In an open letter released by the Future of Life Institute, executives and AI experts are urging the industry to train and develop more capable artificial intelligence (AI) models than OpenAI's GPT-4. The company's latest large language model is currently available in a limited capacity for ChatGPT Plus subscribers and will soon be integrated into Microsoft productivity and security products.
The letter argues that while GPT-4 is impressive, the industry needs to focus on developing models that can better understand human intent and context. According to the signatories, this will require greater investment in research as well as collaboration across different fields such as linguistics, psychology, neuroscience and computer science.
"OpenAI has done an incredible job with their language models but we believe there is still a long way to go," said Max Tegmark, President of the Future of Life Institute. "We need AI systems that can truly understand what people want rather than just generating text based on patterns they have learned from millions of examples."
The call for improved AI capabilities comes at a time when there is growing concern about the impact of automation on jobs and society more broadly. A recent survey conducted by KPMG Australia and the University of Queensland found that residents in India, China, South Africa and Brazil were less critical of continued implementation of AI systems compared to other possible use cases.
75 percent of Indians surveyed between September and October 2022 would place their trust in AI followed by 67 percent Chinese respondents and 57 percent South African respondents. However, experts warn against complacency given current limitations of these technologies.
"We need to be careful not to overestimate what these models can do or rely too heavily on them without considering potential risks," said Yoshua Bengio from Mila - Quebec Artificial Intelligence Institute who co-signed the letter with other AI experts.
The Future of Life Institute hopes to spur more investment in AI research and development in order to build more advanced systems that can better serve society.