How will AI change the world?

Summary

Artificial intelligence is predicted to bring significant changes to both individual lives and the world as a whole. However, opinions differ on the specifics of these changes. In a World Economic Forum interview, AI expert Stuart Russell discusses the common misconceptions about AI objectives and future advancements. The conversation also touches on AI's potential impact on employment, drawing parallels to historical perspectives like those of Aristotle and Keynes. Russell stresses the importance of recognizing and setting appropriate objectives for AI and warns against absolute certainty in these systems. The future development and integration of general purpose AI are debated, with predictions ranging from the near future to several centuries ahead.

Highlights

• AI might drastically change life as we know it, but specifics remain debatable 🌍.
• Human tasks involve flexibility; AI lacks this subtle capability as of now 🤯.
• Concerns about AI-induced unemployment are not new, echoing back to Aristotle and Keynes 📜.
• AI advancements could eliminate millions of jobs, raising dependency risks 🔄.
• Speculation abounds regarding when general-purpose AI will revolutionize industries 🛠️.

Key Takeaways

• AI's potential to change lives and the world remains contentious 🤖.
• Objective-setting in AI must account for all contingencies, contrasting human instinct 🤔.
• Historical context, like Aristotle’s foresight, sheds light on AI-related employment fears 📚.
• The gradual improvement in AI capabilities raises the risk of over-reliance 💼.
• Predictions for general purpose AI's arrival vary widely, reflecting its complexity ⏳.

Overview

Artificial intelligence presents a potential turning point as it promises to revolutionize our world. Yet, the exact ways it will do so are a hotbed of debate, with AI experts, including Stuart Russell, shedding light on both promising possibilities and significant pitfalls. Russell highlights the contrast between human flexibility in task execution and the rigid goal-setting currently characterizing AI systems.

The impact of AI on jobs draws parallels to historical predictions, with figures like Aristotle and Keynes providing timeless insights on technological unemployment. Russell uses present-day examples from e-commerce to illustrate how AI might redefine workplaces, raising pertinent questions about our future dependence on technology and the subsequent societal changes.

Predicting the precise date of general-purpose AI’s dawn proves elusive, as experts offer divergent timelines ranging from the near term to far-flung futures. The ongoing advancements in AI foreshadow an era where flexibility, understanding, and adaptation become necessities in overcoming the novel challenges AI introduces.

