Hacker News Reader

Comment by UncleOxidant | Hacker News Reader

It's interesting to see that neural nets were there from the beginning of the computing age. McColloch, Pitts and von Neumann were thinking of a connectionist approach from the start.

Such a sad end for Pitts. I couldn't help thinking that he was kind of like Will in Good Will Hunting - a complete outsider, a self-taught genius with a very rough upbringing and lots of demons from that past.

View on HN

George Boole believed in a similar idea far before any idea of computers were known. In a sense then the entire endeavor of computation has to do with consciousness from beginning to end.

The spiritual successor to Pitts-- Frank Rosenblatt-- has a similarly dramatic end. https://news.cornell.edu/stories/2019/09/professors-perceptr...

It's not in the article, but Rosenblatt died in a freak boating accident before his views were revived in the 80s.

It's very sad Rosenblatt did not live to see the resurgence of neural networks and his perceptron algorithm. The perceptron algorithm isn't exactly what we use to train neural networks today, but it's similar enough in theory and practice that it still feels very fundamental to understanding machine learning.

And at least in the CogSci space, Rosenblatt's work was instrumental for the PDP (Parallel Distributed Processing) working group in the mid eighties that led to backpropogation methods.

And neural nets are just graphical models which has been around since the early 1900s. I have a theory that everything new in data analysis today has already been invented 50-100 years ago by statisticians.

Reply on news.ycombinator.com