Kasia Borowska, co-founder and managing director at Brainpool AI took us through an interesting analogy between vipassana and machine learning to attain equanimity, at Rising 2020. With a degree in mathematics and cognitive sciences, she is also deeply immersed in the world of vipassana, India’s most ancient meditation techniques which has been in practice for more than 2500 years. And it was during her learnings in meditation, she found that machine learning is in fact, in many ways similar to attaining vipassana. 

“Understanding that your brain is an extremely sophisticated machine learning system can help us acknowledge and come to peace with whatever happens to us,” she said. 

The Uncanny Similarities

Highlighting some of the similarities between meditation and machine learning she said that both have become popular in a decade’s time and while a lot is known about them in a theoretical world, there hasn’t been enough translation of that knowledge into actual practice. Corporates see machine learning as one of the key technologies to be applied in organisations, but incorporating it comes with many challenges — scaling it, integrating it into systems to name a few. 

Further explaining the similarity, she said, everything we experience with our senses is our input data, everything we think and do is an output. What happens inside of the ‘black box’ which we call our brain is still a mystery, and vipassana can give us an insight into these complex algorithms that are trained in the brain throughout our lifetime. 

Talking about how these two work, she said that any ML system works by inputting data, making and bettering the algorithms and generating outputs. It is the same way as the brain works where sensory inputs are the data which is processed by the brain to give output in the form of thoughts and actions. 

Conscious And Subconscious Minds And Machines

Just the way the human brain has conscious and subconscious ways of thinking, the machines may process data in a similar way which are called priming and bias in systems. She pointed out a few areas where humans see bias and may creep in machines in a similar way: 

• Familiarity – Humans often tend to like pictures or pick options that we might have consciously or subconsciously seen before. Even though we may not remember seeing it, the brain processes in a way of familiarity
• Positioning – Human brain work on the concept of the positioning where it works on picking an option from similar sets without a legit explanation of how the brain processed it
• Facial width and symmetry – Human brain often tend to conclude how a person is if he/she is trustworthy by analysing facial features

These biases which are in-built in humans may be found in machines. Similarly, priming is also a way that defines how machines work. It is interesting to note that priming and bias which are the features or humans has also been picked by AI. The way humans are primed about certain content, for instance, in social media, machine learning systems work on these priming as well. 

Further expanding on the comparison, she said, just the way in vipassana practice, one needs to sense and feel everything, which may sound simple but is the most challenging part of meditation. Similarly pre-processing of data in machine learning systems is the most important and challenging part. 

All the concepts of machine learning, such as reinforcement learning can find a connection with how meditation works. She further highlighted that understanding the simplicity of the input-output, observing and accepting the actions can help us find the path to equanimity.