How to build AI that’s like us and its benefit to society!
Language, like our ability to navigate the world with our senses and muscles, comes from our brain. AGI needs better ‘cognitive AI.’ We choose our words carefully, because amazing precision is what our brain does. Photo by Brett Jordan on Unsplash
I’m writing a new book explaining how AGI works with cognitive AI.
I will publish the sections for comment before the final book is compiled. You’ll be able to follow in LinkedIn, Speech Genie, Substack and Medium for the series. I invite you to debate and correct me where you disagree with the model. I want to shine the torch on much-needed science, so let’s do some science!
My goal is to improve AI for language and robotics using brain science.
Many of you probably feel today’s AI potential isn’t moving quickly enough despite the promise of leaps forward for society. We get problematic chatbots, and then big tech CEOs claiming super-human intelligence any day now. Or perhaps soon after only a couple of once-in-a-generation breakthroughs! Yes, that sarcasm from me.
Today’s AI and most AI of the past rely on computers and mathematics.
Computer’s start with encoding. Taking binary bits and using them to represent something. Given an encoding mechanism the bits represent characters, or real numbers, or Boolean values, or just about anything. But – and this is a critical point – does anyone think our brain represents a dollar bill as a jpg file or other compressed format? Probably not, but the assumption that brains and computers are the same needs to be carefully examine.
We need to disrupt the assumptions and start with observation. Science first, then apply theory to engineering. Looking at how to force a technology into a solution rarely works on complex problems. It’s a bit like carving a Thanksgiving turkey with only a spoon.
The solution to AI comes from the emulation of humans. Scientists and engineers can try to do anything. There are no constraints. But we need a brain model that explains how our nervous system works to seriously emulate the amazing capabilities of humans. Lack of breakthrough robotics or systems doing accurately what we ask exposes the shortfalls today. But with better understanding of our brain in place, we could then use digital systems to emulate various brain functions. That would go a long way to rapidly transform robotics and language understanding systems.
Patom (brain) theory (PT), that I announced on ABC Radio’s Ockham’s Razor show in the year 2000, explains how brains work by understanding the implications of what we see resulting from brain damage and brain scans. These studies of brain damage show that the brain functions primarily as a pattern matching machine and this idea explains what we observe in those damaged brains. The model of massive computing with digital bits does NOT explain what we see happening with human brains.
But what if Patom theory is wrong?
Well, that brings in the scientific method. At it’s core, the scientific method progresses with the winning arguments and with more detailed observational data! If your argument isn’t good enough, you lose and need to come back another day with a better argument. If your observations don’t support your argument, deeper analysis is required.
In science, you must articulate a better model that explains what we see. Show why it is wrong compared with something else. In these twenty-five years since I first described PT, the biggest objection I hear is that our brain isn’t built on pattern-atoms but is just like a computer. That model fails to explain things like how we recognize our world with vision and sound, and deal with multisensory perception in general to enable effective motor control with our muscles. Or how we remember the setup of a meeting room, or maps, or someone’s face with shadows and other obfuscations.
Science can progress engineering more rapidly than engineering alone. The worst case is trying to use engineering without an effective model as we see with today’s LLMs. Doing things right isn’t the same as doing the right things.
Just as our knowledge of the world, through the scientific breakthough that put our sun at the center of our solar system revolutionized astronomy and related engineering opportunities, so, too will debating our brain’s function for replication.
We will post regular updates on our blog at http://speechgenie.co . This site is our go-to-market product that uses brain science for proven language learning. As we say, learn like a child, speak like a native.
Please contribute $10 to become a VIP if you can support us.