Elon Musk predicted AI will do any job that does not require touching something physical by next year. He also said humanoid robots will eventually out produce humanity — and in the same breath told governments to keep regulation out of the way. Joining me to unpack it all is Deborah Webster, Founder of AmaniLabs. Deborah, welcome back to Wall Street to Mena.
Thank you for having me back.
First, your thoughts on the school safety AI app we were just discussing — using cameras to flag and prevent student conflicts before they happen.
It sounds great in principle. The issue is what happens behind the scenes — what they are doing with the data, for instance. One person's freedom is another person's licence. You end up having massive surveillance. The path to hell is paved with good intentions. Part of my work is as a Fellow with ForHumanity, which develops audit criteria for AI systems — autonomous systems and algorithms. We just finished all the criteria for educational technology. What these companies should be doing is proactively putting themselves up for audit so that they catch any potential harms before they unleash this technology. It sets them apart and raises the bar.
A 27-year-old safety researcher who spent years at Anthropic and OpenAI publicly warned that working on superintelligence could end humanity. What do you make of that?
He is not the first person to resign. The fact that a safety researcher left one of the leading AI labs — one that says it is focused on responsible and ethical AI — already speaks volumes. They have already had other exits like that. Dario Amodei a few weeks ago said to his people that Anthropic's culture is probably going to change because they will attract more people who just want to join for the money. This is not a one-off. It is a slippery slope. Looking at the quality and calibre of a company's ability to attract and retain talent is a really interesting signal — for people buying this technology, investing in it, and investing their time and talent with these companies.
Musk predicts mass job displacement and asks for less oversight in the same breath. What does that contradiction actually mean?
The space is full of contradictions. The power and potential of this technology is amazing. The fact that it is being led by a handful of people is already problematic. There is a lot of doom ringing going on, and the more people hear it, the more numb they become to it. Meanwhile, real harms are currently being experienced with this technology right now — surveillance technology being deployed in workplaces is shocking. Whilst people are afraid of losing their jobs, if people understood what companies are deploying on them in their places of work, they would not want to work for those companies. Jobs are more than a paycheque. They are means for people to hone their craft, to add value, to be valuable contributors to this world.
Only 20% of the global working population is positively engaged at work. What does that mean for how we deploy AI?
When we look at where AI is being deployed, we need to make sure it does not disengage the 20% who are positively engaged — and that it raises the circumstances for the 80% to become more engaged. We need to look at the world through a different lens. This winner-take-all track we are on is going to harm everyone. Even if AI was not going to wipe out humanity — the circumstances in which humanity is going to live if we carry on this track are not good. Not only will we see an increased potential invasion of privacy, but also a decrease in human agency. And I do not think that gets enough attention.
What do you mean by human agency?
People's free will to make free choices. People do not realise their decisions have been influenced by algorithms for the last ten to fifteen years. The best analogy is a frog in cold water — if you put a frog in cold water and slowly heat it, it stays in. But if you put it in boiling water, it jumps out. Our tolerance for what is okay and acceptable has massively decreased. And on face apps and social media trends — that is biometric data. If people were to read the privacy policies of the apps they use, they would be horrified. If the average person were to read all privacy policies and terms and conditions, they would need 65 working days to get through them.
If you had to bet — does the industry police itself in time or does it take a real crisis?
Looking at history, the likelihood is that only a crisis will make us course correct. We could get good regulation and people miraculously start behaving — very unlikely, frankly. Or there is a big catastrophe that wakes people up. But what could be potentially worse is death by a thousand cuts — where the harms are large enough that people struggle to handle them, but they become numb. Before you know it, privacy is gone, agency is gone, and a lot of power is in the hands of a very, very few.
Thank you so much for being here with us today.
Thank you. Thank you for having me.
