The Road to Conscious Machines: The Story of AI

As someone actively trying to navigate the fields of machine learning (ML) and artificial intelligence (AI), I’m always on the lookout for books that aren’t super loaded with equations. Mathematics has an important place in ML, but I do think there is value in looking at it from the other side. Jeremy Howard and Rachel Thomas, world class AI researchers and the creators of the legendary fast.ai, both advocate for a top down learning approach. They believe that learning is best achieved by doing, and in the case of ML/AI, that means writing code (a lot).

Even though I’m fascinated by AI from a technical perspective and enjoy reading about it from that angle, I still felt that I was missing an insight into the origins, historical motivations and the previous efforts that have taken place in the field. This led me to search for a book that would cover this, and I quickly arrived at The Road to Conscious Machines: The Story of AI. Written by Michael Wooldridge, a professor of Computer Science at the University of Oxford with extensive experience as an AI researcher, the book dives straight away into the beginnings of AI and the influential role played by Alan Turing in conceptualising the legendary test and machine that bear his name.

The author takes the reader back to AI’s Golden Age, a period of time in the 1950s and 1960s when the promise of AI gained significant traction amongst the scientific community. Concepts like ML garnered interest from researchers and exciting progress was made but Wooldridge describes in layman’s terms how complexity challenges like combinatorial explosion threatened AI’s evolution, with the Golden Age ending painfully in a series of critical publications from outsiders who described AI as nothing more than alchemy. The author explains that the early AI pioneers may have promised too much, even the name AI itself may have been misleading. 

Wooldridge moves quickly through the AI winter of the 1970s to a new era of AI research where knowledge-based systems became the area of focus. The book discusses expert-based systems, logic-based AI and knowledge graphs with interesting insights into the state-of-the-art developments of the time. The author makes sure not to go too deep into the technical details which makes reading the book an enjoyable experience. In the last few decades, the rise of deep learning has led to significant breakthroughs in AI. Wooldridge uses DeepMind, the Google acquired AI startup most well-known for AlphaGo, as the basis of his assessment of the topic. The author explores the origins of neural networks, connectionism and what it means for these networks to be deep. He also reveals how datasets like ImageNet and the availability of increased computing power has been so instrumental in the success of deep neural networks. 

The author describes how DeepMind was able to combine two neural networks and Monte Carlo tree search to create a system that beat the world’s greatest Go player, Lee Sedol, 4 games to 1. In spite of such tremendous achievements, Wooldridge is keen to articulate the limitations and risks associated with deep learning and how it is nothing close to artificial general intelligence (AGI), something so often sensationalised by the media.

One of the areas in the book that I really liked was Wooldridge’s open and honest assessment of the current state of AI and where current/future expectations may not be in sync with reality. By addressing common pain-points in the AI discussion like healthcare, automation, bias and autonomous vehicles, the author agrees that care must be taken but that we may be over-estimating the capabilities of the technology itself. A particularly interesting point that Wooldridge raises through the infamous trolley problem is that we may be expecting too much of AI in its ability to make life or death decisions, as most if not all of us would struggle at this task.

The Road to Conscious Machines was fascinating, an enjoyable read which provided a really absorbing view of AI, its origins, challenges and achievements. Wooldridge wasn’t afraid to explain what AI is not and where it’s limitations lie, which I think is invaluable for anyone with an interest in the field.

You can find a copy of The Road to Conscious Machines: The Story of AI here.