The ongoing conversation around AI in the world has so far been limited to competition between labs or abstract debates about the capabilities of the technology. Almost no one is asking what purpose should AI serve? Or are our present institutions and control systems capable of taking this technology in the direction of the welfare of humanity? Recently Pope Leo described AI as a deep threat to human dignity. As an economist—who has long argued that the consequences of technology should not be our destiny, but our choice—I welcome the Pope’s intervention. What most analysts fail to understand is that technology is never neutral; it takes on the characteristics of those who create it, fund it, and control it. In such a situation, the most important question before us today is what should AI be designed to do? Technologies like AI can take many paths, each of which has far-reaching impacts on society. For example, what was taboo just a few years ago—AI-powered mass surveillance or algorithms that select targets for assassination—has now become commonplace. Many in Silicon Valley are now urging that the US consolidate its hard-power through a new military-algorithm complex. But every technology that facilitates attacks without humans seeing faces lowers the ethical limits of conflict. Disarmament of AI is necessary so that it can be freed from the mentality of armed competition. It is today not limited to just the military context, but is also an economic and cognitive phenomenon. We must be aware that technological progress is not necessarily moral progress. If something is technically possible, it does not mean that it will necessarily be good for humanity. Whether a technology is desirable or not depends on who controls it and what ideologies and interests motivate it. While AI increases our productivity by taking over everyday tasks, it also forces workers to adapt to the speed and demands of machines. Whereas what should happen is that machines should be designed to work with workers. The approach of the entire AI industry is focused on mimicking human abilities and automating tasks performed by humans, with the goal of creating an AGI (Artificial General Intelligence) that can do everything a human can do. This philosophy is based on the misconception that machine intelligence and human intelligence are fundamentally the same. Humans are one-shot learners. We form hypotheses from examples, simulate possibilities in our minds, and refine our understanding through the social process of trial and error. This is how children learn language. We’re not very good at taking in information on a large scale or assimilating random data for relevant patterns. In contrast, AI models thrive on huge training sets and become adept at recognizing patterns at scale. But he has not yet demonstrated real creativity. They have no experience of the real world, nor have they shown any capacity to learn through interactions with the physical and social world as we humans do. When two things are so different, you shouldn’t use one to copy the other—and usually you can’t do that. It would be much more productive to use AI for tasks that humans cannot do. So that humans can expand their works. For billions of people in the developing world – where good jobs are the only reliable path out of poverty – an automation-focused AI agenda would be a recipe for disaster. It would be much more productive to use AI for tasks that humans cannot do. For billions of people in the developing world, an automation-focused AI agenda would be a recipe for disaster. (@ProjectSyndicate)
Source link
Deron Acemoglu’s column: Why don’t AI do its work and we humans do ours?
