Maybe ten years ago (2016 or earlier) I read a book authored by Jeff Hawkins with Sandra Blakeslee titled On Intelligence (2004) that changed my life. This book was the product of Jeff’s life-long quest to understand how the human brain works, and his more recent quest to apply it to machine intelligence.
The book proposes a fundamentally different way of thinking about how the brain works which is centered on how the neocortex works. Jeff proposes that the basis of intelligence is in the “memory-prediction framework”. A concept he later calls the ‘cortical algorithm’, drawing on neuroscientist Vernon Mountcastle’s work.
I could go on about how our neocortices work, but its been ten years since I read the book so I’ll leave that for another post. But Jeff argues in his book that intelligence is basically the ability to recognize objects and store models of those objects in our brain using what’s called sparse-distributed representations, and then make predictions about these objects and how they’ll change through time using those models based on what’s been seen before. This could be through touch, taste, smell, sight or hearing.
Jeff recently did an interview with Jiemian News that has been posted in it’s original English version on the Numenta blog, which I highly recommend you read. In it he presents several very good arguments about why current deep learning technology is not the path to Artificial General Intelligence that some people think it is. Instead it is following the cortical algorithm given by our brains, that we will eventually be able to rival these marvelous biological constructions.
I’ve come to believe that all intelligence may be founded on the simply principle of prediction. There may be many forms of intelligence, ranging from the very simple (Venus Fly Trap), to the very complex (the human brain), but all of them, including classic computer programming, follow the principle of prediction. Humans do this at an individual scale when working on their own, when alone, or at a mass scale when cooperating on a project, such as working in a factory or on a farm.
When deciding if you want to go outside and take a walk for example, you have to make a prediction about whether it will rain or not. At a more complex (or perhaps philosophical) level you have to determine if it’s worth it to you. At every point along the way, you make a prediction about whether you will be successful, or whether you will have a peaceful relaxing walk, reaching your state of mind and body that is attractive to you.
I could talk about the subject of discipline, or doing something that hurts a little more now, but is good for you in the long run, but that would be distracting perhaps. I had to make a prediction about whether you would want to read this article after I had written it. Really, everything comes down to predictions with intelligence like that presented in the human brain.
Even ChatGPT has to make predictions, though Jeff argues that it does it in a fundamentally inferior way. To quote Jeff when asked if ChatGPT could be considered as achieving AGI:
No. It is easy to be fooled into thinking that chatbots such as ChatGPT are intelligent like we are, but they are not. Chatbots only know the statistics of text. Human intelligence is based on moving about in the world, touching things, feeling textures, seeing what happens when we interact with the world, and conversing with our fellow humans. A chatbot can fool you into thinking it knows these things too, but in reality, it can only play back text based on what humans have written. Chatbots don’t understand the world as we do.
Jeff is confident that the difference between chatbots that predict the next word and true intelligence is in the algorithm – and having a sufficiently good understanding of our brains’ architecture. I’m very hopeful that the technology at Numenta will be successful and actually implement true intelligence in binary ones and zeros, but I actually don’t believe consciousness and intelligence as the same thing.
If you read the book of Genesis in the Bible you will see that when God created man he breathed “the breath of life” into him and created a being born of his own life and made in his image. In Hebrew this is called the nishmat chayyim (nish-MAHT khai-YEEM) in Hebrew or the “breath of life”. God literally took some dust on the ground, took the constituent molecules and atoms and formed it into a man. then he knelt down and breathed his own life into his nostrils. This was the first man and where the first woman came from.
As such, I don’t believe machines will ever achieve consciousness and may even have a fundamentally different form of intelligence than humans because of their lack of a spirit or soul. Not necessarily in every way, but it may come up in surprising ways. For example, AI may never naturally feel a brotherly or even childlike mutual love for their human creators. They may not possess the same goals unless those goals are programmatically added ahead of time. I’m not saying AGI will be a danger to us, just that it might be very different.
But the main reason I am writing this article is simply to bring to light the ideas that have been brought forward by Jeff Hawkins in his books and by the team at Numenta. His work is much less known than the work by say, Open AI and Anthropic, but proposes a model of intelligence that is fundamentally different that what we have seen before. I hope others take his approach and try to learn from biology.
Leave a Reply