Thread by @IntuitMachine

@pmddomingos It's also the same ignorance that leads to wild expectations when the algorithm games the results. Ignorance like naivety is a two-sided blade.

@pmddomingos The progress we make in deep learning is a consequence of our overall ignorance about general intelligence. There are many alternative ideas on cognition developed by other fields. But these were done without the benefit of computational models.

@pmddomingos It is the combination of empirical AI (i.e. Deep Learning) and theoretical formulation (i.e. Cognitive science, biology, complexity science etc) that lends us a more systematic strategy towards discovery.

@pmddomingos The sad state of affairs is that people trained exclusively in machine learning have outsized confidence in their abilities. The ability to code and run experiments is just half of the problem.

@pmddomingos Science evolves through experimentation, interpretation, hypothesis making, and back again. The low-hanging fruit is experimentation. It's difficult to perform interpretation without adequate models. It's the latter part where many other fields can contribute vastly.

@pmddomingos But like everything of complexity, evolution and development proceeds through messy memetic-driven information propagation. The experiments that gain the most attention are the ones that align closely with the most popular meme.

@pmddomingos But history has shown that a popular meme (otherwise known as consensus) is not necessarily the correct model. The notion of aether was a popular meme at the beginning of the 20th century. The same could be said about GOFAI at the end of the 20th century.

@pmddomingos The reason why Rutherford was right about science progressing one funeral at a time is that memes are propagated by the hosts that adapted them. It's just like a virus. If a virus kills most of its hosts too quickly, it fails to propagate.

@pmddomingos Therefore we must be very vigilant about meme wars in science. Science today is extremely siloed as a consequence of the competition for scarce R&D funding. The popularity of memes is what gets one funding.

@pmddomingos But reality only favors the memes that are useful. Snippets of DNA continue to propagate within a species if they are useful. The detrimental kinds eventually lead to dead ends. This is the same for science where certain memes reach a dead end.

@pmddomingos @threadreaderapp unroll