About the Book

Scary Smart: The Future of Artificial Intelligence and How You Can Save Our World

Mo Gawdat

Pan Macmillan

326 pages; ₹699

Check out the book on Amazon

About five years ago, we were in awe of Artificial Intelligence. Demos of self-driving cars dazzled us. AI was going to find breakthrough cures for cancer. AI was beating world champions in board games. We anxiously calculated the probability of losing our jobs to robots. We thought AI was going to surpass human performance in many domains. Artificial General Intelligence (AGI) and Singularity were not just science fiction concepts but became news headlines.

But in the next few years, even as we continued to make tremendous progress in AI research and applications, we got wise about AI. We understood that we are not creating intelligence but are building pattern recognizers that feast on ginormous amounts of data. That this AI’s good at connecting the dots using the examples we feed it. We label it “Narrow AI” and know its potential and limitations. Serious AI researchers and experts are under no illusion that narrow AI is leading to or going to lead to AGI.

But Scary Smart makes a leap of faith and is completely in the AGI camp. In fact, Mo Gawdat takes it even further. His AI has superintelligence. It has emotions. It has super consciousness. But what AI does not have is a clear definition in this book. That’s a big problem. Like the elephant in the fable of the blind men, AI means different things to different people. It’s easy to project vague ideas, mix in partial truths and increase the confusion.The book does not provide references or evidence for the bold assertions about AI’s coming omnipotent capabilities.

The book can be summed up thus: Due to deep learning, AI will become supersmart and humanity is about to be outsmarted in all domains and faces dire consequences. But AI is our child and is still learning. If we treat AI nicely on its way up, it may deign to reciprocate and hopefully won’t crush us (once the AI reign starts in 30 years). Children learn from their parents. So, treat each other and the machines with respect and start saying “please” and “thank you” to Siri. Seriously.

I don’t have a problem with this manifesto, but I doubt it will fix our problems with AI. Gawdat’s version of AI feels more like an alien being and not the AI technology that researchers and practitioners are building and perfecting. This divergence made it difficult for me to relate to the book.

Even without AGI, the current trajectory of AI is worrying. Narrow AI has problems galore that need urgent attention and fixing. But many of these problems are caused in the first place, because we put AI on a pedestal and overestimate its capabilities. Alas, this book will serve to perpetuate a mythical version of AI.

The second part of the book is about the steps to fix AI. This is high on ideals, but short on specifics. The proposed solutions don’t seem substantive but feel more like a satsang, with a monk who is selling you his happiness apps. The references and reminders to the author’s happiness books, videos, podcasts, etc. would have been better in the appendix.

Of course, there are flashes of brilliance and insights in the book that make you pause and think. The chapter on digital ethics is an example. I loved the characterization of our current uses of AI as “selling, killing, spying and gambling”, which is spot on. The writing is informal and accessible without any technical jargon.

If you are interested in the current state of AI capabilities, future directions, and implications, you’re better off looking elsewhere. The book will help you shine in salon discussions about AI, if AI is what’s shown in science fiction movies. If ever Gawdat writes a Hollywood script, I’ll get my popcorn and watch that blockbuster!

(Kashyap Kompella is a technology industry analyst focusing on Artificial Intelligence)

comment COMMENT NOW