AI (artificial intelligence) is much in the news these days. As someone who has worked in technical areas (mathematics and computer science), I have naturally had some thoughts on the topic.
In a blog post that I read today and titled “AGI Has Not Arrived, and No Tech Company Wants It To“, James McGrath described how he sees today’s AI based on LLM (large language model) methodology. Here, “AGI” stands for “artificial general intelligence”.
McGrath’s background is as a professor of religion. That’s very different from my background as a (retired) professor of computer science. Yet we seem to agree on our views of AI. I would suggest that you read the post by McGrath.
Around 1990 (give or take a few years), a guest lecturer in our department at Northern Illinois University gave a seminar on AI. As I recall, he said that up to now computers had been used to solve clerical problems. But, he suggested, they were now being used to solve intellectual problems.
As I thought about that, it seemed to me that they were only solving the clerical aspects of intellectual problems. Looking at it today, with the LLM methodology, it still seems to me that they are only handling clerical aspects of intellectual work.
To be clear, the LLMs are a huge advance from what we had circa 1990. At that time, people were looking to find ways of using computers to do natural language processing. And it seemed to be a difficult problem. And now the LLMs at least have the appearance of being successful at with natural language problems. They are not completely successful, in that the sometimes make up stories (often described as hallucinations), but the seem to be pretty good at conversation level language use.
Continue reading