The Sentient Machine offers a lively overview of current innovations in artificial intelligence, but disappoints when he actually gets to the topic of sentience. He explains two different types of artificial intelligence, Artificial Narrow Intelligence (ANI) and Artificial General Intelligence (AGI or machine sentience). ANI devices can learn on their own but do not have sentience. We will have achieved AGI when a device has it. ANI devices are all around us, from self-driving cars, to the Internet of Things, to “Mind Hacking” such as in social media advertising. The main part of this book and the part addressed by most reviewers addresses only ANI.
AGI “explodes into existence” when an “I” emerges with a sense of self and an ability to set its own, grander goals. He tells us that “[e]ven leaders in the artificial intelligence community have a hard time conceiving what this Big Bang might look like and what it could bring.” This, I think, needs more attention. My smart phone is not a bit closer to sentience than the pocket calculator I had in high school fifty years ago, regardless of how many apps I download. Self is not achieved by increments; it’s a different sort of thing, a thing that even neuroscientists have difficulty defining. The Sentient Machine does not explore this problem.
Hussain romps through the current state of non-sentient ANI. While he presents some frightening applications, such as AI weapons and mind hacking, he is pervasively optimistic about their promise and our ability to handle them. And with AI weapons systems, for example, hey, others are going to develop them, so we must as well. He may be right, but if we’re going to have weapons that act independently, shouldn’t there be a public conversation about this?
In Part Three, Hussain turns from the ANI we see around us to a future with AGI. The fears about AGI are first, that it might make us obsolete, and second that it might kill us.
ANI is already making us obsolete. 38% of US jobs as of 2017 will be automated by 2030. Against the prospect of massive future unemployment, he suggests that we separate our identity from our livelihood, since the need (opportunity?) to work for a living will disappear. If we are not what we do, who are we? He sets out in two pages (!) to explain the meaning of life. It must, he asserts, lie in that which is uniquely human. We find values like love, charity and loyalty in other species. “Are even these attributes worth preserving at all costs? I’m not sure.” He concludes (with little attention to competing theories) that “the essence of our fundamental purpose is to gain knowledge.” AGI, in his view, can free us from the need to earn a living to focus on gaining knowledge. This is a cheery conclusion from someone in the information business.
In this view, knowledge is an inherent good, rather than being inherently neutral, either good or evil depending on its application. His focus on what is uniquely human sets him on the gnostic path of salvation through knowledge. However, what if our most precious trait is widely shared among living things and has its origins deep in the shadows of our evolutionary past, such as our ability to care for one another? If our starting point is cultivating the inherent dignity and worth of each individual, won’t our exploration lead us to a very different goal?
Additionally, he acknowledges that currently half the world’s wealth is in the hands of eight individuals. He imagines that AGI might create a world where the masses will live in dignified leisure in the pursuit of knowledge. However, with technological advances, we already have the ability to assure that no child in the world goes to bed hungry. We could do this, but we’d rather put our money toward an iPhone X or an OLED TV.
Currently, an increasing share of the value of the world’s goods and services comes from the input of machines rather than people. So the fruit of the world’s production goes to the owners of the machines, rather than the workers. Hussain tells us that the eight richest individuals have as much wealth as the poorest half of the world. Four of the top eight owe their fortunes to information and communication technology. When the singularity comes, these guys, or people like them, will own it. Technology is driving further concentration of wealth. Won’t AGI accelerate this? Shouldn’t we have a public conversation about this? Hussain replaces this public conversation with unalloyed optimism and fairy dust.
A distinguishing characteristic of AGI, Hussain tells us, will be its ability to set its own goals. AI differs from human intelligence in that it explores all possibilities, whereas the mind prunes options for efficiency, and considers only those options. This difference allowed AlphaGo to beat the world Go master. AGI will be a self-teaching machine that sets its own goals and thinks in a way that is beyond human thought, and therefore, he says, cannot be judged by human standards. He gives an unfortunate analogy for this. When a new alpha male takes over a pride of lions, it kills all the cubs. But we as humans can’t judge lions by human standards. Similarly, he says, AGI will be a higher order of being, and not subject to human judgment. (What if it kills all the cubs?)
Hussain is enraptured with the possibilities of creating minds (machines) unencumbered by the restrictions of human biology, human history and common sense, discovering things that we could never conceive of. Since it thinks differently, faster, with more information, how will we possibly evaluate its findings? If it has the ability to act, how can we question its actions? Shouldn’t we at least talk about this?
Hussain is unperturbed by this prospect, since he sees the universe and biological systems as computers, since they can be explained by computational processes. For him, handing the world over to an AGI machine is little more than a system upgrade. But the fact that you can make an analogy between humans and computers does not mean that humans are computers. The explanation is not the thing. A recipe is not a cake. An algorithm may be beautiful, but it does not have inherent dignity and worth.
One might hope that a book called The Sentient Machine would offer a measured consideration of such questions. Instead, cheerleading Hussain ends with a virtual confession of faith. “We must plant the seeds of artificial intelligence as our ultimate creation and give it the ability and agency to become what it will eventually become. . . . Let there be AI.”
Open Pandora’s box. What’s the worst that can happen?
- Hardcover: 214 pages
- Publisher: Scribner (21 November 2017)
- Language: English
- ISBN-10: 1501144677
- ISBN-13: 978-1501144677
- Product Dimensions: 15.2 x 2 x 22.9 cm
- Boxed-product Weight: 386 g
- Average Customer Review: Be the first to review this item
- Amazon Bestsellers Rank: 494,546 in Books (See Top 100 in Books)