Chapters

• 00:00 - 00:30: Introduction to AI's Impact The chapter 'Introduction to AI's Impact' discusses the transformative potential of artificial intelligence in the near future. It notes the difficulty in reaching a consensus on AI's effects. The chapter includes insights from a World Economic Forum interview with Stuart Russell, a respected computer science professor and AI expert, who clarifies common misconceptions about AI. He emphasizes the fundamental difference between human task execution and programming objectives for AI systems.
• 00:30 - 01:00: Human vs AI Objectives The chapter discusses the contrasting ways humans and AI perceive and prioritize objectives. Humans understand informal and implicit constraints when executing tasks, such as getting a cup of coffee, where the task shouldn't override ethical considerations or other valued concerns. In contrast, AI systems are typically programmed with fixed objectives lacking this nuanced understanding, leading to potential misinterpretations and undesirable outcomes if real-world complexity and ethical nuances are not factored in. The challenge is ensuring AI systems can integrate broader human values and priorities into their objectives reliably.
• 01:00 - 01:30: Challenges in Specifying AI Objectives The chapter discusses the challenges involved in defining objectives for artificial intelligence systems, particularly when unforeseen side effects can have catastrophic consequences. An example is given of a catalytic reaction that may operate efficiently but consumes a significant portion of atmospheric oxygen, causing a slow and unpleasant demise for living creatures. Even when attempting to remedy one problem by specifying objectives more carefully, other unintended outcomes, such as poisoning marine life, can occur. The chapter emphasizes the complexity and critical importance of considering all potential side effects when setting objectives for AI to prevent disastrous outcomes.
• 01:30 - 02:00: Comparison with Human Reasoning This chapter discusses the concept of human reasoning and its ability to handle, yet not fully know, numerous variables in decision-making. For instance, when it comes to requests such as getting a cup of coffee at the Hotel George Sand in Paris where the coffee is 13 euros, a human would intuitively know to question the cost before making a decision. This inherent understanding is a key difference between human reasoning and hypothetical rigid directives where every variable would need explicit addressing, such as the consideration for not killing sea creatures like fish and seaweed while performing tasks.
• 02:30 - 03:00: Machine Uncertainty and Control The chapter discusses the limitations in current AI systems regarding their inability to understand unstated objectives or request permissions before taking actions. It uses human behavior as a contrast, highlighting how people naturally ask for consent before making potentially disruptive changes. The chapter emphasizes the need to build AI systems that recognize their own lack of complete knowledge about objectives, prompting them to seek clarification or permission, particularly in scenarios that could lead to unintended consequences.
• 03:00 - 03:30: AI in the Real Economy The chapter discusses the concept of control over AI systems, emphasizing the importance of the machine's uncertainty about its true objectives to prevent undesirable behavior. It draws parallels between AI and human behavior, suggesting that certainty in objectives can lead to negative outcomes. The chapter also raises questions about the impact of general-purpose AI on the real economy, prompting considerations on how society can adapt to these changes.
• 03:30 - 04:00: Technological Unemployment The chapter discusses the concept of technological unemployment, a term popularized by Keynes in 1930. It begins by referencing Aristotle's vision of automated machines doing work without human intervention, exemplifying fully automated weaving and music production. This foreshadows current technological trends where machines are increasingly replacing human jobs, as seen in the partial automation of e-commerce warehouses. The underlying concern is the potential rise in unemployment as automation becomes more prevalent.
• 04:00 - 05:00: Automation in Warehouses The chapter discusses the automation process within warehouses. It contrasts old warehouse systems, where humans rummage through piles of items to retrieve products, with modern systems where robots now retrieve entire shelving units containing the needed items. Human involvement is still necessary to pick the objects from the shelves, as creating a robot capable of accurately picking any item from a diverse range is complex and challenging.
• 05:00 - 05:30: Dependency on Machines This chapter discusses the theme of dependency on machines, referencing a story by E.M. Forster about a society entirely reliant on machines. The central message highlights the risk of losing knowledge and understanding of our civilization when its management is handed over to machines. It draws parallels with the movie 'WALL-E', portraying a scenario where humans become weakened and childlike due to machine dependency.
• 05:30 - 06:00: Loss of Human Understanding The chapter "Loss of Human Understanding" delves into the pivotal role of books in preserving and transmitting civilization's knowledge across generations. However, the limitations of books are highlighted as they cannot function independently to sustain civilization; continuous teaching is necessary to impart knowledge to the next generation. If this educational chain, which spans trillions of person-years and countless generations, is disrupted, there could be unforeseen consequences. The chapter also touches on the unpredictable arrival of general-purpose AI, emphasizing the challenge of pinpointing its emergence.
• 06:00 - 06:30: Arrival of General Purpose AI The chapter discusses the arrival and impact of General Purpose AI, emphasizing that its influence will incrementally grow with each advancement in AI technology. It suggests that most experts predict the emergence of general purpose AI by the end of the century, possibly around 2045, although some, like John McAfee, consider this timeline optimistic due to the complexity of the problems involved.
• 06:30 - 07:00: Timeline for AI Developments In the chapter titled "Timeline for AI Developments," a discussion unfolds regarding the uncertain time frame for significant advancements in AI. When questioned, a person referenced in the transcript humorously or ambiguously estimates that achieving certain milestones could take anywhere from five to 500 years. This broad range underscores the unpredictability in predicting AI progress and highlights the need for substantial intellectual contributions, metaphorically noting that several individuals of Einstein's caliber may be necessary to drive such developments.

