“No other living thing in the history of all life on Earth can bake cookies.”
Though this seems an odd expression of intelligence, Darren McKee’s definition of humanity, as provided at a Carleton University talk, is eccentric. This viewpoint shapes the reader’s perspective throughout his book, Uncontrollable: The Threat of Artificial Superintelligence and the Race to Save the World.
Leveraging his background in artificial intelligence strategic foresight, McKee, fatefully, highlights how artificial superintelligence (ASI) plausibly threatens the future.
McKee’s study looks specifically at ASI, which is the idea of an AI that is so effective, it “will likely be our final invention.” Whether he means that AI will be the greatest human invention, or, contrarily, if it will be our end, remains up for interpretation. But, he does warn us to be aware of these advances.
Laid out in three cohesive parts, McKee begins by defining intelligence. Using the baking cookies analogy, McKee draws a clear distinction between human intelligence and AI.
McKee continues by exploring a human’s unique capacity to imagine. Unlike other life forms, humans have an ability to conjure something that seems simple, but requires significant imaginative power, like baking cookies.
The book is a quick read, even for those unfamiliar with AI, thanks to McKee’s smooth and comprehensible writing. It was interesting and intelligently written while demonstrating an assuring command of the information.
McKee includes graphs and technical updates on AI and software, like analyzing AI’s progression through the years in comparison to human abilities (test scores, ability to decipher handwriting). He also includes precautions from big tech companies like Microsoft and Apple, as well as silly pictures created by AI when it misunderstands a task. The comic relief is needed to remind us that there is no fully capable ASI — yet.
The content is clearly separated to learn as you go. There are many terms and concepts to be explained, but they aren’t oversimplified to redundancy. McKee manages to keep explanations professional while maintaining the readers’ attention.
However, ideas are unnecessarily repetitive at times, considering McKee includes “key messages” at the end of each chapter.
The developments in AI laid out in this book are genuinely terrifying. What will happen if we don’t take precautions? Why have we been so lax on AI safety protocols? Why do we perceive AI differently than any other piece of technology? Do we truly believe we are incapable of creating something so extreme that it’s uncontrollable?
McKee also analyzes the structure of neural networks and how this inspired the development of “artificial neural networks” that copy how human intelligence works. Simply put, these networks are based on our own nervous system and how the body transmits information between body parts. AI uses artificial networks to mimic the way the brain learns, which is how AI develops so quickly. Then, the artificial network imitates the body.
If this seems confusing, do not fear. Uncontrollable thoroughly explains scientific and mathematical concepts relative to both human and AI while providing relatable examples, like comparing the significance of AI to the development of electricity or oil.
McKee outlines several issues that come with continued exploration in AI and urges the public to question whether its certain social benefits are worth it. He argues that humanity is severely “underestimating the harms” that come “as AI becomes more capable.”
Uncontrollable also includes large-scale issues associated with the advancement of AI, like what to do when AI is out of control with its own agenda. Putting all I, Robot jokes aside, McKee was clear that AI is developing at an alarming rate and may be uncontrollable if we push the development further than we know what the outcome will be..
In the chapter titled, “Artificial Superintelligence Is A Risk To Humanity,” McKee states that there is a “10 percent chance… [of] human extinction” caused by advanced AI should we ignore the “risk assessments” corroborated by AI experts.
Yet, there is a huge push to develop AI further and further because of the benefits to society. Is a 10-per-cent chance of elimination worth those benefits?
“We stand on the cusp of incredible advances in AI that carry great potential benefits for humanity, but capabilities come with dangerous risks,” he states.
However, it isn’t all bleak. McKee suggests there are some safe possibilities for the future and ways for humans to get involved. Whether that be by educating yourself and others, donating to organizations for AI safety who encourage safe development and use of AI (like the Center for AI Safety) or petitioning federal institutions for AI safety protocols, it may be up to you to determine our outcome.
Uncontrollable serves its purpose as a call to action against unsafe AI use. McKee actively alerts readers to the potential risks of unchecked AI development, suggesting a collapse to an ASI is not so unbelievable. McKee leaves us to ask ourselves, how high does the chance of extinction need to be before we start asking questions?
Featured image via Igor Omilaev/Unsplash.
