Raw LLM Responses

The more I see people writing that ChatGPT gives out biased, on-rails content, the more I believe that these people's prompting skill just plain sucks. They do not understand that their input needs to be parameterized and explained, you cannot prompt GPT (yet) with minimal detail and expect a human-like level of output. Case in point, just last week there was a recruiter ranting on LinkedIn that their colleagues should not use ChatGPT because it yields non-inclusive job descriptions. She even quoted a "study" that stated that GPT's job description texts had low scores on all sorts of inclusive language parameters. Obviously, that "study" did not mention the prompt used, not showed the outputs, only the final scores. So, I went to GPT (4) and prompted "Act as a tech recruiter for a local startup. Please create a job description for a backend engineer specializing in Kafka and AWS technologies, with about 3 years of experience combined. Make sure to mention some perks such as remote work, WFH allowance, gym membership or language courses. Try to use non-gender specific language and avoid ageist or ableist terms, while highlighting the diverse and inclusive work environment of the company." Guess what, the JD was absolutely spotless and ticking all the DEI checkmarks. Garbage in, garbage out.