How will AI change the world? Transcription

• 00:00 - 00:30 In the coming years, artificial intelligence is probably going to change your life, and likely the entire world. But people have a hard time agreeing on exactly how. The following are excerpts from a World Economic Forum interview where renowned computer science professor and AI expert Stuart Russell helps separate the sense from the nonsense. There’s a big difference between asking a human to do something and giving that as the objective to an AI system.
• 00:30 - 01:00 When you ask a human to get you a cup of coffee, you don’t mean this should be their life’s mission, and nothing else in the universe matters. Even if they have to kill everybody else in Starbucks to get you the coffee before it closes— they should do that. No, that’s not what you mean. All the other things that we mutually care about, they should factor into your behavior as well. And the problem with the way we build AI systems now is we give them a fixed objective. The algorithms require us to specify everything in the objective. And if you say, can we fix the acidification of the oceans?
• 01:00 - 01:30 Yeah, you could have a catalytic reaction that does that extremely efficiently, but it consumes a quarter of the oxygen in the atmosphere, which would apparently cause us to die fairly slowly and unpleasantly over the course of several hours. So, how do we avoid this problem? You might say, okay, well, just be more careful about specifying the objective— don’t forget the atmospheric oxygen. And then, of course, some side effect of the reaction in the ocean poisons all the fish.
• 01:30 - 02:00 Okay, well I meant don’t kill the fish either. And then, well, what about the seaweed? Don’t do anything that’s going to cause all the seaweed to die. And on and on and on. And the reason that we don’t have to do that with humans is that humans often know that they don’t know all the things that we care about. If you ask a human to get you a cup of coffee, and you happen to be in the Hotel George Sand in Paris, where the coffee is 13 euros a cup, it’s entirely reasonable to come back and say, well, it’s 13 euros,
• 02:00 - 02:30 are you sure you want it, or I could go next door and get one? And it’s a perfectly normal thing for a person to do. To ask, I’m going to repaint your house— is it okay if I take off the drainpipes and then put them back? We don't think of this as a terribly sophisticated capability, but AI systems don’t have it because the way we build them now, they have to know the full objective. If we build systems that know that they don’t know what the objective is, then they start to exhibit these behaviors, like asking permission before getting rid of all the oxygen in the atmosphere.
• 02:30 - 03:00 In all these senses, control over the AI system comes from the machine’s uncertainty about what the true objective is. And it’s when you build machines that believe with certainty that they have the objective, that’s when you get this sort of psychopathic behavior. And I think we see the same thing in humans. What happens when general purpose AI hits the real economy? How do things change? Can we adapt? This is a very old point.
• 03:00 - 03:30 Amazingly, Aristotle actually has a passage where he says, look, if we had fully automated weaving machines and plectrums that could pluck the lyre and produce music without any humans, then we wouldn’t need any workers. That idea, which I think it was Keynes who called it technological unemployment in 1930, is very obvious to people. They think, yeah, of course, if the machine does the work, then I'm going to be unemployed. You can think about the warehouses that companies are currently operating for e-commerce, they are half automated.
• 03:30 - 04:00 The way it works is that an old warehouse— where you’ve got tons of stuff piled up all over the place and humans go and rummage around and then bring it back and send it off— there’s a robot who goes and gets the shelving unit that contains the thing that you need, but the human has to pick the object out of the bin or off the shelf, because that’s still too difficult. But, at the same time, would you make a robot that is accurate enough to be able to pick pretty much any object within a very wide variety of objects that you can buy?
• 04:00 - 04:30 That would, at a stroke, eliminate 3 or 4 million jobs? There's an interesting story that E.M. Forster wrote, where everyone is entirely machine dependent. The story is really about the fact that if you hand over the management of your civilization to machines, you then lose the incentive to understand it yourself or to teach the next generation how to understand it. You can see “WALL-E” actually as a modern version, where everyone is enfeebled and infantilized by the machine,
• 04:30 - 05:00 and that hasn’t been possible up to now. We put a lot of our civilization into books, but the books can’t run it for us. And so we always have to teach the next generation. If you work it out, it’s about a trillion person years of teaching and learning and an unbroken chain that goes back tens of thousands of generations. What happens if that chain breaks? I think that’s something we have to understand as AI moves forward. The actual date of arrival of general purpose AI— you’re not going to be able to pinpoint, it isn’t a single day.
• 05:00 - 05:30 It’s also not the case that it’s all or nothing. The impact is going to be increasing. So with every advance in AI, it significantly expands the range of tasks. So in that sense, I think most experts say by the end of the century, we’re very, very likely to have general purpose AI. The median is something around 2045. I'm a little more on the conservative side. I think the problem is harder than we think. I like what John McAfee, he was one of the founders of AI,
• 05:30 - 06:00 when he was asked this question, he said, somewhere between five and 500 years. And we're going to need, I think, several Einsteins to make it happen.