Artificial intelligence expert and advisor Darren McKee is concerned that humanity is not moving fast enough to protect itself from malicious, rogue or indifferent AI.
Source: Darren McKee
Artificial superintelligence will either enslave us or free us.
Artificial superintelligence will fulfill our cosmic destiny, or it will never happen.
Artificial superintelligence will always be inferior to us or it will be our next god.
Choose what you like. There are an overwhelming number of predictions about the future of artificial intelligence (AI) from trusted experts, great philosophers, science fiction writers, and random people with YouTube channels. It is difficult to think clearly about the emergence of true artificial intelligence. Because it will be historically unique and unprecedented. It may happen this century, it may never happen, it may happen ten minutes from now. Indeed, we are constantly embracing and struggling with new technologies. But the AI ​​is not a modified Acheulean hatchet or light flyer. For better or worse, its arrival will be a pivotal turning point in the human story.
Darren McKee is one of the trusted experts mentioned above. He is an advisor to his AIGS Canada (Artificial Intelligence Governance and Safety Canada) and author of the following books: Out of Control: The Threat of Artificial Intelligence and the Race to Save the World. This book is fascinating, comprehensive, thoroughly based on sound research, and terrifying. McKee advocates for reasonable vigilance and meaningful regulation as we seemingly move forward. important moment.
Guy Harrison: Explain how true artificial superintelligence would be new and different.
Darren McKee: Remember when you were little? Could you imagine how much power you have now? No, even if someone patiently tries to explain it to you, it is impossible for you. The gap between you when you were younger and who you are now may be similar to the gap between you and artificial superintelligence. This dramatic difference in ability comes with risks.
While we tend to think of computer systems as tools, and they are tools in their own right, artificial superintelligence is agentic, acting as if it had autonomy to complete a wide range of tasks. . Autonomous systems are not new to humanity, but we have never developed a system as functional as this one.
GH: The Terminator scenario of armed robots hunting humans may be unrealistic, but what are some credible ways AI could destroy civilization?
DM: It helps to consider specific scenarios, but more importantly, the expected outcomes. If you played against the world's best AI chess-playing system, you would lose. You can't tell in advance exactly how you'll lose, but you'll lose. Similarly, when humanity confronts something that can think faster than we can, with a world-expert level understanding of military strategy, human psychology, economics, cybersecurity, power grids, basic physics, chemistry, biology, etc. Then you'll probably win.
A possible path is to integrate AI systems more and more into our lives and society due to the benefits they bring. As AI systems are able to operate with greater agency, they may be able to deceive and manipulate humans to gain more power and resources. Ultimately, just as we pay little attention to ants when we walk around or clear land for construction, autonomous AI systems with advanced capabilities will It may be that they have no regard for their welfare at all. It is important to understand that malicious intent is not necessary for advanced AI to bring down human civilization. Simple indifference is enough.
Mackie's book calls for urgent and immediate regulation and international cooperation to make us as safe as possible from super-artificial intelligence.
Darren McKee
GH: Your book makes a strong case for doing everything we can now to prevent AI catastrophe, but how do we do that when there is so much mistrust between nations and companies? Can meaningful safeguards be achieved?
DM: With a lot of effort, care, and persistence. Although we have had sensitive and worrying power relationships between nations and corporations for many years, we have still managed to cooperate on many important initiatives. The world came together decades ago in the Montreal Protocol to protect the ozone layer, reduced nuclear weapons by tens of thousands, and humanity continues to strive to detect and deflect asteroids. With advanced AI, everyone is at risk. No nation or company can be safe from an uncontrollable artificial superintelligence.
GH: Can AI-enabled policing catch up with the criminal use of AI?
DM: Harassment and crimes using AI are already a big problem. There are apps specifically designed to alter an uploaded photo of a person wearing normal clothes and use AI to simulate a nude image of that person. These scary apps are already causing damage in schools and around the world.
Educational institutions are scrambling to keep up and need the resources to do so. However, it should be noted that the usual complications associated with increased police powers arise here as well, such as how citizen surveillance can be used in both positive and negative ways. .
It is important to raise awareness, including reminding families that everyone should have a special password to avoid falling victim to audio and video impersonation scams that imitate loved ones and steal money.
Artificial Intelligence Essentials
GH: At least five of my books have been used to train AI systems without permission or compensation. What do you think about this?
DM: Artists are perfectly justified in feeling outraged when their work is used to train AI systems without any consideration or compensation. This kind of plagiarism will not stop on its own, so creatives need to come together to fight such practices. If you don't get involved, others will decide your future.
GH: Would it be better for world security if governments or corporations built artificial superintelligence first?
DM: Given that private companies have a greater responsibility to their shareholders than to the entire population, technology equivalent to nuclear weapons should be subject to democratic control.
GH: When do you think the first artificial superintelligence will exist?
DM: Artificial superintelligence will probably exist within 10 years, but something more universally functional could be created within 2-5 years. As the old saying goes, “It's hard to predict, especially about the future,” and it applies equally to the field of AI. Given the risks inherent in the existence of super-artificial intelligence, it is better to be cautious and prepared than to be caught by surprise.