Generative AI for Genealogy – Part V

The Cost Problem: Why Not Just Use Reasoner v1?

At this point, a reasonable person might ask:

“Why not just let Reasoner v1 do the matching?”

And yes it could. GPT suggested it too.

But here’s the catch: Reasoner v1 is not the long‑term plan. The future is cloud‑based AIs, and cloud AIs charge per token.

If Reasoner v1 has to:

  • read a list of example questions → 243 tokens
  • or read all questions and their examples → 169 words (≈ similar token count)

…then the cost of asking it to reduce tokens is more tokens than just sending everything to the main model.

This is the AI equivalent of paying £10 for a taxi to drive you 200 metres to avoid a £2 parking fee.

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *