2024-01-24
The Real Risks of AI – and the Imagined: “Like Changing the Entire Functionality”
AI has emerged as the prominent buzzword in the tech industry. The discussions around what AI is, what it aims to become, how it will assist us, how it might take over the world, and everything in between occur on a near-daily basis. However, what constitutes a genuine risk and what is purely imagined? These are the questions addressed by author, podcaster, and digital ethics expert Per Axbom during his appearance on the Digital Influencer Podcast.
– The best analogy I’ve come across comes from Alan Blackwell, who has just written the book ‘Moral Codes.’ He talks about this notion that an algorithm will suddenly gain consciousness – a subjective emotional life and social abilities. Arguing in that way is akin to saying that if we make airplanes fly faster and faster, they will eventually lay an egg. It’s like you’re changing the entire functionality, as if it were a Darwinian evolution, states tech veteran Per Axbom in Consid’s Digital Influencer Podcast.
According to Per Axbom, the idea that an AI algorithm would suddenly function like a human-like cyborg is both far-fetched and, in his opinion, impossible. However, today’s generative AI can mimic certain human behaviors, posing a risk in itself.
– The Eliza effect is about our strong desire to believe that this is something that truly understands us and genuinely cares about us. We crave for that human connection. This is exploited by the chatbots we are discussing now, says Per Axbom, highlighting another, more programming-based, risk:
– It’s what people are exposed to as these systems are built. That is, training these machines on information that can be quite outdated, biased, and contain thoughts about people that we have long dismissed.
Is legislation keeping pace with rapid developments?
– Is there legislation that is ahead, where someone is gazing into a crystal ball? ‘We believe this will happen next year, so let’s enact laws now.’ It’s a very strange way to view legislation. Also, it’s important to note that the EU AI Act, passed here in December, began its work in April 2021, long before AI was on everyone’s lips. That signals a bit that they are on top of things, I think.
The implementation of AI is another hot topic. The fact that AI can penetrate most industries, all types of businesses, and contribute something is an absolute truth. AI can be a valuable support in education and can certainly be a useful tool. The key is to be aware of the risks, says Per Axbom.
– What I hear is that tools are neutral – they are harmless in themselves – and all tools are comparable. But we know that if a baby bottle contains toxic plastic, maybe I don’t want to feed my child with it. It doesn’t mean the baby bottle is unethical, but there is still a problem there – a dilemma, says Axbom.
He also notes that there are plenty of opportunities with the new technology.
– AI is fantastic when we talk about machine learning and these systems used to find patterns in very large amounts of data. I look at agriculture – being able to predict where to plant what based on weather and how people eat. There are significant gains in large datasets.
New episodes of the Digital Influencer Podcast are released on Thursdays and are available wherever podcasts can be found.