Machine Learning & Hunter-Gatherer Children

This is an interesting article by Gul Deniz Salali. A British Academy research fellow and lecturer in evolutionary anthropology/medicine at University College London. She does research on human behavior and health using evolutionary approaches. She studies modern hunter-gatherer communities and teaches evolutionary medicine at UCL.

Machine learning and its relation to human learning.

AlphaGo is the first computer program to defeat a professional human Go player, the first to defeat a Go world champion, and is arguably the strongest Go player in history. Go is known as the most challenging classical game for artificial intelligence because of its complexity. 

Go originated in China over 3,000 years ago. Winning this board game requires multiple layers of strategic thinking. That is why when DeepMind’s AlphaGo program defeated its human competitor the AI scene was shaken.

There are striking similarities between new-generation machine learning technologies and how children learn skills without access to formal education.

Salali does her field research in hunter- gatherer communities in the Congo. In these communities people do not often give instructions directly when teaching their children. Instead, they create a learning opportunity, for example, they provide tools and monitor the child without interfering. With feedback about his/her performance, the child can adjust behavior. Learning skills through practice and feedback, proves that empiricism is best!. This kind of learning, without a set of pre-written instructions is how the AlphaGo program was trained to beat a professional human Go competitor.

In AI research, the ultimate goal is to generate artificial general intelligence (AGI), that is a machine that can understand and learn as humans do. AI researchers such as the DeepMind team, believe that this will be possible through more independent learning strategies. Salali gives the example of unsupervised learning. Machines learning by observing data without a predetermined goal or explicit guidance.  It is a form of learning parallel with how hunter-gatherer children learn most skills.

It’s interesting to learn about how we pass on skills when formal education is not available. Understanding this can help us to understand how complex cultural practices evolve, such as the Go game. We have evolved a great capacity for learning by imitating others. Researchers found that when information is transmitted faithfully, cultural practices remain in the population long enough so that they can be modified to generate more complex practices. This is how human culture progresses. Our cultural traits are built on the legacy of past information. This also means that acquired traits may also be restricted by them.

New training algorithms in machine learning have parallels with how human children learn but the key difference is that they have the capacity to surpass human culture. These new algorithms are not bound by the legacy of our cultural history. In 2017, DeepMind  introduced AlphaGo Zero, the version of AlphaGo that became its own teacher by learning from self-play. It is now considered the best Go player in the world.

Human Go players have been building their strategies on 3,000 years of built up knowledge. AlphaGo Zero became the best Go player by setting itself free from this knowledge.

So what can we learn from this? There’s a lot being done with how we understand artificial intelligence that runs parallel with how we understand ourselves. The ways in which we have evolved to be complex thinking creatures has created a solid foundation for how we continue to grow. However, this legacy can hold us back. The AlphaGo program was able to beat a professional human Go player by teaching the game to itself. From this we can understand younger people’s eagerness to experiment and develop on their own as a way of evolving, and we can help them and give them feedback. No clear-cut instructions necessary.

Meanwhile, I’m left thinking about the concept of machine learning becoming more advanced than human learning.  A nagging thought – humans make computers! Yet, in the search for whatever it is that we are looking for… Creating machines that learn beyond our understanding, does it give us pause?  I just wonder, are we creating an all knowing being to look out for us? ‘Wall-E’ comes to mind!  Conclusion?  To paraphrase Browning, a man’s reach should exceed his grasp.  A new frontier is a welcome site.

All this talking about machine learning is exciting. Want to know what else is exciting in technology? Augmented and Virtual Reality! American Movie Company has the best quality AR/VR Production in NYC.  Check it out!