If you have any interest at all in technology, and possibly even if you do not, it is very likely that over the past few weeks you have read or heard something about OpenAI’s ChatGPT chatbot, which was released for public testing on 30 November 2022. OpenAI describes itself as ‘an AI research and deployment company’ with a mission ‘to ensure that artificial general intelligence benefits all of humanity.’ It counts Microsoft as a major sponsor, to the tune of a US$1 billion investment that gives OpenAI access to enormous computing resources, in exchange for which Microsoft gains privileged access to OpenAI’s breakthroughs. ChatGPT’s conversational and question-answering skills are – superficially, at least – impressive. Naturally, this has led some breathless commentators to suggest that we are on the verge of seeing various professional jobs, including those of lawyers, being replaced by AI machines. But is ChatGPT really that good, or are people merely being blinded to its limitations by the fact that it is undeniably much better than anything they have seen before?
If language models like ChatGPT are going to replace professional advisors, such as lawyers and patent attorneys, then they will need to demonstrate the standards of competence and reliability (not to mention ethical conduct and responsibility) that the public expects from such advisors. As we shall see, my experiments with ChatGPT demonstrate that it is nowhere near achieving these standards. Indeed, there is a serious question as to whether it even represents a viable approach to developing AI with such capabilities. In any event, there are no signs that machines will be replacing professional advisors in the foreseeable future.