[ Blog ]

The Great AI Debate

The revolution wasn't televised. The change is already here. That is great news!

Figure 4-1. Functional block diagram of automated video ingest, encoding pipeline, and downstream distribution network. (KI Flight Operations / MSC Internal Note 67-FS-269)

There is a great debate going on amongst computer programmers about using AI as part of the development process. This blog entry is not about that because listening to computer people argue about which tech is better got boring back in the Apple/TRS-80 debate days.

What isn't a debate is how applications are designed and built. When the internet started, it was every programmer for themselves, grabbing whatever technology would work. The change was rightly invisible to non-technical people, but over time, but the programming languages used on the web went from jury-rigged to built for web application from the ground up. That made it easier for companies to build secure access into their systems while they were writing those systems. This is the real revolution, and it is totally invisible. Developers can use standardized tools to connect disparate systems together, so custom tools don't need to be built just to connect to each system. With that problem solved, developers can focus on the real issue: giving users applications that make their job easier.

One example is a WordPress client who publishes a site that has a lot of video. The videos are edited in Frame.io, then, when approved, downloaded from Frame.io and uploaded to JW Player for hosting and serving the video. Once a video is ingested, a new video page is created in WordPress. They wanted to add the records to the file Google uses to rank videos, as well as update news feeds like Apple News, with the new content.

They had the hard parts of the business figured out. They have great editorial that people want to read and an audience eager to read it. The problem was all the steps in the process meant they couldn't keep their content as fresh as it needs to be, and they weren't able to update their SEO or content partners in a reasonable amount of time. What if just approving the video in Frame.io could start a process that would end up with the new video added to the site and optimized for SEO?

How hard is that to build? It turns out, not nearly as hard as building an online publishing brand, and parts of it can be done without code. Frame.io and JW Player both have systems built to communicate with other systems. A "workflow automation tool" called Zapier was used to manage the communication between Frame.io and JW Player. A custom plugin was built for WordPress, which told WordPress that when it received information from JW Player, a new record of type "Video" should be created and the various sitemaps and syndicated feeds should be updated. At every level, the systems are set up to handle the hard parts; they just need code to act as a switchboard operator to move information around.

I hear from a lot of people interested in improving their processes, or people with processes that don't work but who have no idea that they can even be improved. Because this revolution was invisible they don't know the tools are here to solve their problems.

OK, I started with this article being about the AI debate, but only because nobody is going to read a blog article about APIs. But i think this is more interesting that discussing what type of hammer to use.