Throughout history, certain technologies have revolutionized the way we live, work and play. This often happened gradually, but occasionally also through dramatic, profound changes in society. The technology behind the Industrial Revolution—for example, the steam engine, the spinning mill, and the cotton machine—accelerated economic growth by an order of magnitude and sparked an explosion of innovation that still reverberates around the world.
In contrast, other technologies have had more modest impacts. The invention of electricity, cheap air travel, computers, etc. were all important and revolutionary, but the economic growth they enabled has always been limited by other factors that prevented the explosive growth of the Industrial Revolution.
This raises an interesting question about the impact of artificial intelligence. This novel technology is currently taking the planet by storm, and there’s no shortage of hype about its potential to change the way work is done. But will this revolution lead to explosive economic growth or will some other factor limit the ultimate impact of AI?
Bottleneck science
Today we get an answer of sorts thanks to the work of Ege Erdil and Tamay Besiroglu, both at Epoch, an Open Philanthropy-funded research group that studies and predicts the development of advanced AI. They investigate how AI can bring about explosive growth and what mechanisms could prevent this.
“We conclude that explosive growth seems plausible if AI is able to largely replace human labor, but high confidence in this claim does not seem warranted at this time,” they say, adding that potential bottlenecks are better must be understood.
The main reason AI could lead to explosive economic growth is that it changes the nature of work. Currently, economic growth is possible through reinvestment and improvements in the factors that lead to growth in the first place. These include land, energy, computing power, access to capital, human labor, and so on.
Economic activity requires all of these factors, and through reinvestment it is always possible to buy or redevelop land, build more power plants, increase computing power, and so on. Economists refer to these as accumulating resources.
But human labor cannot be accumulated because it is limited by demographic factors such as birth rates and immigration. The availability of human labor therefore sets fundamental limits on the speed of economic growth.
But artificial intelligence changes this calculus as soon as it is able to replace human labor. At this point, it’s just as easy to double or triple the workforce as it is to double or triple the computing power or space or energy consumption. Just buy more of it, sit back and watch the explosive growth it enables.
Economists are fascinated by this idea. However, many have pointed out that there may be other factors preventing explosive growth. That’s why Erdil and Besiroglu examined these arguments to estimate how likely they are to thwart the parties’ plans.
It is highly unlikely that many of these arguments could prevent explosive growth. One argument is that growth is fundamentally limited by the laws of physics and the limited resources of our planet. For example, there are limits to the speed at which information can flow from one volume of space to another and a limit to the amount of energy we can extract from the environment.
In response, Erdil and Besiroglu say that there is no question that these are real and important borders, but that we are so far from them that they cannot represent significant bottlenecks in the foreseeable future.
For example, the Sun currently floods the Earth with an output of about 1016 W. But we currently use significantly less, about 1013 W. So we have three orders of magnitude to play with and numerous efficiency improvements that we can implement. Access to energy is unlikely to slow economic growth in the foreseeable future.
Another possibility is that humans decide they want human-produced goods and therefore refuse to engage with AI products, limiting growth. In their analysis, Erdil and Besiroglu admit that there may be a preference for certain man-made goods such as works of art. But on the whole, people are unlikely to mind if computer chips, cell phones or cars are made entirely by intelligent machines, especially if they are significantly cheaper.
But other arguments are harder to dismiss. One is that achieving human achievement will be slower than expected and that this will limit possible growth. This argument suggests that some tasks will be easy to automate, but others will be much more difficult and require orders of magnitude greater investment.
One example is that it may prove much more difficult than expected to build robots with human-like abilities and even more difficult to integrate them safely and productively into society.
But even if this were the case, it is likely that the problem will be resolved at some point. So such a problem will delay the explosive growth instead of preventing it. “We assume that this objection is unlikely to block explosive growth,” say Erdil and Besiroglu.
One of the biggest potential obstacles is the problem of aligning AI systems with human expectations and goals. AI targeting is “the challenge of controlling artificial intelligence systems to behave according to intended goals and avoid unintended harmful behavior.”
It is already clear that current AI systems suffer from problems such as hallucinations, which require human oversight to resolve. Other safety issues may arise with robots, etc. So human supervision may always be required. And that this creates a bottleneck that prevents explosive economic growth.
Erdil and Besiroglu point out that an important factor is the importance of the problems that arise. AI systems are already better at preventing hallucinations and may be able to overcome other problems as well. But it is difficult to estimate how these factors will play out in the longer term. “Overall, we believe that this argument is unlikely to block explosive growth, but its influence cannot be ruled out,” they say.
Riddles about the legislature
Perhaps the most important factor that could prevent explosive growth is regulation. Policymakers around the world are currently scrambling to enact laws to prevent the use of artificial intelligence by malicious actors and protect consumers. But regulation could also serve to protect jobs and guarantee employment for people.
Regulation could also limit the material that AI systems can use for training, for example by preventing AI systems from using professionally prepared text or images without significant compensation to the owners, thereby reducing the incentive to invest in AI systems. investing in systems. Regulations could limit the use of AI in certain activities, such as in courtrooms and medical facilities, or prevent the technology from being exported to prevent competition.
Researchers look to history to find out how regulation has influenced the spread of previous technologies. They point out that much of the technology for the Industrial Revolution was developed in England in the 18th and 19th centuries.
“While England tried to prevent the spread of some key technologies of the Industrial Revolution by banning the emigration of skilled workers and the export of machinery, these protectionist measures proved largely ineffective,” say Erdil and Besiroglu. Still, they conclude that regulation has up to a 1 in 3 chance of preventing explosive economic growth.
Given all these potential problems, the authors say the chances of AI-powered explosive growth by the end of the century are far from a foregone conclusion, but are equally likely. And they say it could start soon if advances in AI continue at their current pace.
This is interesting work with profound implications. If the analysis proves broadly correct, the importance of directing this economic force to improve human flourishing cannot be overstated, especially given that the outcome depends heavily on the choices we make now and in… meet in the coming years. Policymakers are taking note.
Ref: Explosive Growth Through AI Automation: An Overview of the Arguments: arxiv.org/abs/2309.11690
Comments are closed.