Many years ago, when search engines first emerged on the Web, finding the right page felt like a guessing game, trying to identify the perfect combination of keywords. Early players like Lycos, Altavista, Xcite, and Yahoo were eventually eclipsed by Google’s PageRank technology. In a sense, the guessing game ended, or at least we learned how to play it better.
Today, LLMs’ current state of maturity is not much different, I think. Instead of keywords, we experiment with different combinations of full sentences to get the content we want. In fact, it’s becoming increasingly common to see job postings for prompt engineers.
As we learned in university, software should be deterministic, meaning the same input always produces the same output. So, when you input “1 plus 1,” the output is always….. whatever the programmer defined, but it remains consistent for that input. This principle has long been one of the core axioms of programming.
When Google introduced PageRank, determinism became more apparent, the guessing game faded, and we all began enjoying more accurate search results. It also gave rise to the online advertising industry, allowing businesses to bid on keywords that aligned with specific search queries to promote their interests.
Following that Chain of Thought (CoT), it’s safe to conclude that guessing and value creation are inversely correlated: the less guessing, the more value, and vice versa. This might suggest that the demand for prompt engineers may be short-lived, and a tremendously prosperous future for some AI companies.
That said, I recently came across a particularly clever prompt online that I must admit I tried and really liked.
Here it goes:
Your goal is to name a new podcast that will cover AI, provide content on what AI is, and offer commentary on recent news, trends, and books.
Use the “playoff” method to find the ideal name.
Please follow these steps:
Step 1: Generate Names
Create a list of 32 distinct podcast name ideas. Each name should reflect the focus on AI, setting the foundations of the subject, and providing insight into news, trends, and books. Be creative, ensuring that names are clear, catchy, and relevant.
Step 2: The Playoff Rounds
Round of 32: Pair the 32 names into 16 matchups. Compare each pair, choosing a winner based on:
• How memorable and catchy the name is.
• How well it reflects the podcast’s goals (education on AI, commentary on news, trends, and books).
• Additional qualities like creativity and originality.
Round of 16: Form 8 matchups from the previous winners, select the winner from each pair, and explain your reasoning.
Quarterfinals: Form 4 matchups with the 8 remaining names. Choose the best from each pair, providing reasoning.
Semifinals: Narrow down the 4 remaining names into 2 matchups and select the best name from each, explaining your choice.
Final Round: Compare the final 2 names and select the overall best name. Explain why it’s the strongest choice, based on how it reflects the podcast’s goals, its appeal, and other key factors.
Step 3: Final Output
Provide the original list of 32 names ranked from best to worst. Include the final chosen name and an explanation for why it was the top choice.
The explanation might or not, make sense to you, or at all, but it is undeniable that now you have 32 options to choose from.