Member-only story
Unleashing the Power of Large Language Models: Exploring their Limits
Yesterday, I listened to specialists discussing the impact and limits of large language models. As I was listening, I understood that at least currently and probably for years to come, large language models will only be able to summarize existing information, and know-how statistically and not generate content if this information base does not exist. They work fast, but conduct basic tasks.
Let us take a recent example. A tunnel on the road to a small village in Switzerland has just collapsed. To report on the incident, we will need to delegate a reporter on-site who will interview whoever is required to explain the reasons for the collapse or the impact it will have on the inhabitants’ lives.
As the event just took place, Chatgpt could not know about it. In that sense, I tend to agree with R. Kurzweil who mentions in one of his last speeches that AI is an enhancement for our brain, very similar to a mobile phone. https://www.youtube.com/watch?v=Iu7zOOofcdg&t=1745s
What is at risk are all the activities that provided this summarization of information, may it be in law, medicine, architecture, journalism, biology,…
What is at risk are the entry-level or lower-level white-collar jobs. The question that our society has to ask is how do you allow…