How Smart Machines Think by Sean Gerrish. Sean is a Senior Engineering Manager at Google leading machine learning and data science teams. He holds a PhD in machine learning from Princeton.
This book is providing readers with a wonderful overview to advances in artificial intelligence, and specifically how machine leaning is now the most popular subset of AI.
How Smart Machines Think is addressing three key areas that reveal the leaps in advancements of machine learning development: The DAPRA Grand Challenge, the Netflix recommendation engine, and Neural Networks.
Each section is well written, providing above all, deep insights tied to objectives driving new business models.
While Sean is certainly providing a solid grounding in algorithms and their methodologies, I was certainly surprised at the depth of autonomous vehicles, recommendation engines, and game-playing. The larger lessons from his book include amazing progress in neural networks.
Machine Learning for autonomous vehicles
Clearly Sean understands the full picture of how this emerging technology began. The 2004 initial contest found team only to achieve a small distance, perhaps less than twenty five percent of the course before their AI systems failed.
To be fair the route was kept secret from each team. Only hours before the race was the actual route revealed. However, the following year Sebastian Thrun leveraged machine learning for Stanford’s car. This permitted for the first time autonomous vehicles to avoid obstacles in the road.
Machine Learning for Netflix recommendation engine
Another topic is the 2007 Netflix recommendation engine challenge. This section focuses on predictive modeling. Moreover, the amazing lesson is Netflix was not in the streaming business at the time of this contest. The contest generated 20,000 team submissions. The stories of those teams, and their fun team names makes this a particularly enjoyable read. Sean also shows how a recipe for Holiday Baked Alaska fits into the picture. At first glance, many would not see the correlation.
Machine Learning for Neural Networks
Accordingly, the third section addresses Neural Networks. Chapters 8-16 reveal the history of computer games like Atari to the Google acquisition of DeepMind in 2014. In fact, the lesson begins with Atari games and how an algorithm defines a time-adjusted reward. There is a small bit of a dive into biology and neurons.
This serves as another base foundation for neural networks. There are similarities to stories and lessons from Artificial Intelligence: A Guide for Thinking Humans by Melanie Mitchell and Grokking Artificial Intelligence Algorithms: Understand and apply the core algorithms of deep learning and artificial intelligence by Rishal Hurbans. This fits perfectly into my understanding of neural networks from my previous reads. Sean closes the book in chapter 17 projecting the next 50 years.
In conclusion, the role of machine learning is certainly transforming the plant in ways most do not understand yet. This book reveals a deep understanding of where we can take this subset of Artificial Intelligence.