The Unlikely Advancement

I’ve been contacted by almost a 100 people of late telling me of their “great new AI that breaks all things out there” moments. Sure, but where’s the proof? That’s what is usually missing and no papers that have been submitted for review as well happens a lot in this instance. If you search for my name, you will find more papers that state McClain, J. than a lot of other people whom have published work. I have some hits, but a lot of misses as someone else beat me to the punch, my ideas were too heavy for production, the market was past my idea by the time I published, or maybe I was just plain wrong. That’s the point, admit your failure before proclaiming success in something you have no training in.

I started out in 1999 working on machines that cooled modular tower equipment. People passed on it, some may have taken some ideas, but it did not work out. Over the years, I created a product that hid email addresses, meta, and header information through proxies and was well encrypted. I was contacted by a group of people from Yemen and the FBI got involved, that died quick. Then I got into voice recognition, yep, another failure; however, I can state I did have a working computer that I could talk to and it did what I wanted it to do. Not as good as today’s technology, but for 2000, not bad. None of these ideas came to market, nor did I say, eureka!

Based on that I had some good ideas, but needed to learn more, I did something radical, I went to school. College was different and easier than other experiences. I found a place where I belonged and that fact mattered. Showing how your product worked mattered. I took years off and on, I had a child, married, and was working on the next big thing. I listened and watched learning as I went on through life. I’m sure a lot of this sound familiar in your own lives, but for me I was a calm experience. I was hired to do some work that was above my college experience, but I was able to do it. Confidence, the correct kind and not the pretend from AI giving me a pat on the back, came and I learned what was important. I worked on my degrees, a total of over 220 credit hours with 3 honor societies memberships along the way. Professors took notice and contacted prior students which got me to the place I am now.

We lose a lot in translation with AI as it brings skills we do not possess to life and it makes us feel confidence we are not ready for under rigor. I have had to go in front of professors and explain my ideas, thesis, and knowledge in a rigor that is not an easy defense for anyone. I do protect my IP, but I state that I will have an independent third party examine my results and have them try to reproduce the results. If they cannot, then I do not want that on my resume, nor do I find it a failure. I was caught before I was made a fool and that’s worth more than running into the marketplace with something that will fail.

Some of my hardest classes were in mathematics. Calculus and linear algebra did not excite me at first, but then I found the purpose of knowing the subjects. They lit up the room with ideas, and I learned that knowing the subjects matter more than ever as the need for my own uses was baked into education. Others guided me to something I thought I did not need.

Many, if not most, have degrees or other higher education of some sort; however, this platform is flooded with people who do not have such knowledge. They read something spit out from ChatGPT and they think they know more than others. Let me be clear, investors are like those professors and are prepared. Your idea may be running or working now as you think it is, but it may be just reorganizing data from established ideas. I personally use methods that have already found and work on most of my projects. I never oversell and state that my AI frameworks will be better than a human. Anyone who states that humans are going to be replaced by coming AGI does not know their elbow from their nose. AGI in the knowledge space can only simulate and cannot feel and if it’s an LLM, it’s symmetrical in nature. Adaption cannot be achieved in a quick and easy way, nor is it going to happen with such approaches.

Merely, the facts are this, AGI is a level of computing that a program runs. It is NOT alive nor is it doing anything other than simulations. Furthermore, AGI is a statement of what achievements it may be able to achieve, not the outcome of a full human mind. We must fight and rage against such ideas as they are false and create an illusion that is not true.

You may have noticed that many AI companies that have closed models are not really improving in a traditional sense. They have additions, they roll out new products, but are they really improving past what is considered AGI? I would state they will hit a wall. Many others, who are like me, consider Dr. LeCun who left the LLM world for a whole new approach. He was correct and that is the right move. More data is needed than just a written word. AI can not hear you, it only processes the information (sounds you make when you speak) and transcribes them into words to process. It then makes sounds (talking) based on the transcription it gives to another model. I can hear you, but AI needs transcription, that’s a huge difference.

Asymmetrical approaches towards AI are the route I have taken. From the World Model to some custom approaches that I have come up with (some still work in progress) I think some of the answers are there; however, the idea that the entire idea is in one model and that is all you should stick with for everything is a farce.

Would you use a mechanic to make your coffee? Want to get fed by a construction worker who never worked in a restaurant? I think not. It’s all about domains and specific AI as no one size fits all.