Excerpt from Heidi Cohen
While it's easy to forget that ChatGPT is still in beta because the results appear so good, as the saying goes, 'Buyer Beware!' - Why? Due to ChatGPT's structure and data training set, it has some critical issues.
ChatGPT’s problems include:
- Output accuracy can be unreliable. Since ChatGPT is based on major internet websites such as Wikipedia, it can repeat existing misinformation and false facts contained in the original content.
- Key concepts may be lost or misinterpreted during the generative process changing the meaning of the output. Or, worse, it can just make things up which AI researchers refer to as hallucinations.
- Events and information that appeared or occurred after its initial training set was last tested aren’t referenced. According to the OpenAI website disclaimer: Limited knowledge of world and events after 2021. Before using any ChatGPT results, you should check the facts carefully.
- Bias contained in the underlying information creates a critical problem when using ChatGPT. The Internet contains information in all content formats that reflect societal attitudes and language that is no longer acceptable in general discourse.
- Facilitates plagiarism, other forms of theft and cheating. To help with this widespread problem, tools are being created to detect whether or not a piece of content was written by a human.
- Lack of copyright for published content in any format including text, images and/or sounds. Given the fast growth of this new form of AI, expect new tools and laws to sort out this challenge.
Click here to read complete article at Heidi Cohen.