JetTV™ - Leveraging Indirection - Getting What We Want or Need by Intermediary

This was a big week for me, as I finally defended my Ph.D. dissertation and obtained the approval of all four of my committee members. I'm tired, but the excitement of the week leaves me struggling to wind down.

One of the highlights of my research this past year was discovering that pointing a camera at a specialized sensor, rather than a complex natural scene, can be an incredibly efficient use of resources. I called this "biomimetic processing," and it saves a lot of training time while simplifying the hardware needed for making decisions in real-world situations, called 'inferencing' in the trade.

Roll Pitch and Yaw in the Inner Ear - We were made to fly!

Although my work includes a mathematical innovation in machine learning, I believe that the biomimetic principle that facilitated my work is more important for breaking down a robotic design into manageable pieces of functionality.

This introduces the first example of 'indirection' in this article – not directly getting what you want but achieving it through a route with an agent in the middle that makes life easier for everyone involved in the design, creation, and manufacturing process. It even simplifies things for those who service the robots, but that's a tangent.

In my dissertation work, I developed an autopilot for aircraft using Machine Learning (ML). Initially, I planned to mount a camera outside the aircraft and process the images it captured, feeding the data to both traditional and reinforcement learning controllers for comparison.

However, halfway through the project, I found that having the camera outside the aircraft:

• Was easily fooled into seeing a false horizon
• Generated a lot of unnecessary data
• Created a massive ML training set that was expensive to process
• Required pricey hardware for real-time corrections
• Limited the aircraft to only flying in good weather conditions (VFR)

Moreover, after all this trouble was taken, one never obtained more than a 'VFR', that is Visual Flight Rules, more aptly called, 'Fair Weather Only Flight' which was disconcerting, to say the least.

Surprisingly, moving the camera inside the aircraft and focusing on a single cockpit instrument – the roll indicator – had several benefits:

• It was more error-tolerant
• Generated only the necessary data (the bank angle)
• Created a smaller, more affordable ML training set
• Allowed for deployment on cheaper hardware
• Enabled flight in various weather conditions, except tornadoes and hailstorms (IFR)

Moreover, when this simple act of putting a specialized sensor - the roll indicator - between the world and the ML, one immediately obtained 'IFR' compatibility, that is, Instrument Flight Rules, more aptly called, 'Flight in Any Condition Except Tornadoes and Hail'.

This counterintuitive solution was a game-changer. By placing an intermediate specialized sensor (the roll indicator) between the world and the ML system, we achieved "indirection" – a term used in computer science for storing the address of something rather than the thing itself. 

I encountered a second example of 'indirection' while trying to organize my resume files, 85 of them collected over 37 years. Initially, I named them in a way that didn't sort by year, which didn't satisfy my perfectionist tendencies. Instead of using traditional text-processing tools like awk, grep, and sed, I tried using ChatGPT4 (which my wife jokingly calls "Jet TV") to rename the files. To my surprise, it worked quickly and efficiently, saving me time and effort.

However, when I attempted to have ChatGPT4 organize the content of my resumes, we struggled. That's when I realized that the key to using this tool is not to have it do the work for you but to help you understand the process and, when necessary, provide the code for you to run independently. This subtle shift in perspective is another example of 'indirection,' and I plan to explore it further in the days and weeks ahead.

As we enjoyed our Indian dinner, my wife and I discussed the rise in JetTV's "Opportunity SPAM" – numerous offers for passive income through affiliate marketing. However, for programmers, creators, or designers seeking leverage or a fresh perspective, JetTV can be a game-changer. It's like having the personal assistant or advisor we always needed, now at our fingertips. It's important to approach it as a tool that provides us with the best workflow before embarking on our mission to create and innovate, rather than solely relying on it to do the work for us. In these exciting times, embracing indirection and learning to make the most of this powerful AI can help us save time, increase productivity, and ultimately, make a greater impact in our respective fields. Perhaps this is a return to the presence of an agent or intermediary, formerly called 'the middleman', that the first web took away. But in doing so we could see a return of what was formerly called the 'man in the middle' attack. We'll see.