30s Summary
Researchers developed a computer model “Centaur” that can accurately simulate human behaviour using data from the Psych101 database, which contains over 160 psychological experiments. Traditional models only operate in specific areas but Centaur was further refined with Meta’s AI language model, “Llama 3.1 70B”, to predict behaviour in untested scenarios. The model can work in real time, adapt, and learn from its surroundings, lining up with the theory of cognitive scientist Alan Newell.
Full Article
Researchers from several top universities have come up with a computer model that they claim can pretty accurately predict and simulate how humans behave. The model nicknamed “Centaur” could be really useful for saving time in scientific research and experiments.
The group of researchers trained Centaur using a data goldmine known as Psych101. This database has a ton of information about human behavior from over 160 psychological experiments involving over 60,000 individuals who made over 10 million choices on something.
Before now, no computer model could authentically capture what the human mind is like. The human mind is unique – it handles everyday decisions about breakfast choices and clothing selection, as well as complex problems like finding a cure for cancer or exploring outer space.
However, most modern computer models only work in specific areas and are programmed to solve certain problems. According to Marcel Binz, one of the researchers, they refined Centaur using Meta’s AI language model, Llama 3.1 70B, which honed its ability to predict how people behave even in untested circumstances.
Cognitive models are basically ways to simulate how human and animal brains work – things like perception, reasoning, and memory. Binz says that after refining Centaur, it was more in tune with how human brains work, despite the fact that it wasn’t specifically trained to do that.
Centaur can work in real time, adapt logically, and learn from its environment. The team’s excitement over it being a unified model of human cognitive processes was shared by Binz who said it’s exactly what cognitive scientist Alan Newell had once envisioned.
Alan Newell was a computer science and cognitive psychology researcher who worked for the RAND Corporation.
Source: Cointelegraph