Do new ai tools like ChatGPT really understand language the same way humans do?
It turns out that even the inventors of these great new language models are debating that very question, and the answer will have enormous implications for education and for all aspects of society if this technology can get to a point where it achieves what it aims for. known as Artificial. General Intelligence or AGI.
A new book from one of those ai pioneers delves into the origins of ChatGPT and the intersection of research into how the brain works and building broad new language models for ai. Is called “ai/” target=”_blank” rel=”noopener nofollow”>ChatGPT and the future of ai,”And the author is Terrence Sejnowski, a professor of biology at the University of California, San Diego, where he co-directs the Neural Computing Institute and the NSF Center for Temporal Dynamics of Learning. He is also the Francis Crick Professor at the Salk Institute for Biological Studies.
Get free EdSurge journalism delivered to your inbox. Subscribe to our newsletters.
Sejnowski started out as a physicist working on the origins of black holes, but early in his career he says he realized it would be decades before new instruments could be built that could adequately measure the types of gravitational waves he was studying. So he turned to neuroscience, hoping to “open the hood” on the human brain to better understand how it works.
“It seemed to me that the brain was as mysterious as the cosmos,” he tells EdSurge. “And the advantage is that you can do experiments in your own laboratory and you don't have to have a satellite.”
For decades, Sejnowski has focused on applying findings from brain science to building computer models, sometimes working closely with the two researchers who just won the Nobel Prize this year for their work in ai, John Hopfield. and Geoffrey Hinton.
Today, computing power and algorithms have advanced to the level where neuroscience and ai help inform each other, and even challenge our traditional understanding of what it means to think, he says.
“What has really been revealed 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 something. “We can record from neurons, but that doesn't really tell you how it works and what's really happening when you're thinking.”
He says new chatbots have the potential to revolutionize learning if they can deliver on the promise of being personal tutors for students. A drawback of the current approach, he says, is that LLMs focus on only one aspect of how the human brain organizes information, while “there are hundreds of parts of the brain that are left out that are important for survival, autonomy for to be able to maintain activity and consciousness.” And those other parts of what makes us human may also need to be simulated for something like mentoring to be more effective, he suggests.
The researcher warns that there are likely to be unintended negative consequences for ChatGPT and other technologies, just as social media led to a rise in misinformation and other challenges. He says regulation will be necessary, but “we won't really know what to regulate until it's actually available and being used and we see what the impact is and how it's used.”
But he predicts that soon most of us will no longer use keyboards to interact with computers, but will instead use voice commands to dialogue with all kinds of devices in our lives. “You'll be able to get in your car, talk to him and say, 'How are you feeling today?' (and might say:) 'Well, we're running out of gas.' Oh okay, where is the nearest gas station? Here, let me take you there.”
Listen to our conversation with Sejnowski on this week's EdSurge podcast, where he describes research to more fully simulate human brains. He also talks about his previous project in education, a free online course he co-teaches called “learning to learn,”, which is one of the most popular courses ever created, with over 4 million students enrolled in the last 10 years.