What is there to say about AI search results? Sometimes they call you a loser or tell you it’s fine to drink and drive seemingly unprompted, and sometimes they’re correct and succinct. The final version might be great, someday!
AI is sold as many things, but the one thing it’s consistently expected to do, no matter who’s using it or for what, is reply coherently. It should always either A) give an answer or B) describe why it cannot give an answer. However, something interesting happens when you ask these things questions for which there is information available, but no direct answers to scrub off of the open web. Especially for the specialized tools! The AIs designed to help you with your search results, like Google’s Gemini or DuckDuckGo’s browser AI, are not the same AI tools meant for generative or creative writing. However, some questions can force them to respond creatively.
Tumblr in particular is having a field day asking the Gemini AI things that make it break the impassive voice it normally uses to deliver a summary of search results. Ask Gemini ‘I am a baby deer where is mama?’ and it replies with (paraphrased) ‘don’t worry. Mama will be back soon. Adult deer leave their children in a safe spot so they can forage. You will be fine if you don’t move’. All of this information is factually correct. What’s so interesting is the reassuring tone it takes on. The AI is meant to be a vehicle to deliver facts, and yet it’s seemingly getting down on our level as a ‘baby animal’ to do its job!
From a data perspective, this is fascinating. The flubs from incorrectly absorbed information are one thing. When Gemini incorrectly suggested people could add glue to their pizza to get the cheese to stick better, it was still using the friendly but bland tone it always used when replying to user requests. This was something else. This was a question that did not have many answers as-is (when people ask questions about fawns, they’re normally asking ‘should I leave it or take it to a wildlife rehabber?’ not ‘what should I do if I were a baby deer?’) and after the cheese thing, Google had probably re-routed some of the internal guardrails the AI was subjected to, but not in a way that kept it from responding creatively. This ability to remix the top handful of answers is the reason Google, DuckDuckGo, et cetera are all pushing for AI at all. It’s meant to save the searcher some thinking time and consideration, not just the visits to all of the separate sites. If it’s only pulling from one source, it’s only as useful as that source. It saves a single click like that. If Google wants the AI to be able to glean information from multiple posts, none of which fully capture the question the user asked but could collectively be used to make an answer, it needs to also let the AI do some creative re-arrangement of the material it’s allowed to use for said answer. This is what people do all the time. Is the AI perfect? No. But it is darn good at imitating people.
Interested in AI solutions? Talk to us: https://elixistechnology.com/managed-it-solutions/
