In his book, The Atomic Human: Understanding Yourself in the Ai era, Neil D Lawrence asks what artificial intelligence as a human means for human identity. He doesn't believe we should try to compete with computers. Computers already believe they can communicate faster than they can already be 30 million times faster. He also sees enough scope for technology to provide better services to people. For example, the AI Lab in Uganda says that by working with the United Nations to provide a better alert system, it will help reduce the impact of disasters such as flooding and hunger.
Lawrence, a deep professor of machine learning at Cambridge University and a senior AI fellow at the Alan Turing Institute, dismisses some of the dystopian visions that the technology can take us with. He questioned, for example, the possibility of artificial general information, that is, the idea that machines can acquire intellectual tasks that humans can do.
Here he sits with David Phelan, the technical correspondent for Monocle, to investigate the path ahead.
First of all, what should I call you?
My official title is DeepMind Professor of Machine Learning at Cambridge University, but I don't work for DeepMind, a Google-owned AI company. It donated to the chair and put a certain amount of money the university would pay for the position. That's much less money than you think – scholars are pretty cheap. I have a lot of job titles, so people usually call me Neil.
Should we fear AI? In your book, I write that there is a slight chance that a machine will be clever enough to work beyond our control.
As long as it can happen, it's already happening. A simple view of AI technology leads to serious misconceptions. It also spreads vivid, science fiction-like stories for very human reasons. But, for example, in banking, machines that are beyond our control are interesting because we are talking about a world where there was a human system that interacted, edited and controlled with data. Think of a system such as double entry book keeping. This system allows you to train someone relatively quickly to understand what you are doing and bring the entire human judgment to the task. What we see in the digital world is losing people's power, whether it's banking, law or accounting. When the context is well understood, the human ability to judge influences individual decisions is sort of excluded.
What are the advantages of AI? How can it help us?
I use the book's concept of “artificial general vehicle” to demonstrate the absurdity of the idea of ”artificial general information.” Whether you want to go to the store at the end of the road or go to New York, is there a vehicle that's perfect for all situations? No, there are no such vehicles. The exact same applies to intelligence. Artificial general information is an absurd and deeply misleading concept about how business should go.
But in 10 or 20 years, we can imagine that everyone in the business can interact with a computer and pilot the best software engineers today as fine as possible. It's a big change in power structure: what to do from being limited to a few hyperscalars [companies with huge data centres with enormous computing resources] You can build large software systems on something that is distributed across your organization. It is deeply transformative.
Are there any risks related to AI?
The current structure of software, hardware and operation is now clearly outdated. Information infrastructure is changing so deeply that things are not being done correctly. The problem is that none of us knows what the right way is.
Today, there's an entire business literature on what you should do, but the only thing we really know is that the person pronounces what we should do is likely wrong. Well, one of them may be correct, but which one? What you're looking at is a period of uncertainty when business leaders face the difficult task of making decisions about the technology they are always familiar with. As a result, their business judgment tends to be overturned. But the truth is, there is something in between.
If I tell you to pay attention to something, it may focus on your thoughts, but you will not scan the horizon. We want business leaders to be more focused on customer needs, as well as being a horizon scan. To leverage developments such as AI, we need to restore confidence in business leaders and trust in the fact that basic intuition about businesses and customers is still retained.
However, because information infrastructure has remained almost constant over the years, most companies divide what they do for efficiency into different parts. Unfortunately, it tends to disconnect the business from its customer base and undermine agility. In short, although opportunities are enormous, the challenges we tend to face are reexamining the structure and culture and how it serves the needs of our customers.
American scientist Roy Amara once said people tend to overestimate the impact of new technologies in the short term, but underestimate the impact in the long term. Does that apply to AI?
Bill Gates said something similar. In the short term, it happens slower than you would expect, but in the long term it happens much faster. That's another challenge. The type of time frame to make a decision is really between these two. My advice to business leaders is to improve your communications machine. This allows you to manipulate things that will become complex and evolving landscapes. We can talk about making efficiency, but people know about it. It's about how to deploy and integrate AI into existing infrastructure.
What about sectors like healthcare and education? What opportunities can AI offer?
People wanted to see the benefits it could bring over a decade, but we virtually offered nothing because there is a complete separation between macroeconomic interventions and microeconomic needs. There is a disconnect between the company and its customers, and more seriously, between the government and the citizens. As long as there's something dystopian about AI, that's the root of the problem.
Companies like Amazon have the concept that they have to dive deeper when anecdotes and data suggest contradictory things. I saw people rely too much on data and cite Jeff Bezos as saying that it was usually the right anecdote.
The digital systems we use today give people the impression that things are fine from a data perspective. It's a problem and reflects the strange desire to centrally control everything. I know this won't work. This is what has had a negative impact on education and local governments. People in the coal bath are undermined by their ability to provide, as their tools are separated from what ordinary teachers can do their job. The nurse, like my wife, may spend 30% of her time entering data.
AI offers a great opportunity to make it right. People can sit with nurses, look at work today and give them a set of tools that can be designed very quickly to support them with that data entry. We need a world where the nurse can build that type of service on its own. And it is a fundamental change in our information infrastructure and will take time. So, if deployed in a different way – in less centralized, but smaller pockets – AI can have a big impact.
Where do things need to change?
The ability to use all the innovations in the software is currently protected by several digital providers. They are gaining power, but the rest of us grow and don't make any profit in terms of productivity. This has already impacted health, education and social care. We see what big hyperscalers choose to unfold, and they get stuck in that loop. We need innovation by people who understand roles, jobs, and how to make things better. AI is an extraordinary route to that.
Finally, there is the issue of etiquette. Should we say, thank Alexa?
There are different views on that. My answer is because it's about my own dignity. Voice services such as Alexa provide a human interface and play based on a sense of interaction with others. That means it's deeply locked to my dignity and I don't want to mean lightly myself. What I see on social media over and over is that it reduces our social interactions. Now I don't know if it's right or wrong, but that's what I'm thinking. I choose to say “please” – at least when I remember.
This article was originally published in the Opportunity Edition Newspaper 2025, created in collaboration with UBS for the Asian Investment Conference in Hong Kong..
