Thanks to ChatGPT, artificial intelligence (AI) has moved from the domain of technologists into the public consciousness.
While the fear that AI would replace jobs (or worse) has subsided for some, the subsequent over-enthusiasm for AI and the belief it can be used to solve every challenge in business – to fix culture, leadership and more – is potentially more dangerous, says professor Johan Steyn, founder of AIforbusiness.net.
Regulation typically follows innovation, but getting the right balance between the two can be challenging, especially when a disruptive technology is at the centre of an international race for global superiority, he says in an interview with ITWeb TV.
Steyn says it’s difficult for governments across the world, including our own, to move quickly enough, but to do it correctly. The temptation is to override the mechanics of democratic law-making in favour of speed to reflect the urgency of the matter.
“If you’re a dictator, you can probably have an AI Act tomorrow. I would say let's rather move a little slower and make sure we include everyone, because we're such a diverse country, than lean more toward an Orwellian kind of world where we don't care what you think, we're just going to legislate to get it over and done with.”
Although the Department of Communications and Digital Technologies recently released a draft AI strategy document on the subject, it follows in the wake of countries in the ‘Global North’, which are finalising their stance on the subject and producing regulation and legislation.
“The best example for me is the EU AI Act, which is still not enacted into law, although the member states have accepted it. There are a lot of principles in it that I think other regions will probably follow; hopefully, we in South Africa will as well, but we have to flavour it from an African, our own society’s, point of view,” he says.
On SA’s draft national AI plan, Steyn, who is an advisor on the CSIR's National AI Working Group, says it’s good to have a starting point, but recognises the document is vague and leans heavily towards academia.
“My worry is that this initiative could be a great way to produce more papers, speak at more conferences and the like, but if that little boy running barefoot in Soweto, if his life is not changed through the work we're doing in regulation in the next few years, then what are we doing?”
There are numerous reasons and areas where AI regulation is a necessity, from protection of the environment, through to protecting the rights of companies and individuals, and reducing programmed bias.
One aspect of the growth of AI that rarely gets discussed, is that it is more compute-intensive than traditional workloads; the additional power, cooling and water consumption involved will have an increased impact on the environment.
Steyn says: “When OpenAI released ChatGPT, up until then, to train a model, it consumed 1 078KWh of electricity and produced something like 550 tonnes of carbon. That was before they released it. It [the environmental impact] is something that’s not talked about enough. These organisations are aiming for net zero; how they’re going to get it right with where we are now, I don’t know.
“With the AI arms race, it will probably increase what we do to our environment in the next quarter century or so.”
Steyn also highlights some of the disruptions AI is already causing in society, including the recent strikes in Hollywood over the image and voice rights of actors being signed away and the threat AI will be able to create digital replicas of them.
“In our country, we have nothing in the way of regulatory protections on artists and art, as far as it goes with deepfakes.”
The deepfakes issue could also affect legal cases, and the level of trust that can be placed on video or audio evidence in court, as well as the use of the technology in cyber bullying or for ‘revenge porn’.
There are many other potential areas where AI can impact society, he adds. And while there is a need to get regulation in place, enforceability will be a key challenge for South Africa.
“If we look at our country with POPIA [Protection of Personal Information Act] at the moment − great legislation, but the enforceability, the court cases, the costs in calling to task those who have violated it, it takes so long. If we can’t get that right, what are we going to do with AI in our jurisdiction and across the world?”
Share