A law school in the US gave exams to a chatbot which received information from several online sources. It wrote papers about the Constitution, taxes, tort law, and nearly everything else.

A chatbot energised by tonnes of internet data was able to pass tests at a U.S. law school. It did this by writing essays on topics like constitutional law, taxes, and torts.

To use artificial intelligence (AI), ChatGPT from OpenAI, a U.S. company that got a big funding boost from Microsoft last week, turns simple suggestions into streams of text.

Due to the good results, teachers have warned that they could lead to a lot of cheating or even mean the end of traditional ways of teaching in the classroom.

The same 95 multiple-choice questions and 12 article questions were given to ChatGPT by Professor Jonathan Choi of the law school at Minnesota University.

ChatGPT Explainer is an AI tool for writing

On a gadget, you can see a ChatGPT prompt. Peter Morgan of the AP says

In a white paper called 'ChatGPT goes to law school,' which came out on Monday, he and his coauthors said that the bot got an overall grade of C+.

Even though this was good enough to pass, the bot always got scores in the bottom half of the lesson and 'bombed' multiple-choice math tests.

The authors said that ChatGPT 'had a good understanding of basic legal rules when writing essays and always used good structure and composition.'

However, when provided an open-ended task, the bot 'often had trouble finding flaws, which is a basic skill on law school exams.'

In New York as well as other locations, it is against the law for schools to use ChatGPT, but Choi said it would be a good way to teach.

On Twitter, he said that ChatGPT 'wasn't a great law student working alone.'

But we think that language models like ChatGPT will be extremely helpful to both lawyers in practice and law students taking exams when they work with people.

In responding towards another Twitter user, he also said that two out of three graders had noticed that the document had been written by a bot, which made it less likely that someone had cheated.

Choi says the team had a sneaky suspicion, and it proved to be correct because ChatGPT used perfect English and was very repetitive.