AI visionary and speaker Walter de Brouwer talks to Wavelength in this first of a three-part series
Walter de Brouwer is one of the world’s most qualified experts on Generative AI. He happily identifies as a ‘serial entrepreneur’ having built and sold several AI businesses and was recently awarded “AI Visionary” by Midjourney. While most of us race to catch up with what technology can do right now, Walter is one of the few people on the planet who can see, and shape, what it will be able to do in the future.
In a recent interview Walter said, “I think reality is severely overrated.” We started by asking him what he meant. “The internet was great, but it’s now 30 years old. It was made for text, which is static, and based on economics. Now, the way we have progressed with AI, we want it to be fast and cheap and high quality and dynamic. The internet we have doesn’t fit very well with that anymore.”
It’s tempting to assume that everybody is now familiar with Chat GPT and has a basic working understanding of what LLMs (Large Language Models) are. But Walter cautions against such assumption, “The latest report from executives of corporations says that only 18% have used Chat GPT once and another 20% who have used it but very superficially. A good leader is adaptable – s/he will use it.”
Outside the engineering firms in Silicon Valley, much of the practical application of AI remains a mystery. The balance between the promise of possibility and the reality of what’s available now is tipped heavily in favour of the former. On the maturity of the market for AI products Walter says, “We are at the end of the beginning. But nobody likes the middle part. Airlines can’t sell the middle seats for a reason! But things are going so fast that the middle will be over quickly because the systems will start to build themselves.” Meaning that the products available for companies to invest in, and the ease with which they can build their own, will increase rapidly in the next 24 months. Organisations need to ensure that it’s not just product and technology people who are abreast of what’s coming down the pipeline – these products will affect marketing and sales, HR, finance and CEOs too.
Walter’s joke is that developing generative AI is not rocket science because, “with rocket science, we know exactly how it works. With this, we don’t.” It sounds dystopian, the classic Terminator metaphor, but Walter’s enthusiasm for what he calls “earned dopamine” – that which comes from learning something new as opposed to that which comes from buying something online or watching something on TV – is infectious. “Humans and machines are both in the learning phase together.”
Search results for de Brouwer online throw up a term called “robo-doctor” – another somewhat terrifying Schwarzenegger-esque image – but counterintuitively, Walter’s vision of the future of healthcare is the opposite of de-humanising. He paraphrases Eric Topol’s book Deep Medicine which explains that AI will “give doctors the gift of time.” He recalls the days a couple of generations ago where a doctor would visit his/her patients every few days at home. If doctors can rely on AI to bear the brunt of diagnostics, prescribing, monitoring and so on, the doctor can focus more on the therapeutic aspect of human interaction – the “bedside manner” so to speak. Walter cites the popularity in the States especially of so-called “concierge doctors”; clinicians who are paid a monthly retainer to remain on-call by a fixed number of patient “clients.”
We asked Walter to explain the focus of his project xy.ai, which is experimenting in the field of agents. “It’s still experimental but in one or two years they will be all over the place.” An agent is a piece of software that is built to “think” like a human being would. The business applications are clear, as agents can be trained to execute tasks that would otherwise be done by a team of people. The example given is market research and competitive analysis. In order to cover the ground much faster than a team of human beings can, “the agents replicate themselves – and you can even link them to Eleven Labs [which turns AI generated text into speech] and make them have meetings with each other, you can hear them talking. Just like in real life, sometimes you have to kick their ass.”
If your mind isn’t blown yet, Walter goes on to describe a live project at Stanford, called Simulacra of Human Behaviour, in which 25 software agents “sit in a box and they think it’s real life. Because we have told them that this is their job, this is their partner, this is where they live, this is the post office, this is the park and every day they go to parties, they go to dinner.” The agents “exist” in lines of natural language text, which is how the project team can see whether the agents are “happy” or not. At the end of the day, each agent “reflects” by writing a sort-of diary, after which they can re-write their own code in order to adjust “the rules of the game – it’s a very dynamic world.”
One of Walter’s most spectacular projects was Starlab, in which he brought together a couple of hundred brilliant scientists to work on world-changing, experimental science in seven divisions. “We had protein folding, superluminal travel – all with clients – and even time travel.” Starlab delivered a crucial insight that Walter applies to the way he does business. He had meetings with each experimental team every day and after a couple of years, he started to see analogies across all the apparently disparate divisions. Each individual problem couldn’t be solved on its own, but by joining the dots across the different projects. “A lot of people make a start-up and go to VC and that’s the normal route to market. I set up companies that work inside large corporations, so again I’m in that position where everybody from across the company comes to me and shares intimately their business problems. If you have never seen a client from the inside, how can you design something that works? Before you know it you go bankrupt and then you have to say, well it wasn’t a market fit.”
After Starlab closed down, Walter spent some time working in financial services in Monaco. He is sanguine about the experience. “The easiest job in the world is being a banker. You buy and sell money. It’s also a very human job. After a year I knew everything, and by everything I mean 70%. You don’t need to know more than that, otherwise you become boring.”
For someone who spends so much time with machines, Walter is very attuned to the peculiar joys of being human, and the way storytelling is the key to keeping boring at bay.
“I once met Douglas Adams [author of Hitchhiker’s Guide to the Galaxy], he came to visit me at Starlab. And also Dyson, you know the guy with the vacuum cleaners. I didn’t believe in his idea – big mistake! Anyway, students deserve professors who are great storytellers like Adams. If you’re teaching archaeology, you have to make everybody an explorer. You have to make everyone feel like Indiana Jones – to feel the poetry and the romance.” It’s a crucial and admirable skill, and one that is relevant not just to professors and scientists but for business leaders too.
So what’s next? The AIs that Walter talks to everyday are about two years old. Humanity has a hundred thousand years of experience. He is certain that General AI, which communicates and thinks as well as humans do and possibly better, will come. “When that arrives, we have to have a common language. The children of Generation Beta will be born in 2025. We humans have to build a new language together with AIs so when AGI comes, it understands the fundamental complexity that has developed in human language over 100,000 years. Emotion is not noise, it’s a rich signal. And this new language needs to be spoken and understood naturally, and understood computationally as well.”
Walter foresees a scenario where humans of the future will look at the planet, with its oceans blocked with cables and the sky clogged up with satellites, in the same way that we look at the Great Pyramids today, “Who were these people? How did they do this?” He goes on to say, “We are a reckless and aggressive species that never look back. We always progress.”
Nonetheless, Walter is optimistic about where technology is taking humanity. “We are before a bifurcation in the road where we have to take some decisions. What we learn from history, every time we are at the bifurcation by some coincidence of evolution, we take the right decision.”