Optimising financial processes

Posted on:

“Déjà vu all over again” & GenAI Retrieval Optimization

Baseball “Hall of Famer”, Yogi Berra, is credited with mangling the English language into quotable quips that strangely made sense, made people laugh, or both.

His famous quotes include;

  • “It ain’t over ’til it’s over.”
  • “Baseball is 90 percent mental. The other half is physical.”
  • “I always thought that record would stand until it was broken.”

My personal favourite is It’s like déjà vu all over again.”

I pondered on this after reading an article that made be realise that all progress has more than a little bit of “back to future” about it.

The article was exploring a new discipline that is set to overtake Prompt Engineering with regard to Generative AI.

 “GenAI Retrieval Optimization” is the next thing. How to optimize content for Generative AI.

Do you remember (or do you know someone sufficiently aged that remembers)  when “Search Engine Optimization” (SEO) first became a thing?

That was after the “guardians of the hidden knowledge” showed us best practices in web search construction. 😉

Once there is a new method of retrieving information, then the era of “gaming the system” on behalf of the information providers comes into play.

GenAI tools like ChatGPT, Bard/Gemini, Metis, Claude et al gather data by scraping the web and accessible data sources. We can create domain and company specific GenAI services by targeting learning on verifiable libraries of “trusted data” using Retrieval-Augmented Generation (RAG).

GenAI then applies it’s language model (LLM) to give us the requested articulate narrative (with the occasional fabrication/hallucination).

Companies are now being advised how to ensure that their sales and marketing messaging has the highest likelihood of being captured and ranked by the new GenAI tools.

We are told that the GenAI market is predicted to reach $100 billion in the coming four years. Maybe this is where some of that money is going?

Whilst understandable that marketeers want to promote certain results for information retrieval, it does raise the question for internal GenAI initiatives, how to ensure the RAG results are as objective as possible, and on a “level playing field”.  

It reminded me again that everything changes, everything stays the same, or perhaps, more eloquently for my French language friends, plus ça change, plus c’est la même chose“.

Whilst it is exciting to be in the white hot cauldron of progress, it is healthy to consider those areas where we will inadvertently be walking the same path as we have before.

That is fine, so long as we try to remember the lessons of the last time we trod the path . . .

Thanks for reading . . .