laitimes

ChatGPT actually lies! AI has a long way to go to replace the human brain, and now there are many problems and loopholes

author:Zhou Xian Hong Kong

Text | Zhou Xian

ChatGPT actually lies! AI has a long way to go to replace the human brain, and now there are many problems and loopholes

As far as ChatGPT is concerned, it emphasizes creation, and the answers are different from time to time, which is very useful in work such as writing cover letters. (Profile picture)

In an August 2019 case in which Roberto Mata, a man flying from El Salvador to New York, was hit in the knee by a flight attendant's cart and injured himself, accusing Colombia's national carrier Avianca of personal injury, Colombia's national carrier Avianca was charged with personal injury, while Colombia argued that the statute of limitations had expired and asked for the case to be dropped.

A few days ago, Steven A. Schwartz, a lawyer in New York, filed a 10-page litigation document on this case, but the judge found that the six precedents cited in the document "did not exist at all."

He had to admit that he used ChatGPT to study precedents, and this was the first time he used ChatGPT to prepare court materials, and "did not realize that the content may be false", because he had repeatedly asked whether these "non-existent cases" were true, and ChatGPT kept giving affirmative responses, even claiming to be found in credible legal databases such as "Westlaw" and "LexisNexis".

ChatGPT actually lies! AI has a long way to go to replace the human brain, and now there are many problems and loopholes

He and his colleagues are charged with ethical violations and face disciplinary action. Since similar incidents are unheard of, judges also hold hearings to discuss what punishment should be imposed.

The lawyer, who has been practicing for 30 years, could not have imagined that the old cat would burn his beard, because he never dreamed that ChatGPT could actually lie, and deceived so thoroughly, there is no doubt that this incident not only caused him great harm, but also lost confidence in the recent popularity of ChatGPT and artificial intelligence.

When ChatGPT first launched, I installed it, but I didn't think it worked. To put it simply, I think it's too "watery" and too "low-energy" to help at all.

In 1999, I worked in Xiao Ruoyuan's "Skynet" and witnessed the rapid evolution of the Internet from a very simple situation to today's situation, and I believe that the development of artificial intelligence will be the same, but this cannot hide the fact that it is still very simple.

ChatGPT actually lies! AI has a long way to go to replace the human brain, and now there are many problems and loopholes

As far as ChatGPT is concerned, it emphasizes creation, and the answers are different from time to time, which is very useful in work such as writing cover letters. However, if you are asking simple questions such as "Who is Zhou Xian", the answer will be very annoying every time, and if you are asking technical questions such as law and medicine, if there are different answers, it will be even more confusing.

For example, "DoNotPay", founded in 2015, is known as "the world's first robot lawyer", but robots appear in court to fight lawsuits, should not be accepted by the court, if only write legal documents, in fact, and the use of archives to copy and change, is no different.

Copyright statement: This article is the author's original article, the picture material comes from the Internet, the text content of this article is strictly prohibited from illegal reproduction without authorization, if you need to reprint or quote, you must obtain the consent of the author and indicate the source.

Read on