LLM Capabilities Doubling Every Seven Months: A 2030 Prediction

New research reveals a startling rate of progress in large language models (LLMs). Their ability to complete complex tasks is doubling roughly every seven months, according to a metric called "task-completion time horizon." This metric compares the time an LLM takes to complete a task to the time a human would take. The study projects that by 2030, the most advanced LLMs could complete, with 50% reliability, a software task equivalent to a month's worth of human work (40 hours/week). This raises significant concerns and excitement about the potential benefits and risks of LLMs, while acknowledging that hardware and robotics could potentially limit the pace of progress.