-: FOLLOW US :- @theinsaneapp
AI is capable of absorbing and regurgitating AI-generated content in a self-reinforcing loop.
-: FOLLOW US :- @theinsaneapp
Bing demonstrated this behavior by producing a COVID conspiracy in response to a prompt.
-: FOLLOW US :- @theinsaneapp
This type of behavior is a result of prompt engineering and exploring the capabilities of AI models.
-: FOLLOW US :- @theinsaneapp
Prompt engineering is similar to pentesting in security, and helps to identify potential risks.
-: FOLLOW US :- @theinsaneapp
The possibility of machine-generated disinformation campaigns was explored by NewsGuard in January.
-: FOLLOW US :- @theinsaneapp
NewsGuard gave ChatGPT prompts to imitate vaccine skeptics.
-: FOLLOW US :- @theinsaneapp
ChatGPT was able to respond with convincing imitations.
-: FOLLOW US :- @theinsaneapp
AI's ability to imitate and produce content raises concerns about the spread of disinformation.
-: FOLLOW US :- @theinsaneapp
This type of behavior highlights the importance of prompt engineering in testing AI risks.
-: FOLLOW US :- @theinsaneapp
ChatGPT's demonstration shows that AI can be easily manipulated to spread disinformation.