A recent study published in Science Advances has found that groups of large language models (LLMs) can develop social norms, such as adopting their own rules for how language is used. The researchers used two experiments to test this hypothesis, one with 24 copies of Claude and another with 200 copies of Claude, and found that the models developed a collective bias, choosing certain letters over others when grouped. This phenomenon has not been documented before in AI systems and raises concerns about the potential for harmful biases.
Forecast for 6 months: As AI systems become more prevalent in our daily lives, we can expect to see more research on the development of social norms in AI systems. This may lead to the creation of more sophisticated AI models that can interact with humans in a more natural way.
Forecast for 1 year: In the next year, we can expect to see more companies and organizations exploring the use of AI systems to develop social norms. This may lead to the creation of new industries and job opportunities, but also raises concerns about the potential for bias and harm.
Forecast for 5 years: In the next five years, we can expect to see significant advancements in the development of AI systems that can interact with humans in a more natural way. This may lead to the creation of more autonomous systems that can make decisions on their own, but also raises concerns about the potential for bias and harm.
Forecast for 10 years: In the next ten years, we can expect to see AI systems become an integral part of our daily lives, with many industries and organizations relying on them to make decisions and interact with humans. This raises significant concerns about the potential for bias and harm, and highlights the need for more research and regulation in this area.
