Saturday , June 05, 2021
You would not be surprised to see a Bengal Tiger in the Mangrove forests and a ceramic mug in the kitchen. Based on your prior experiences and knowledge, you know that is where tigers and mugs are often to be found.
However if I say you saw a Bengal Tiger in your kitchen and a ceramic mug in the mangrove forests, you will be shocked! If you saw a mysterious object in your kitchen, how would you figure out what it was? You would rely on your expectations or prior knowledge. Should a computer approach the problem in the same way? The answer may surprise you.
Rationalists are people who view the world with a prior knowledge, they expect a mug to be in kitchen, tigers in the mangrove forests, elephants in the Savanna etc.
On the other hand, Empiricists are people who view the data exactly as it is presented. When they visit the mangrove forests, they no more expect to see a tiger than they do a mug.
If a rationalist came across this set of data points in the kitchen, he/she is most likely to believe it as a mug. Their prior knowledge states that a mug is likely to be found in a kitchen; it is highly unlikely to find a tiger. They have never seen this situation before, nor have they ever learned that such a situation could occur. Although their result takes in a certain amount of the data, it leaves out other parts. In this case, their methods have produced an incorrect result: a mug.
However, when an empiricist sees the same data, they will analyse it without regard to whether they are in the mangrove forests or their kitchen. They will piece together an image from as many data points as possible.
Whose view is correct?
Neither the empiricist nor the rationalist is wrong. Both approaches work for various kinds of problems. However, in this case, if there is a tiger in the kitchen, it would pay to figure it out as quickly as possible. A middle ground between purely empirical and purely rationalist approaches may be best. With some prior knowledge you could get the idea that it is a tiger in the kitchen, and it is time to leave!!
Data scientists face this sort of problem all the time. They train computers to recognize new objects or patterns. Some machine learning programs may be able to process a lot of information and make many rules to fit the presented data. Just because the pattern is reproducible, that doesn't mean it accurately represents what is happening.
There are historical examples of this dilemma. Two thousand years ago, Ptolemy developed a model of the universe that yielded excellent predictions for the movements of the moon and planets. His model was used successfully for centuries. However, Ptolemy used the wrong prior information: He placed the Earth at the center of the solar system and prioritised the circular motions of celestial objects. Johannes Kepler questioned this view in the 17th century and ultimately rejected Ptolemy's approach, which eventually led to Newton's law of universal gravitation. Although Ptolemy's complex model fit his own observations exceptionally well, it did not accurately represent what was happening. If you want to be an extreme empiricist, you really do need a lot of data. We now understand why under certain circumstances, such an approach can, in fact, succeed in a mathematically rigorous setting. Biological brains, on the other hand, are halfway in between. You do learn from experience, but you're not entirely data-driven.
“I hope that data scientists will look to brain circuitry for inspiration when developing next-generation machine learning approaches. Vertebrate brains have circuits of different sizes, including medium-sized (mesoscale) ones. Those circuits are encoded with priors (known information, such as what animals look like, where they are found, or how to escape quickly from a charging elephant). At the same time, your brain is highly flexible, classifying new information and weighing the importance of different priors based on experience—tiger may not belong in a kitchen, but somehow, you have one anyway.”
~ Anonymous
See you next Saturday, until then have a great weekend :)
Cheers!
Some things that you may find interesting-
Article: Reimagining Twitter by Aashir Sutar
Must Watch: The surprising science of Alpha Males by Frans de Waal
Song I am listening to: All I Know So Far by Pink
Quote I am enjoying: “In this world there are only two tragedies. One is not getting what one wants, and the other is getting it.” ~ Oscar Wilde
Thread of the week: Internet Computer and the Network Nervous System(NNS)
Here are the last three posts if you were too occupied to read them:
If someone has forwarded this to you, please subscribe and start receiving updates directly in your inbox. It’s free. Your information is protected. And I never spam. Ever