Inside AI

Let's use science to uncover how brains work, so we can copy them!

AGI is around the corner……or is it?

Artificial General Intelligence (AGI) re-emerged as a talking point a few years ago propelled by variations on machine learning, leading to possibly the big question of 2024: ‘What if it is just around the corner’? Maybe it is already here, and hiding on the Internet?! Dream on. Dream on.

Even though we are encouraged by the media that it will emerge from statistical systems like large language models - no, instead, AGI will come from brain science, such as Patom Theory, not from the statistical manipulation of, and prediction from, lots of data.

Why brain theory? What is new?

If we emulate brains properly, we have AGI.

Patom theory is a model first published in the early 2000s. It is the theory that explains how brains work: using hierarchical, bidirectional linkset patterns only. Previously, scientific theory has been used to explain planetary motion. It also explained physics with relativity. But how do brains work? That’s our goal. Let’s sort through the theory to build AGI scientifically.

By the way, I don’t love that three letter acronym, AGI, but that’s a blog for another day.

A quick look online and you’ll see that brains send signals to other parts of the brain to understand the world. They are complex processing systems, aren’t they? We are ‘just starting to understand how brains work.’ Brains evolved as the organ of computation. And so on.

Recently, my Palo Alto-based company in Silicon Valley completed a range of tests using Natural Language. The tests show how easily brain theory solves many AI problems. We’d like to share our success with you as we move into the commercialization phase. And it has been my life’s goal to see this happen! Emulating the computer from Star Trek in the original series is just one step on our roadmap.

Building the AI community

Language use is the final frontier for our machines and is the last interface we will ever need. It’s the fastest way to communicate and it can be automated for any human language. So is this the beginning of AI or AGI? If so, we should be able to store knowledge centrally for lossless retrieval on demand via our native language, whichever that is. This comes with the expectation that by voice, we can drive our computers, devices, and machines. After that, it should be able to read endless data and identify the variety of opinions stored within.

So what’s the best way to improve AI? We need a community to iterate on and drive the potential of AI for society, starting with trustworthy language-based systems.

You are invited to subscribe and provide your views along this journey to help improve our thinking for the best outcome.

How often will we post?

The goal is to publish a new story at least once a week which will include our new science in the context of what’s currently happening in the world of AI.

Free subscribers will get the majority of news and many articles, while a paid subscription will also get some selected in-depth analysis. We are opening up comments for engagement to build our community, and most articles will be free.

Brains should not be magic

Brains perform several functions and we shouldn’t rely on the ‘we don’t know how it works’ model. At a micro level, we know that brain cells - neurons - activate when their inputs hit a threshold. At a macro level, we know that brain cells are organized into regions, with specialization related to the input received. By way of example, input from the eyes to a brain region cannot recognize auditory input and output to muscles initiates motion, not recognition of sensory input.

An AGI model should include the brain, its nerves, sensors and connected muscles, not just the neurons in our skull. Image: Adobe Stock.

Articles will include scientific references and engineered solutions to explain why pattern matching, in particular, is effective at solving AI problems today.

Join us next time

Next time, I’ll look at the fundamental reason brains match patterns, along with feedback that prefers a processing model. The power comes from how we can emulate the theory with pattern matching, but not processing, to lead us into AGI.