The Next Big Thing After The Internet and Artificial Intelligence

Can we envision a future beyond AI and the internet? I even had the temerity to give a talk at Google Brain research in nearby Mountain View — an hour from my house — on the topic “Today we Google, Tomorrow we …”. I asked the audience to fill in the dots. No surprise, most folks said “Google”. That’s the wrong answer, IMHO. There will come a time when we view web search as a quaint artefact, just like young folks roll their eyes when I said my first exposure to computing was using punched cards to program a computer in FORTRAN.

So, what could make web search obsolete? What lies beyond AI? Let’s first understand the fundamental limitations of both AI and web search engines.

I can think of a million questions I can ask a three-year-old that would simply befuddle any AI system or a search engine. Here are a few: 

1. if people had wings, would a wall separating one country from another (e.g., Mexico from the US) be useful?

2. if we had eyes behind our heads, like flies, would cars need driving mirrors?

3. If Californians decided to build their houses out of ice, like Eskimos, would that work?

Note all these questions are “what if?” imagination questions. Any three years old could breeze through dozens of questions. No search engine, heck no AI system of any stripe or colour, would answer most of them correctly. Fundamentally search engines are dumb. They do a massive lookup of an indexed database.

AI systems and search engines can generally answer “what is?” questions that statistically summarize the state of the world today (e.g., how long would it take me to drive from my house to the main Google campus in Mountain View?).

They have almost no capability at imagination or answering “what if?” questions or common sense reasoning. I’ve been studying AI for 40 years. The problem of common sense reasoning has been discussed since The 1970s, and we are no nearer a solution today than we were 50 years ago.

So, where does this leave us? Some folks believe that building a deep learning network with ten quadrillion weights and 500,000 layers will solve both the AI problem and the common-sense problem. This is an empirical question, obviously. I wish them all the luck in the world. I don’t share their optimism. But who knows, perhaps this could work. Meanwhile, let’s ponder a fundamentally new approach.

A startup in the UK called Improbable has, I think, the beginnings of the right answer. In short, they want to build “The Matrix”.

Just imagine for a moment if we could build even a tiny part of the Matrix, perhaps for the San Francisco Bay Area. Any question we want to answer, even what-if questions, we could pose it to our Matrix. WIRED magazine had a nice write up on Improbable that’s worth reading. It tells you succinctly the limitations of data science and machine learning, why solely focusing on historical data cannot by itself solve either the common sense or the imagination problem.

If we're living in a simulation, this UK startup probably built it

Could this even work in principle? What’s the business plan for Improbable? They started out as a gaming company, releasing a core software package called Spatial OS that lets designers simulate massively complex interacting systems with millions of components. If you’re going to simulate the Bay Area, you’ll need to figure out how to model the interaction of hundreds of thousands of cars, people, cats and so on. Google is not oblivious to the possibilities emerging here. In fact, they are teaming up with Improbable to simulate the world using AR (augmented reality).

Google teams up with Improbable to recreate the real world in VR

Let’s play along with this scenario for a moment. So, say we have a fancy Oculus like headset we can wear and presto, you’re transported to The Matrix, or some part of it — say you’re simply exhausted from staying at home in the COVID pandemic and want to visit your favourite travel destination, perhaps the Galapagos island.

No sweat, The Matrix “takes you there”. You can see how this possibility will make our current text-based web search feel like a primitive anachronism, like punched cards for programming computers. The following conversation with the founder of Improbable, Herman Narula, is worth reading. He’s commented that today AI gets all the publicity, but tomorrow building The Matrix may supplant this into an even larger goal.

How Improbable’s SpatialOS Will Use Google’s Cloud To Democratize MMO Development But how will this solve the AI problem? How do common sense reasoning and the rest emerge from The Matrix. To get a hint of the answer, I want to turn to discuss the ideas of Rodney Brooks, a pioneering roboticist who used to teach at MIT. Brooks popularized a view of AI that’s been forgotten and deserves to be reintroduced to modern younger AI enthusiasts. Even earlier than Brooks, Herbert Simon, one of the founders of AI and a Nobel laureate in economics, commented in his pioneering book Sciences of the Artificial that an ant wandering on a beach looks incredibly complicated in its behaviour, but much of that complexity is due to the environment, and the ant is actually following a very simple finite state machine program. Brooks implanted this vision of Simon in his amazing robots, one of which was called “Herbert” in honour of Simon. Herbert has a unique brain that had no central representation. Instead, it had hundreds of simple behaviours that synergistically produced what looked like goal-oriented behaviour. It had no maps, yet could wander around the MIT AI collecting soda cans.

In a nutshell, Brooks argued in his famous Computers and Thought award-winning lecture called “Intelligence without Representation” that the secret to why animals are so smart is not because they do complex math in their heads, but that they rely on the world as its own representation. A simple way to think about this is to consider shopping for groceries. You can plan what to get and make a list. Or wander through the aisles and pick up items as you see them, which triggers your brain (right, I need to stock up on milk!).

So, The Matrix is an ultra high definition simulation of the world that makes a different approach to AI possible, where intelligence emerges from the synergistic interaction of simple entities embedded in complex environments. We ask questions of The Matrix by simply simulating “what if” questions, like building an igloo in Mountain View and watching what happens in the heart of California’s summer.

In this view, we think of intelligence, not as an ability innate to a creature, but as a composite of the interaction of the creature with its environment. Perhaps this may lead to new insights, along the lines of what Simon won his economics Nobel prize for, showing that businesses don’t follow the theoretical maxims of expected utility theory, but rather use simple heuristics that produce satisficing solutions. This view rings so true when we see society’s response to the COVID crisis. Do the past few months increase your belief that humans are rational beings who take actions that maximize their expected utility? Herb would not have been surprised at all by our response to this pandemic.

Sigmund Freud, the great Viennese psychoanalyst, one not particularly known for modesty, said that mankind had made three great scientific insights in the past 500 years. He listed these as follows.

1. The earth is not the centre of the universe, but just an infinitesimally small speck in a boring corner of one of hundreds of billions of galaxies. 

2. Man is not unique but descended from apes and other animals.

3. Man is not rational, but driven by instincts and desires, just like any other animal.

What if the Internet Not Existed? Click Here to know

Freud, of course, claimed credit for the last insight. Freud’s point, and Simon and Brooks’ reasoning all point in the same direction. The goal of AI should not produce some mythical rational superhuman reasoner, but to understand that like other animals, our “intelligence” owes a lot to the complexity of the world we live in, and we are more driven by our drives, our fears and our imagination than much of modern statistical data science would have you believe. The Matrix may lead us to a more achievable technology in the long run, however Improbable (no pun intended!) it may seem in the short run. Stay safe and healthy! See you in The Matrix!