Artificial Intelligence Can Think Like Humans: Scientists Find First Evidence
Favicon 
anomalien.com

Artificial Intelligence Can Think Like Humans: Scientists Find First Evidence

A study of large language models (LLM) has shown that popular AI systems from OpenAI and Google organize information on their own and in unexpected ways, even though they were not trained to do so, The Independent reports. The authors of the paper, experts from the Chinese Academy of Sciences and the South China University of Technology, concluded that such models “have deep similarities with how humans construct concepts.” This challenges the common belief that artificial intelligence only reproduces answers by recognizing patterns. In the experiment, OpenAI’s ChatGPT-3.5 and Google’s Gemini Pro Vision were asked to perform a task to find the odd object out. As a result, the models created 66 conceptual dimensions for classifying objects. “These dimensions could be understood and described, indicating that large language models and multimodal LLMs form human-like representations of objects,” the researchers noted. “This provides a strong argument that although these representations are not identical to human ones, they share common foundations and reflect key features of human knowledge of the world.” The authors hope that these results will help create artificial intelligence with more “human” cognitive abilities – one that can collaborate better with people. At the same time, systems even closer to human nature are already being developed around the world. For example, a startup from Australia recently presented the world’s first commercial biological computer based on living brain cells. The Cortical Labs system is a “body in a box”: it is based on lab-grown neurons located on a silicon wafer that can receive and transmit electrical signals. According to the developers, this combination of biology and electronics allows them to learn and adapt much faster than conventional computers. Even at an early stage, 800 thousand neurons in humans and mice were able to independently master the game Pong. As reported in a paper in the journal Cell, the neurons showed signs of consciousness when interacting with the game environment. “We combine biology with classical computing to create a universal learning engine,” the company’s website says. “The neuron is self-tuning, has nearly infinite flexibility, and is the product of four billion years of evolution.” The study detailing the research, titled ‘Human-like object concept representations emerge naturally in multimodal large language models’, appears in the scientific journal Nature Machine Intelligence. The post Artificial Intelligence Can Think Like Humans: Scientists Find First Evidence appeared first on Anomalien.com.