Do new AI tools like ChatGPT actually understand language as well as humans?
It turns out that even the inventors of these new large-scale language models are debating that very question. And if this technology can reach the point of achieving what is known as an artificial language, the answer will have a profound impact on all aspects of education and society. General intelligence, or AGI.
A new book by one of these AI pioneers delves into the origins of ChatGPT and the intersection of research into how the brain works and building new large-scale language models for AI. It is called “The future of ChatGPT and AI,The author is Terrence Sejnowski, a professor of biology at the University of California, San Diego, where he is co-director of the Institute for Neural Computation and the NSF Center for Temporal Dynamics Learning. He is also the Francis Crick Professor at the Salk Institute for Biological Studies.
Get EdSurge’s journalism delivered free to your inbox. Sign up for our newsletter.
Seinowski started out as a physicist studying the origins of black holes, but early in his career he realized that it would be decades before new instruments were built that could adequately measure the kind of gravitational waves he was studying. He said he realized that. So he switched to neuroscience, hoping to “pop the hood” of the human brain to better understand how it works.
“It seemed to me that the brain was as mysterious as the universe,” he told EdSurge. “The advantage is that you can do the experiments in your own lab and you don’t need to have a satellite.”
Sejnowski has been focused for decades on applying discoveries in brain science to building computer models, and he is working closely with John Hopfield and Jeffrey Hinton, who just won the Nobel Prize this year for their work on AI. I have worked closely with the two researchers from time to time.
In recent years, computing power and algorithms have advanced to the point where neuroscience and AI are mutually informing each other, even challenging traditional understandings of what thinking is, he says. Masu.
“What’s really become clear is that we don’t understand what ‘understanding’ is,” Sejnowski says. “We use the word and think we understand what it means, but we don’t know how the brain understands it. We can record from neurons, but we don’t know how it works. You don’t actually know what’s actually going on when you’re thinking about whether it’s going to work.”
If new chatbots can fulfill their promise as personal tutors for students, they have the potential to revolutionize learning, he says. One shortcoming of current approaches is that LLM focuses on only one aspect of how the human brain organizes information, whereas LLM focuses on only one aspect of how the human brain organizes information, whereas 100 parts of the brain are being left out,” he says. Activity and consciousness. ‘And for things like tutoring to be most effective, he suggests, other parts of what makes us human may need to be simulated as well.
Researchers warn that just as social media has led to an increase in misinformation and other challenges, ChatGPT and other technologies are likely to have unintended negative effects. He said regulation is necessary, but “we don’t really know what to regulate until we actually see it out in the world and used and what the impact is and how it’s used.” ” he said.
But he predicts that soon most of us will no longer be using keyboards to operate our computers, but instead will be using voice commands to interact with all kinds of devices in our lives. Masu. “You’ll be able to get in your car and talk to your car and ask, ‘How are you feeling today?’ [and it might say,] “Well, we’re running low on gas.” Oh, okay, where’s the closest gas station? Here, let me take you there. ”
Listen to our conversation with Sejnowski on this week’s EdSurge podcast. There he describes his research to more fully simulate the human brain. He also talks about his previous projects in education, a free online course he co-teaches.learn how to learn,” is the most popular course ever created, with more than 4 million students enrolled in the past 10 years.
