(These are excerpts from my book "Intelligence is not Artificial")
Appendix: The Cloud of Invisible Robots
We now thrive in the "data economy". A credit card company is no longer a financial company: it has become a data company (it knows what you buy). Social media are obviously data companies: they know what your interests are, whom you are connected with, and what you like. What do these technologies have in common: Internet of Things, biotech, wearables, social media...? They produce data. Billions of data. According to IDC, 44 zettabytes of data by 2020. Just about every new technology is producing huge amounts of data, and deep learning is one technique to analyze those data. It turns out that "big data" is also what Artificial Intelligence needs to train neural networks. Deep learning needs large datasets; large datasets need deep learning. Traditionally, neural networks were about pattern recognition (such as image or speech recognition). As data become available on such a large scale, a new natural application (the "killer application") becomes increasingly tempting neural: profile recognition. If we put together what you buy, what you do, whom you befriend, etc., we obtain your digital behavior, and that's pretty much your mind (desires, beliefs, needs).
Invisible robots can analyze all the data that we live behind on the Web and learn a lot about us. These invisible robots can deduct intimate facts about your life. Not only does the invisible robot know your digital behavior: it also knows the digital behavior of everybody else. It knows the correlation between circumstances and human behavior; and it can guess which behavior will follow from your circumstances. The next step is to affect your circumstances, to "persuade" you to adopt a certain behavior.
The invisible robot can "persuade" you because it knows your mind. It can customize a message for your mind that will generate the desired behavior. Information can be crafted to "design" your behavior. This is nothing new: marketing specialists have always done this. The difference is that now a) it can be customized for each individual; and b) the "bots" can find out a lot more about you than any marketing study could dream of.
The machine knows more about you than your best friends, and has more data about everybody else than you do; so the machine can use knowledge about humans to understand how humans act, and then use your data to understand how to manipulate you.
Sounds scary? It gets worse: thousands of invisible bots will collaborate on this project.
The bot doesn't understand your conversations and doesn't really "learn" anything; but it can do magic with the digital footprint that you leave behind when you buy something from an online store and when you engage with people on social media.
Invisible robots are the future not only of marketing. A bot trained with a dataset of voter behavior can customize a political candidate's message to maximize the number of people who will vote for that candidate. A bot trained with a dataset of performance appraisals can be more efficient than a human recruiter in finding the best person for a job. A bot, trained with the historical record of repeated offenders, can even advise a judge on the optimal sentence because it can guess the probability that the convicted defendant will commit the same crime again once released from prison.
Bruce Tognazzini, the influential designer of early Apple, written that in our age an invisible computer works for an invisible user. I would rephrase it as: "an infinite crowd of invisible robots work nonstop for an infinite crowd of invisible users".
The engineer of the future is the "behavior designer". The engineer of the future will need to also understand the language of machines, not just the language of humans; and the language of machines is computational mathematics.
You want to be the one who shapes Artificial Intelligence, not the one who is shaped by Artificial Intelligence.
Back to the Table of Contents
Purchase "Intelligence is not Artificial")