The Future of AI - A Debate Between Jeff Hawkins and Ben Goertzel

Notes

The Emergence of Intelligence from the Neocortex
Key takeaways:
(* The neo cortex is the outermost and most recently evolved layer of the mammalian brain., * It occupies about 70 % of the volume of your brain., * It's responsible for all aspects of your intelligence, that is to say, your sense of vision, touch, hearing, language and all its forms, and even abstract thinking such as mathematics and philosophy., * The complexity of the brain is in the content of its connections and its wiring, which is an emergent property of a simple learning algorithm.)
Transcript:
Speaker 2
Hawkins says that the neo cortex is the outermost and most recently evolved layer of the mammalian brain, a bit like a wrinkly napkin which wraps around the old brain. It occupies about 70 % of the volume of your brain, and it's responsible for all aspects of your intelligence, that is to say, your sense of vision, touch, hearing, language and all its forms, and even abstract thinking such as mathematics and philosophy. The neo cortex is surprisingly different to other parts of the brain. It has no visually obvious divisions. The anatomical organization is strikingly similar. Nevertheless, different parts of the neo cortex still perform different functions like vision or hearing. For example, the complex circuitary of the neo cortex looks remarkably alike in visual regions and language regions and even touch regions. It looks similar across species such as rats, cats and humans. And the variations between regions are relatively small compared to their similarities. Now, hawkins argues that the complexity of the brain is in the content of its connections and its wiring, which is an emergent property of a simple learning algaritm onwe look at the open ai microscope, their tool for visualizing trained convolutional neural networks, as an example. The learning algarithm is simple, but all this beautiful complexity emerges as a result of training. Even after this type of visualization, it's not really understandable by us humans. It's possible that any successful knowledge representation or substrat for some domains of data may never be understandable by humans. Now, the neo cortex is arguably one of the main organs of intelligence.

The Future of AI: A Debate Between Jeff Hawkins and Ben Goertzel
Key takeaways:
(* Jeff Hawkins believes that artificial general intelligence will be a hybrid of many underlying algorithms, not a single learning algorithm., * Ben Goertzel believes that artificial general intelligence will be a single learning algorithm.)
Transcript:
Speaker 2
He knew the biggest prize was understanding human intelligence and then using that knowledge to create human level machine intelligence. So in two thousand and five he cofounded numenta in redwood city in california. Nementa is a machine intelligence company that has developed a cohesive theory, coarse soffwar technology and applications based on the principles of the neo cortex. Its dual mission is to understand how the brain wor s and to apply those principles of real intelligence to create intelligent machines. Neuro scientists were publishing thousands of papers a year covering every single detail of the brain, but there was a lack of systemic theories that tied all those details together. Nementa decided to first focus on understanding a single cortical column. They knew that cortical columns were doing something physically ex and therefore must be doing something complex. Now last week we had ben gertzel on the show, and he was convinced that artificial general intelligence must be a hybrid of many underlying algarisms, not a single learning algarism. Jeff hawkins doesn't agree. Hawkins thinks that all the magic of intelligence could emerge from a single cortigal learning algaritm andrew n g

The Mindful Brain
Key takeaways:
(* The mindful brain is a book about the brain written by two prominent scientists in 1978., * The neo cortex got big by making copies of the same basic thing, the same circuit.)
Transcript:
Speaker 2
The mindful brain a small book. It's about a hundred pages long and published in 19 78, and it contains two essays about the brain from two prominent scientists. One was written by vernan malcastle, a neuro scientist at john hopkins university. Now jeff hawkins cites mountcastle as being one of his biggest inspirations. Jefs says that it remains one of the most iconic and important essays ever about the brain. Mountcastle proposed a new way of thinking about the brain that is elegance, a hall mark of great theories, but it's also kind of surprising, and it continues to polarize the neuro science community. Now, mountcastle noted that the brain grew really large by adding new brain parts on top of old brain parts. The older parts control more primitive behaviorsw the newer parts create more sophisticated ones. However, mountcastle goes on to say that while much of the brain got bigger by adding new parts on top of old parts, that's not how the neo cortex grew to occupy 70 % of our brain. The neo cortex got big by making copies of the same basic thing, the same circuit. He says that every single part of the neo cortex is the same basic it. Different parts of the neo cortex are different, not in their intrinsic function, but rather in what they are connected to.

The Importance of Voting Connections in the Cortex
Key takeaways:
(* Our perception is composed of many different things, including sounds, vision, and touch., * Our models of the world are based on visual models., * Our cortex has voting connections that allow different parts of the brain to reach a consensus.)
Transcript:
Speaker 1
We have a singular perception it. You know, we think, oh, i'm just here, i'm looking at you. But it's, it's composed of all these things. There's sounds and the and theres a this vision, and theres touch and all kinds of imports. Yet we have the singular perception in what the thousand brain series says, we have these models that are visual models. We have bought models, har orditory models, models, tokdo models and so on. But they vote. And so a, the in the cortex, you think about these columnss like little grains of rice, a hundred, 50 thousand stacke, next ut. And each one is its own little modelling system. But they have these long range connections to go betweene and we call those voting connections. Are voting norans. Ah. And so thethe different columns out try to reach the consensus, like, what am i looking at?

up:: 📥 Sources