ChatGPT is just the beginning: How advanced AI is set to enter a new era
Artificial intelligence is entering a second wave of development as signalled by the likes of ChatGPT, an expert in the field explained Thursday.
Developed by OpenAI, ChatGPT stunned observers in late 2022 with its ability to produce answers to abstract and complex questions. Speaking at the 2023 Consumer Electronics Show in Las Vegas on Thursday, Zico Kolter, chief scientist of AI at Bosch, explained how the system's ability to respond to input from a variety of sources move away from application-specific AI systems and toward a more general system that works in a variety of cases.
The breakthroughs were perhaps best demonstrated by Don Burnette, CEO of Kodiak Robotics and a fellow panelist. Burnette explained how businesses can better leverage AI than ever before, and the general public's understanding of the technology has improved.
"I actually asked GPT to answer that question," Burnette revealed. "That's the answer it gave me."
A neat party trick, but Kolter explained that it's indicative of a broader shift. Kolter has worked in the industry for around 20 years, and it's part of a second wave in emergent AI technology. The first was around 10 years ago, with the rise of deep learning. IBM explains that deep learning aims to mimic the human brain, with similar matching abilities.
The second shift, as highlighted by the likes of ChatGPT, is a shift away from task-specific AI models to more general purpose models. ChatGPT, he explained, is an example of how a more general AI can interpret language to come to new conclusions — the next step will be to apply more general purpose designs to other areas.
"I think the next big challenge for AI is how we are going to develop these general purpose models," Kolter said at CES 2023.
Try talking with ChatGPT, our new AI system which is optimized for dialogue. Your feedback will help us improve it. https://t.co/sHDm57g3Kr— OpenAI (@OpenAI) November 30, 2022
ChatGPT: What comes next
The comments hint at something that's held up as the holy grail of artificial intelligence: an AGI, or artificial general intelligence. This is a system so general that it's able to learn almost any task in a similar fashion to a human. Tech entrepreneur Elon Musk has sounded the alarm on such a development, as he has warned that such superintelligences could be "the single biggest existential crisis that we face."
Kolter isn't describing something quite as ambitious as this. Instead, he sees the next step in development as ChatGPT-like AIs that can take input and solve new and unexpected tasks. Instead of language, this could be inputs from sensors like radar or lidar.
"In a similar way of ChatGPT, let's just take the input from these sensors [and solve new problems] without having to retrain the model," he explained.
ChatGPT may be answering homework questions and general queries today, but it could lead to a future of even more ambitious systems.
IE talked with Shohini Ghose about how quantum computers might transform our future, the mysteries of quantum mechanics, and what the quantum scene will look like in 2027.