In the late 90s I had a short stint doing freelance work for a "stupid money" investor who funded the work of this weirdo post-modernist programmer/philosopher claiming to be developing something akin, I guess, to what we would now call a deep learning system.
The claim was that the system this person was developing, when introduced to a lot of information, would emerge with "interesting" behaviour and insight. The problem was that nobody except for the author knew how to interface with that system, which was just a dense and highly idiosyncratic codebase in C manipulating data structures. And so my role was to learn about that system and build a UI that will allow mere mortals to use it.
After a few frustrating and very confusing weeks not really managing to understand what that system actually does (and therefore what an appropriate user interface would be) I finally got to read through the code (taking advantage of the main author, who normally guarded his creation and didn't allow anyone to look at it too closely, being absent for a few days). It quickly became clear to me that this system doesn't really do anything useful at all. It was more post-modern poetry than code.
When I tried to talk to the investor about the situation I was made to understand that this is not something we're going to be discussing openly. I guess he sort of knew but for whatever reasons liked keeping things ambiguous. In order to be able to deliver _something_ for my fees, I prepared a cute little Windows UI that technically connected to that system but really only used it as a random number generator, had some buttons and sliders, and presented all kinds of cool graphics. They were very happy with the result and I soon moved on to other works.
The claim was that the system this person was developing, when introduced to a lot of information, would emerge with "interesting" behaviour and insight. The problem was that nobody except for the author knew how to interface with that system, which was just a dense and highly idiosyncratic codebase in C manipulating data structures. And so my role was to learn about that system and build a UI that will allow mere mortals to use it.
After a few frustrating and very confusing weeks not really managing to understand what that system actually does (and therefore what an appropriate user interface would be) I finally got to read through the code (taking advantage of the main author, who normally guarded his creation and didn't allow anyone to look at it too closely, being absent for a few days). It quickly became clear to me that this system doesn't really do anything useful at all. It was more post-modern poetry than code.
When I tried to talk to the investor about the situation I was made to understand that this is not something we're going to be discussing openly. I guess he sort of knew but for whatever reasons liked keeping things ambiguous. In order to be able to deliver _something_ for my fees, I prepared a cute little Windows UI that technically connected to that system but really only used it as a random number generator, had some buttons and sliders, and presented all kinds of cool graphics. They were very happy with the result and I soon moved on to other works.