Next week, governments, AI companies and experts in research will come together at Bletchley Park in the UK for the AI Safety Summit - a series of talks about the safe development and use of frontier AI technology.
Frontier AI is a recent term that describes highly capable AI models which can perform a wide variety of tasks and get close to human-level intelligence. Because these frontier models could evolve from today's large language models, some reasearchers have even claimed that frontier AI could match - or even surpass - human intelligence in the next 3-5 years.
However, there's one challenge that remains unsolved with large language models - they lack common sense. I explain more about what that means and why we might not reach human-level intelligence in the near future.