What is Moravec’s paradox and what does it mean for modern AI?

Paradox, noun. An absurd or contradictory statement that proves well-founded or true.

It seems almost paradoxical to suggest that a technology ruled by logic — such as AI — could fall prey to paradoxes.

And yet, artificial intelligence isn’t immune to the occasional illogical truth. Moravec’s paradox is one such contradictory reality within the development of AI.

But as artificial intelligence grows in ability, how has this paradox influenced its development? Here, we take a closer look at Moravec’s paradox and what it means for modern AI.

What is Moravec’s paradox?

Moravec’s paradox is a phenomenon surrounding the abilities of AI-powered tools. It observes that tasks humans find complex are easy to teach AI. Compared, that is, to simple, sensorimotor skills that come instinctively to humans.

In the 1980s, Hans Moravec, Rodney Brooks, Marvin Minsky and others articulated and discussed this AI paradox. As Moravec put it:

“It is comparatively easy to make computers exhibit adult level performance […] and difficult or impossible to give them the skills of a one-year-old.”

For example, artificial intelligence can complete tricky logical problems and advanced mathematics.

But the ‘simple’ skills and abilities we learn as babies and toddlers — perception, speech, movement, etc. — require far more computation for an AI to replicate.

In other words, for AI the complex is easy, and the easy is complex.

The logic behind Moravec’s paradox

But why does AI struggle with the simple? The explanation behind Moravec’s paradox revolves around evolution, understanding, and perception.

For a start, the skills that we define as ‘simple’ — those we learn instinctively — are products of years and years of evolution. So, while they may appear simple, it’s only because of billions of years’ worth of tuning.

In other words, the complexity of the simple abilities we take for granted is invisible.

Plus, AI ‘learns’ through us telling it how to do things. We’ve consciously learned how to do mathematics, win games and follow logic. We know the steps (computations) needed to complete these tasks. And so, we can teach them to AI.

But how do you tell anything how to see, hear, or move?

We don’t consciously know all the computations needed to complete these tasks. These skills are not broken down into logical steps to feed into an AI. As such, teaching them to an AI is extremely difficult.

Moravec’s paradox and the AI of the past

The history of AI has seen an impact from Moravec’s paradox. In fact, it’s arguably a factor that held back development and contributed to the AI effect.

The AI effect is a phenomenon that has seen AI-powered tools lose their ‘AI’ label over time, due to not being ‘true’ intelligence. Moravec’s paradox could have contributed to this. That is, the reason these tools lost their ‘intelligent’ status is that the tasks it does are simple, once you break them down. 

No matter how good AI tools and programs got at games and logic, thanks to Moravec’s paradox, they couldn’t complete ‘basic’ human tasks.

How could anything that can’t replicate the behaviour of a toddler be ‘intelligent?’

Moravec’s paradox and modern AI

Moravec’s paradox explains why AI capable of adult-level reasoning is old hat, but AI vision, listening, and learning is new and exciting.

Indeed, things are changing.

Now, AI is overcoming Moravec’s paradox. Higher-level artificial intelligence is beginning to replicate our evolutionary abilities.

For instance, we are beginning to see AI tools like image classification and facial recognition — that is, a machine’s equivalent of sight. 

Meanwhile, personal assistants like Alexa are an example of AI becoming capable of ‘hearing’ and understanding us. This is thanks to natural language processing (NLP).

Similarly, AI is becoming capable of speech, as with these assistants, or advancements like Google Duplex.

AI future

Artificial intelligence has seen highs and lows. It’s a field saturated with ethical questions, problems, and the occasional paradox for good measure.

But modern AI is answering Moravec’s paradox. We’re starting to teach AI the complex ‘simple’ skills. In doing so, AI is finally becoming human-like. In turn, this could mean a more stable future for the technology — one free of disillusionment and AI winters.  

Perhaps we just needed to overcome Moravec’s paradox before we could truly have AI. 

Further reading