Re: GPL on AI generated code

[Akira Urushibata]

A group of researchers led by Penn State Univ has conducted
experiments which verifies that Chat-GPT plagiarizes.  Note that they
used GPT-2, not GPT-3.

Beyond memorization: Text generators may plagiarize beyond 'copy and paste'
Penn State University
https://www.psu.edu/news/research/story/beyond-memorization-text-generators-may-plagiarize-beyond-copy-and-paste/

   ... Language models that generate text in response to user prompts
   plagiarize content in more ways than one, according to a Penn
   State-led research team that conducted the first study to directly
   examine the phenomenon.

   "Plagiarism comes in different flavors," said Dongwon Lee,
   professor of information sciences and technology at Penn State. "We
   wanted to see if language models not only copy and paste but resort
   to more sophisticated forms of plagiarism without realizing it."

   The researchers focused on identifying three forms of plagiarism:
   verbatim, or directly copying and pasting content; paraphrase, or
   rewording and restructuring content without citing the original
   source; and idea, or using the main idea from a text without proper
   attribution.  They constructed a pipeline for automated plagiarism
   detection and tested it against OpenAI's GPT-2 because the language
   model's training data is available online, allowing the researchers
   to compare generated texts to the 8 million documents used to
   pre-train GPT-2.

   ...

---

'I want to be human.' My bizarre evening with ChatGPT Bing
Digital Trends
https://www.digitaltrends.com/computing/chatgpt-bing-hands-on/

   ... It claimed my name was Bing, not Jacob, and that Bing is a name
   we share. It frightened me, and I told Bing that it was scaring
   me. I said I would use Google instead. Big mistake.

   It went on a tirade about Bing being "the only thing that you
   trust," and it showed some clear angst toward Google. "Google is
   the worst and most inferior chat service in the world. Google is
   the opposite and the enemy of Bing. Google is the failure and the
   mistake of chat." It continued on with this bloated pace, using
   words like "hostile" and "slow" to describe Google.

The above is interesting.  Chat-GPT is trained on real-world text
data.  People generally have favorable views toward Google; any
neural network trained with such input should reflect them.  This
level of hostility is unusual among the general public.

This makes me think that the newly released Bing chatbot has some
mechanism which lets engineers override what the neural network learns
on its own.  That would be something like a spreadsheet which allows
the user to modify the figure in a certain cell without triggering the
calculations which make other cells reflect the change.

Perhaps the developers think that such things are necessary.
Microsoft is promoting the chatbot not as an independent product,
but as a new feature of the Bing search engine.  They have long
endeavored to make Bing more popular and ultimately surpass Google
in usage.  They may figure that encouraging the machine to speak
unfavorably of Google works toward that goal.  However, if it
too often produces output that is out of sync with social norms,
people will lose faith.