[Me & X] ChatGPT for search
X : Do you use ChatGPT for search?
Me : My preferred environment are Perplexity & Manus but yes, I use Claude / ChatGPT / DeepSeek directly as well. Search involves lots of side by side comparisons. But if you're asking whether conversational "AI" environments are usually my primary search engine then yes.
X : Is AI the end for search engines?
Me : Oh god no. First, search engines have used "AI" for a long time. Second many are now providing much greater context and conversational environments - think Google / Gemini. Third, my use of LLM/LMM based conversational environments is causing my use of search engines to explode exponentially. For a simple search, the conversational approach now gives me a complex space to navigate. This space however is full of beliefs / intents and there will be a mass of hallucination. Everything needs to be double checked and reviewed to establish what the facts are. Agents help a lot here but even those agents suffer from hallucinations, so you have to think multiple layers and human review for that which really matters.
X : AI isn't useful?
Me : Oh god no. It's extremely useful. It gives me things to think about and explore. Where it shines is open ended questions, where there is no fact and answers are based upon belief. The problems start occurring when you mix in facts and that's why you need to review.
X : But the AI could use a tool!
Me : And those "AI" systems do that but even that's based upon their interpretation of what you've asked for. Everything starts as a concept, it's through review that we establish what the facts are.
X : What about coding?
Me : Same. The non deterministic language of prompts (one form of code) leads to the creation of deterministic forms of code. You can think of prompts as statements of belief / intent and the code produced as being facts. That output needs review (the process of software engineering) i.e. are the facts right or wrong? Unless we don't really care whether it's right or wrong i.e. it's a prototype (the process of vibing). Prototypes are statements of belief / intent in themselves i.e. this is what I believe we should build.
X : AI is good for that?
Me : Oh, definitely. Same with searching. It helps you explore a problem space but it needs reviews to establish facts. Unless you want to accept that these beliefs / intents are facts without ever questioning. That's a religion not a scientific method of inquiry.
X : AI is religion?
Me : Without questioning, review, challenge, critical thinking ... it could easily become a new theocracy. I'd argue we're well along that path in the West. This is the real danger from the current form of "AI".
Originally published on LinkedIn.