In the early 80's where I worked the room was filled with as many as 60 personnel all sorting through stock trade invoices, to then mail to the respective customer and their broker. Commissions were at least $110. Quotron was the system used for Information Management of customer account records and allowed for us to read the newsticker, read quotes, or send short hand email messages as a way to reduce bandwidth and cost.
One day an inner office memo was issued simply stating how the entire department was going to be replaced by a computer. At that time the Dow struggled to reach 1000.
I was eager to change jobs and in the process increase my pay by 30%. As I adjusted to the new environment I found that given the age of the institution that some areas of the office were still in, or as original as the day the office first opened. There at a wooden switch board was likely one of the oldest workers manning the switch board with dedicated line(s) to the exchange floor in order to execute trades with the floor traders. If it was sent morse code, I'm not sure.
Years later, and with an additional background in computer modeling (Fischer Black/Myron Sholes - The Rational Price of an Option) I pursued a new interest as it pertained to an IBM fellows work regarding a developing technology in which the premise was to use computers as a method to program other computers. That study was refereed to as Machines That Learn.
Before delving into a years learning experience playing backgammon with “mloner” a machine that learned, I kept a ear close to the development of the options models used to compute the rational price of an option. Years before "Black Monday"
We had developed on our own idepenently, the model based on the first paper written and submitted to the CBOE at Kiewit Computation Center at Dartmouth College, in Hanover, NH. In order to input data I had to use the services of Interactive Data Services in Waltham, MA to capture stock prices and then transfer those prices to punch cards that we could load into the database in order to test the pricing of options as a function of price movement. If you like main frames then seeing an IBM System 70 was always worth the visit. We determined that the only way to determine variance was to solve for it. In short, it left us believing the strategy, yet short the confidence to follow through with investment. After discussing the work with a floor trader at the CBOE he made comment that they all used the model and that enough capital was required to keep the trades up to the pace of the models output each day. Around $20k.
On the other front of trading was a system billed as the largest super computer system this side of the Mississippi. It was designed to follow every trade and calculate every position held.(example: Their high powered scanners look at every single tick and trade on every US market )
In total the experiences left me overwhelmed. The first computer system replaced an entire office floor of workers. The second system was dwarfed by an even larger system.
After which in time I would play backgammon with mloner a neural net system that as it learned my errors could present problems I would fail to solve. Whereby the more one plays a neural net computer the more it learns its opponent. Most would expect the machine is programmed to play the most effective strategy in order to win the most points. Which it always accomplishes. But, mloner would fix the dice rolls and leave you standing in your shoes everytime wondering, suffering at the heat of your own inconsistencies of learning to play Backgammon the way it was really meant to be played.
Todays market, the bailout, bankruptcy, too big to fail, is reminiscent of the old times so often forgotten.