Comments by "Louis Giokas" (@louisgiokas2206) on "Veritasium" channel.

Mathematics conquers all! In the mid-1990s I was working for IBM. I had a neighbor who was a seasoned currency trader. He was planning to use machine learning algorithms to automate trading and with a few friends who were also currency traders, went out on his own. As a favor to him I reviewed and analyzed the technologies and algorithms he was licensing. I found them to be good. So, they set up their system. First, they trained it and then let it "trade" without actually driving any of their actual activity. It performed brilliantly. Then they started to use it, with one caveat. One of them would have to actually authorize whatever action the system indicated. They let their years of experience get in the way, and often didn't follow the algorithm's indicated actions. Thus, they didn't do all that well.
7
What Prof. Lo says at about 30:15 brings to mind another IBM related story. At one point they were involved in building what would have been the fastest supercomputer in the world at the University of Illinois. I went to a seminar there and got to see the system being built and got information on the technology. I was, from my time at IBM, familiar with a lot of this. One of the salient points of this new supercomputer was that it was made of a number of large modules. At one point, IBM determined that one of the main requirements for the machine was unobtainable and withdrew from the project. This was not a loss for them at all. Each module, consisting of two large cabinets joined together, was a formidable high-performance computer in itself. They sold a lot of them to Wall Street trading firms.
4
In the 1980s I was working at an aerospace and defense firm. Although I was already a senior engineer, I did not have a degree. I was taking classes at night (paid for by my employer). I was using FFTs, and had books on them. I wanted to get a more theoretical background, so I took an advanced calculus class, which was about transforms. When we got to the Fourier transform, I was perplexed. I went to my professor and asked when we would get to the FFT. He looked at me puzzled and said he didn't know what that was (he was a theoretician). Now I was concerned. So, next class I told him it was a fast way to compute the Fourier transform. His response was that such a thing was of no interest to him. At least I was relieved that I was totally crazy.
1
The whole premise is silly. What matters about math is not some intrinsic property, but what it can be used for. It is used quite successfully. For mathematicians exploring these things makes sense and is natural. Saying this is a flaw in this venue is alarmist and does not really mean anything to someone using math or benefiting from math.
1
@disputin9930 Oh, come one man (as President Joe likes to say), it was just a joke. I do hope you took it that way. I do like your example.
1
@PinkeySuavo What do you think all these quantitative traders are doing? What do you think statistics is all about? It is about predicting the trend and the odds, not an individual next event. Actually, though, the code can influence the share price if it controls enough of the market. This has been a concern for some time, and you should have gotten that from the video.
1
We only made progress when we went to digital. Analog computers were used in things like simulators and spacecraft control systems, and these had limitations. That work went to digital computers (I was part of that, in both cases). It improved accuracy and speed. Most of what you do on a computer is managing information, not processing images. Analog circuits may become like a co-processor function, just like quantum computers. Even quantum computers need digital computers to prepare the data and the state of the system.
1