Moravec’s paradox and generative AI
AIs have been generating text, audio and images for many years. However, the quality and capability of AIs have improved significantly in recent years due to advances in AI algorithms. Prior to these advances, AIs could not generate quality content as they do today, and it was difficult to imagine how they could do so in the future.
Today, generative AIs have become more sophisticated and capable, leading to the creation of AI systems that can generate high-quality original content in a variety of ways.
However, generative AI could be reinforcing a very famous idea in the AI field: Moravec’s paradox.
Moravec’s paradox
In the 1980s, many AI researchers noticed that seemingly difficult tasks involving advanced reasoning were becoming increasingly easy for computers, and that seemingly simple tasks involving sensorimotor skills could become very difficult for computer systems. This is the idea behind Moravec’s paradox.
If you want to beat Magnus Carlsen, the world chess champion, you choose a computer. If you want to save the pieces after the game, you choose a human — Larry Elliott.
The paradox is named after robotics researcher Hans Moravec. Although other scientists such as Rodney Brooks and Marvin Minsky also postulated this principle. Moravec wrote that “it is comparatively easy to get computers to show adult human-like abilities on intelligence tests, and difficult or impossible to get them to possess the perceptual and motor skills of a one-year-old baby.”
Also, Minsky wrote “In general, we are not aware of our best abilities,” adding that “we are more aware of small processes that cost us than of complex ones that go smoothly.”
Hence, we can summarize Moravec’s paradox as the idea that hard problems for humans are easy for computers and hard problems for computers are easy for humans. But what does the Moravec paradox have to do with generative AI?
Generative AI and Moravec’s paradox
It was millions of years of evolution that contributed to our sensorimotor skills necessary for our survival. Abstract thinking (thinking involving logic and mathematics) is more recent to us and easier to design into AI systems since its basis are logic and mathematics.
The tasks that generative AIs perform, such as writing text, generating images, generating audio or generating music, are tasks that require a lot of abstract thinking for humans to perform.
And by Moravec’s paradox, it is easier to design AI algorithms to do these tasks on a computer than to design them to perform the same tasks but making use of sensory and motor skills. That is, it is comparatively easier to have an AI system compose and generate an original song, than to have a computer system or robot perform or play a song.
And that is how generative AI reinforces Moravec’s paradox. Generative AI has made more evident the idea that things that are difficult for humans are easy for computers and vice versa.
What kind of future do we want to achieve with this technology?
A study conducted by OpenAI to study the potential impact on the labor market of large language models. One of the things the study found was that the less education a job requires the less danger AI has to affect it and the more education a job requires the more danger AI has to affect it. So, what does this idea resemble? Exactly, the Moravec paradox.
It is the jobs that require more education that are in danger of being replaced by these intelligent systems. Gardeners, receptionists and cooks have jobs secured for the time being. But is this what we want?
With a technology with the potential of Artificial Intelligence it is very important to ask “What kind of future do we want to achieve with this technology?”. A few years ago it seemed that the idea of what we wanted to achieve was that with machines, boring, manual or monotonous work would be automated and that humans would be doing the creative and decision making work. However, it seems that we are heading in the opposite direction.
Conclussion
In conclusion, generative AI reinforces and makes more evident Moravec’s paradox that machines are becoming more capable and skilled at tasks that require a lot of abstract thinking and creativity for humans. Thus, computers are becoming increasingly skilled and capable at things that seemed to be reserved for humans. However, we cannot predict precisely how all this will play out, because if we can learn anything from the history of AI, it is that it is impossible to predict what comes next.
