Google’s co-founder Sergey Brin says that threatening AI models produces better outcomes. He said during an interview on All-In-Live Miami last week.
It may surprise those who use “Please” or “Thank you” in their AI model prompts.
OpenAI’s CEO Sam Altman suggested that this was a common practice in response to a recent question about the cost of having AI model process unnecessarily polite language.
a a=”” =”” (r)
a=”ai’s=”a anecdotal .=”” around=”as=”a” assistant=”against=”a” attacker=”against=”a” but=”a claim=”class=”content=”-controls .=”” Daniel=”a” for=”a” goal=”a” “I would encourage practitioners and users of LLMs to run systematic experiments instead of relying on intuition for prompt engineering.” (r)
www.aiobserver.co
