Skip to main content

Futurist: Moving Ai (Mi) Past the Human Limits

I am a long-time Futurist, and technologist. In my career, I have spanned the birth of personal computers, to the rise of Cloud Computing.

What if we are limiting intelligent machines with the way we design them?

What if we are limiting intelligent machines with the way we design them?

I wonder if we limit Machine Intelligence by building it the way we think.

I'm not to start today with my new futurist tagline, although I'm still using the tag, just a little further on the text this time. Instead, I want to talk about the reality of machine intelligence and the continued push of machine intelligence into the mass market. Part of this is continuing my discussion of automation and why machine intelligence or AI is good. I understand that many people are concerned about AI replacing humans. In particular, they are more concerned about automation that will replace actual human workers. I understand and support that the impact of machine intelligence on some people could be significant. However, the value for many more people could also be extremely important!

One of the things that I think about right now when I think about machine intelligence is the limitation of a neural network. Neural networks simulate or emulate the human mind. The human mind itself is a marvelous machine. While it's not exactly a machine, it's a biological creation, but it functions effectively. So for the sake of this initial argument, a human brain is a machine. I do understand the context, and please don't play me. I do understand that it is not a machine. That is a biological construct. However, in the area where we are talking machine intelligence, the human brain is emulated as we build neural networks. The various Neural networks are replicas of the way humans think.

It might be better to design machine intelligence systems to take advantage of the processing power available everywhere. Rather than the somewhat linear paths of a neural network, allow the machine intelligence to use a less neural and more broadcast-focused network. I do not forget the reality of the fictional "Skynet." In Terminator, Skynet was a machine intelligence created to "prevent war." It also had control of the entire Nuclear Arsenal. Skynet finally realized that it removed humans when everyone was. Humans were the problem in the computations of Skynet. Skynet was a ubiquitous intelligence that spread throughout an existing system. So while I'm advocating for that, I am aware of the potential risk.

From an intelligence perspective, if you think about the way human beings create thoughts, we are taught at a very early age to perform linear tanks. As second and third graders, we memorize the addition and multiplication tables, not how additional worker modification work comes later. Linear education is how the young human mind is best able to learn. 5×5 is 25 that linear thinking and critical for the development of the neural network of a young child. Giving a seven-year-old an equation derived from Einstein's theory of relativity, E=MC2, would probably cause most seven-year-olds to struggle. Some would get it, and all those who later in life would probably be the creators of new theories expanding the reach of physics. But the reality is that most seven-year-olds would struggle with that equation.

A child learns by taking small bits of information and building a framework.

A child's learning capacity is a known quantity. Would that be limited to machine intelligence? The ability to break information into smaller pieces and act on that information to consume information faster because many smaller nodes are interconnected. Instead of utilizing neural pathways, which is how the human mind works, allow more of a broadcast network. A Broadcast network means that information becomes read once broadcast many. A broadcast network expands both the network and the capabilities within the network to act on information. As a result, the system could consume and share information at a much faster non-linear rate.

With this new network, the system would be able to expand continuously. Instead of being limited to single or multiple paths within a structure, it can create a broadcast network and allow information to be moved quickly amidst this distributed intelligence. If you exceed the capacity of someone's ability to learn, they're not going to be able to learn what you q43 teaching successfully. That's why you break the information down into smaller pieces and, over time, build up the ability of the person to handle that intelligence or information. That's why distributed machine intelligence would be far better off. First of all, it could break data into smaller chunks. Then, they could integrate with, develop and consume information much faster. Now, the reality of machine intelligence today is based on models built in the machine learning world or sometimes called deep learning.

I do understand again, going back to Skynet's risk, that there are risks in doing this. The concept would allow machine intelligence to both concentrate and distribute data quickly. A broadcast network expands the amount and capabilities of the MI (AI) to process the information in question. The goal of leveraging Machine Intelligence is to improve the world for humans. We should stop starting with how humans operate as a starting point.

This content reflects the personal opinions of the author. It is accurate and true to the best of the author’s knowledge and should not be substituted for impartial fact or advice in legal, political, or personal matters.

© 2022 DocAndersen