Client Server and LLMs E008

Подробнее о видео

Client Server Computing and LLMs
A Bit of Security for February 22, 2024
A bit of history - the origin of Client Server and Cooperative Processing - and a note on why an LLM could never develop such research.
A few months after I joined Gartner, we have a team offsite meeting. The agenda went a very smoothly and we had some extra time at the end, so our manager, Jeff Schulman, asked if anyone had a topic to address. I asked if we could have Roy Schulte and Tony Percy each speak for 30 minutes on “What Client Server Computing Means to Me” – because whenever they talked, the conversation spiraled out of control. They each grabbed a marker and built a chart or two … and then the unthinkable happened. They both looked at me and asked “What do you think?”
I had called the meeting, so it was my task to resolve it! I said, “The problem is that you are discussing two orthogonal sets of concepts, but you are using the same vocabulary.” Our team leader said we needed to develop a research report on this. We worked for a few weeks (including some time at my home with an editor, Kelly Humiston nee Clark, to have uninterrupted space) and wrote “Client/Server and Cooperative Processing: A Guide for the Perplexed” (A tip of the hat to Moses Maimonides there). In that document we described an application as having a presentation layer, an application layer, and a data layer. We discussed the five styles a two-tier application could have: Distributed Presentation, Remote Presentation, Distributed Application, Remote data, and Distributed Data. The difference was where the parts split. That document published in 1991 and laid the groundwork for n-tiered application design, as well as mapping applications to the most suitable computing platform.
The big question is, could that research have been generated by an LLM? The answer seems to be “NO” – because the LLM would not explore the differences between semantic nuance, but rather try to blend those differences in a bland word porridge. LLMs build statistically average strings of words, so poetry would be smoothed into uniform bland utterances with on semantic sensitivity. If your job requires creativity, and if your boss values creativity, you are entirely safe from any AI taking your role.