In this interview written for Jiemian News, our co-founder Jeff Hawkins shares his insights on ChatGPT, the brain-based approach to building intelligent AI systems, and the future of humanity in an increasingly AI-integrated world.
The Q&A was originally conducted in English and later translated into Chinese for Jiemian News’ readership. Recognizing the global interest in these topics, we’ve decided to publish the original English version here.
Jiemian News (JN): There has been a global buzz surrounding ChatGPT. What are your thoughts on these recent advances in AI?
Jeff: The recent advances in AI are impressive, but they seem impressive only because so little progress was made in AI for many years. When computers were first invented, people were amazed by their speed and accuracy in searching through data and doing mathematics. Over time we got used to computers, and we no longer marvel at their abilities. However, some problems, such as image recognition and language processing proved difficult for computers, so we came to believe that these problems might never be solved by a computer.
The recent advances in AI are not based on fundamentally new algorithms. Progress is mostly because we can now train AI systems on very large datasets. The largest dataset we have is all the text on the internet. ChatGPT and other language AI systems work well because they can be trained on billions of language samples. Over the coming years, language-based AI will be integrated into lives as just another application of computers. We will feel as comfortable using them as we do when using a calculator or a smartphone.
JN: ChatGPT’s human-like language capabilities have been impressive. Would you consider this an example of artificial general intelligence (AGI)?
Jeff: 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.
JN: The term “emergence” often comes up when discussing ChatGPT. Do you think ChatGPT is capable of generating higher intelligence?
Jeff: Without being able to move and interact with the world, ChatGPT will forever be limited in what it can do. Chatbots will get better over time mimicking human language, but they will remain limited in what they understand. To create intelligent machines that understand the way we do requires copying some of the principles that the brain uses.
The Brain and Artificial General Intelligence
JN: The two primary avenues of AI development are developing general-purpose and specialized AI systems. With many companies currently engaged in deep learning and foundation models, your theory stands out as it is clearly dedicated to building general-purpose systems. Why is this the case?
Jeff: Starting at a young age I wanted to know how my brain worked. I couldn’t think of any question that was more interesting or more important, so I dedicated my life to that pursuit. I also believe that to build truly intelligent machines, we need to start with an understanding of the brain. Today there is a lot of confusion about what intelligence is. The progress we have made in understanding the brain, the Thousand Brains Theory, provides a clear understanding of what intelligence is and how we can build intelligent machines.
This doesn’t mean I am against more limited forms of AI, such as ChatGPT. In fact, my company, Numenta, has created technology that greatly lowers the cost of running language models such as GPT. We are excited by how we can not only make these models less expensive, but also greatly reduce the energy required to run them. But ultimately, we will build machines that are intelligent in the same way we are, and those machines will work similarly to the brain.
JN: You’ve previously stated that knowledge representation is the key to creating AGI. Could you elaborate on that?
Jeff: All AI systems learn by observing some part of the world and storing what they learn. How knowledge is stored inside an AI system is critically important to what the AI system can do. Humans store knowledge in a way that reflects the three-dimensional structure of the world. We learn by moving our bodies, eyes, and limbs. As we move through space, our brain stores what it learns in reference frames that reflect the physical structure of the world and where we were when we learned something. As I explain in my book, A Thousand Brains, the brain’s reference frames explain how we can rapidly learn almost anything and why we are able to solve problems in novel ways.
JN: What is the meaning behind the title of your new book, “A Thousand Brains”?
Jeff: The title, A Thousand Brains, refers to a surprising attribute of our brains. When we learn something, such as what a coffee cup feels and looks like, we actually learn many models of the cup, not just one. Your brain is a distributed modeling system. It contains hundreds to thousands of models of everything you know. We didn’t anticipate this when we started studying the brain, but it is an inevitable conclusion. I explain this in my book.
JN: In your book, you suggested that the world we perceive may not be the true reality, but one created by our brains from constant streams of sensory inputs. Does this mean we can understand high-level concepts like religion and philosophy from a biological and neurological standpoint?
Jeff: Most people who study the brain, including me, believe that everything we experience, including consciousness, is caused by activity in the brain. There are no magical forces that make cells do things that can’t be explained by biology. So, by studying the brain we can understand how the brain and its cells create religion, philosophy, and our beliefs. I cover all these topics in my book. Some people don’t want to believe that we can be reduced to biology. Personally, I don’t see any problem with this. Humans and human experiences are just as fascinating regardless of whether we can understand them as biological processes.
AI and the Future of Humanity
JN: We have been seeing a surge in AI research across various industries. Do you think this will lead to a new technological revolution?
Jeff: AI will transform the 21st century similarly to how digital computing transformed the past century. The changes that AI will bring will impact all aspects of our lives. We can’t know exactly how this will play out, just as the original architects of computers couldn’t anticipate GPS, smartphones, or the internet. But we can be confident that the world will appear very different in fifty years.
JN: What are the risks associated with AI? Could it pose a threat to humanity?
Jeff: The changes brought about by AI will be largely positive. However, there will be negative consequences, too. Some of the obvious challenges are using AI to spread misinformation and using AI in warfare. I don’t see any direct existential risk of AI. AI will not turn on us and destroy humanity, I explain why in my book.
JN: What industries do you think AI will likely replace first? Conversely, what new industries are likely to emerge?
Jeff: All technological revolutions change the way we live and work. These changes are mostly for the better. Of course, some industries and jobs will be negatively impacted, but more new industries and new jobs will be created. We have seen this trend play out over and over again. We must help the people who lose jobs due to AI, but more importantly, entire new industries and intellectually interesting jobs will be created by AI. I am excited to be part of this future.