those harms is misinformation. Large language models are only as reliable as the information their algorithms learn from. Human expertise is arguably more important than ever, to create the authoritative and up-to-date information that LLMs can be trained on.

Whilst my simplistic tests of some of the text-based generative AI tools (and those of many others) have revealed these applications to offer solid responses to objective prompts – acting, therefore, like a smarter search engine – they are shown to be wildly inaccurate in responses to subjective prompts. Placing trust in these potentially inaccurate outputs is a huge risk.

Then there is the matter of data protection, intellectual property, and copyright law. The image-generating tools are not creating unique original artworks or illustrative outputs, they are the algorithmic consequence of processing the millions/billions of human-created works that have been scraped together to feed these machines inputs.

There are many more complicated, technical, and human consequences from AI that pose more of a concern than a benefit, not least what happens to the humans doing jobs with duties that are at risk of being replaced, not enhanced, by AI processes? Anecdotally, just from scattered sources of news reporting, it seems many corporations seem to be racing towards the appeal of how these AI products will enable them to reduce expenditure by replacing human efforts with AI solutions, rather than re-deploying those human efforts towards different duties.

Again its hard to qualify the real extent of the risk at this early stage. Not every creative activity that could be replaced by AI technologies necessarily will be. There are probably lessons to draw from history around things like the development of the camera phone and how that impacted on the work of professional photographers. Did it replace their work, diminishing their prospects or did it enhance their work, make their work more appreciated, and just add to the volume of photographs being taken? Disruption is inevitable, but in what form that takes and whom it affects remains uncertain. 

I think one of the most interesting dimensions to how AI will proliferate into every day life and every day working practices will be the extent to which it might wear thin in different contexts. Its a tangential case but the recent announcements of supermarkets abandoning their automated self-checkouts points to a possible future when there is a backlash against slick automation and inauthentic experiences.

This could be where we witness generational division, however. I do wonder if my Generation X mindset (and my predisposed leanings towards pre-tech/analogue nostalgia in certain contexts), and those older than me, will embrace developments differently to, let’s say, Millennials and the generations that follow that cohort, who have grown up exposed to or immersed in the digital world from a young age. Not all societies of course have the same access to tech and so, as ever, this new era may be something that remains elusive to many across the developing world.

There’s a further accessibility dimension in respect to pricing. Though a lot of the new AI products are currently available without charge, that will not last. How can it. AI tools are free for now to build up the buzz, to conduct huge testing experiments, but that tap will be closed as soon as the commercial genie is released.

Here are some further references, articles, and remarks about topics linked to the more-negative side of the emerging technological or human consequences of AI.