Will answer engines replace search? Four builders on the web's future
Marc Andreessen asks whether AI will replace search, Aravind Srinivas explains why Perplexity builds an answer engine and avoids Google's game, Sam Altman weighs search and ads, and Jimmy Wales considers what language models mean for Wikipedia. Their views show how the way people find information online may change and who stands to lose.
Do the four builders think answer engines will end search? Show answer
No. Three expect answers to take over much of what links do, but none predicts search simply disappears.
What is an answer engine, as Aravind Srinivas describes it? Show answer
A system that runs a search, reads the relevant pages and writes a concise answer with every part backed by sources.
Why does Srinivas think Google is slow to push answers? Show answer
Google earns its highest margins when people click links, and an answer that needs no click reduces that revenue.
Where does Srinivas admit Google is still better? Show answer
For navigational queries, meaning reaching a known site, and for real-time information such as sports scores.
What is Marc Andreessen's supply problem? Show answer
If people stop making web pages because nobody visits them, a major source of AI training data dries up.
How does Andreessen describe hallucination? Show answer
As what we call it when we don't like it, and creativity when we do. He notes a law firm wants it for exploring theories.
Why does Sam Altman dislike ads in AI answers? Show answer
He fears a question being answered with a nudge toward a product, and likes that paying users get uninfluenced answers.
What is Jimmy Wales's worry about assistants using Wikipedia? Show answer
Donations could fall if users never learn the knowledge came from Wikipedia.
What is the citation loop Wales warns about? Show answer
An error becomes its own source, as when a magazine printed a false detail that came from vandalism on his Wikipedia page.
What four ways can a retrieval-based answer still go wrong, per Srinivas? Show answer
The model misreads the query or passages, the snippets are stale or thin, irrelevant detail confuses it, or the wrong documents are retrieved.
What should you check before trusting a tool on a purchase or contested topic? Show answer
How it is paid for, since whoever pays influences what is shown.
Why should claims here about ads, models and law be rechecked? Show answer
The conversations date from a fast-moving period, and the speakers have a stake in the outcome.