All large language models, such as ChatGPT, are renowned for “response hallucination,” generating a
Published December 27, 2022
All large language models, such as ChatGPT, are renowned for “response hallucination,” generating a response that seems very credible and coherent but is factually false.
Open AI has admitted that ChatGPT is susceptible to providing “plausible-sounding but incorrect or nonsensical answers.”
As you can see in this illustrative example, ChatGPT can generate what we can call “unintentional computational misinformation”.
Many (way too many) people will actually firmly believe that the response is 73 years.
While we should be concerned about the generation of blatantly false information, the real danger comes from people blindly relying on such information.
(picture source: Peter Yang)
#technology #gpt3 #chatgpt #openai #data #information #misinformation #bias #decision #ai #artificialintelligence #people Thomas Reuner HFS Research
Follow me if you are looking for an honest, pragmatic and funny perspective on technology, automation and security https://lnkd.in/gm253xN
Originally posted on LinkedIn · 813 likes · 148 comments
