By Dennis D. McDonald

Years ago I sometimes spoke at professional conferences on how electronic publishing and networked information retrieval systems were impacting institutions like libraries, publishers, and academia.

I once titled a speech, "Do You Want Your Electronic Research Results to Be 'Thick and Juicy' or 'Thin and Crispy'?"

That question resurfaced with OpenAI's announcement of "Search GPT," which might eventually challenge Google's advertising business, especially if Google is no longer the default search engine on Apple products.

Search GPT summarizes online search results and provides links to source material. However, these features aren't unique; Perplexity, for instance, displays both concise answers ("thin and crispy") and a clickable source list ("thick and juicy"); as a result I use it a lot.

My own (limited) research suggests that ChatGPT's responses can be less verbose than Perplexity's, although this finding admittedly was based on a very limited experiment.

Some AI companies aare entering into formal licensing agreements with potential search targets to sidestep legal issues experienced by some LLMs having used online intellectual property without explicit permission.

I expect such negotiations will eventually be resolved, but the resulting agreements could lead to a "Balkanization" of web access, with different search engines having specific rights to crawl, summarize, and link to various resource sets. Also, once governments start regulating these AI-linked licensing and access agreements,, we might see not only a decline in online advertising revenue for some but also a shift in the concept of a 'worldwide web' open to all. Apple’s emphasis on privacy might eventually prevail, but that’s uncertain, given the lax attitude many users have towards online privacy.

For those managing complex bodies of professional knowledge—like in medical, financial, or engineering fields—I'd advise caution in allowing AI systems to access your information, considering the potential loss of control and revenue due to AI's advanced querying capabilities.

One approach for owners of specialized bodies of knowledge might be to license tools that can be used to provide their own intelligent, proprietary search interfaces. This could safeguard intellectual property and ensure quality control over content updates; it might also heighten the "walls" of the Internet’s "walled gardens," further damaging the concept of a “world wide web” open to all.

Back to the metaphor of thin and crispy versus thick and juicy. Sometimes, I prefer straightforward, concise information ("just the facts, ma'am") when checking a fact or number. Other times, I seek detailed, nuanced searches to uncover context and understand facts independently.

If a search engine can offer both through a plain, user-friendly interface at a reasonable price, I’m all in. Whether that might be SearchGPT in collaboration with Apple remains to be seen.

Postscript: Recently, I asked online for a "scary ghost story for a six-year-old girl interested in gymnastics, math, unicorns, and Harry Potter." Claude, Perplexity, and ChatGPT all provided engaging "thick and juicy" responses. Unfortunatele, she was disappointed as none truly frightened her!

Copyright © 2024 by Dennis D. McDonald