It’s… Challenging. Like the pet eating thing, there are many sources saying it’s true and many saying it’s false. Official sources can lie (Russia came to mind for no reason whatsoever), so we rely on sources we already trust, which is tricky and even subjective.
I imagine that “if in Fox then False” is a good start, but aside from that I can only think it getting extra sources, also a challenge without real time web crawling of the internet, were google and Microsoft are already light years ahead.
But if Mozilla can, for example create a sources list and even charge for the ability to be a default on said sources list, wouldn’t that be a double win? The problem with things being unreliable can be dealt with via language. Like big red text saying don’t trust this blindly.
They can also do intelligent searching and simply surface links.
Do I trust LLM summaries? Not fully. But how about the strategy used by an app like BeyondPDF for Mac:
Think: Firefox does the search, then gives you the sources and the most likely relevant excerpts from each. Consequences of it searching wrong? A small waste of time, but no misinfo.
Sidebar!
One can be against environmental costs of great machine-learning powered search, and offended by the arguable IP theft that created the tools, but it’s unlikely all those who say they “don’t want AI anything!” really mean that entirely.
“I don’t want or need the current version of ChatGPT for my use cases” is very fair though. Maybe they don’t have any SQL queries or Excel formulas - on the edge of their abilities - to build, or text to beautify, or quirky esoteric philosophy to bounce off a robot…
It’s… Challenging. Like the pet eating thing, there are many sources saying it’s true and many saying it’s false. Official sources can lie (Russia came to mind for no reason whatsoever), so we rely on sources we already trust, which is tricky and even subjective.
I imagine that “if in Fox then False” is a good start, but aside from that I can only think it getting extra sources, also a challenge without real time web crawling of the internet, were google and Microsoft are already light years ahead.
But if Mozilla can, for example create a sources list and even charge for the ability to be a default on said sources list, wouldn’t that be a double win? The problem with things being unreliable can be dealt with via language. Like big red text saying don’t trust this blindly.
They can also do intelligent searching and simply surface links.
Do I trust LLM summaries? Not fully. But how about the strategy used by an app like BeyondPDF for Mac:
Think: Firefox does the search, then gives you the sources and the most likely relevant excerpts from each. Consequences of it searching wrong? A small waste of time, but no misinfo.
Sidebar!
One can be against environmental costs of great machine-learning powered search, and offended by the arguable IP theft that created the tools, but it’s unlikely all those who say they “don’t want AI anything!” really mean that entirely.
“I don’t want or need the current version of ChatGPT for my use cases” is very fair though. Maybe they don’t have any SQL queries or Excel formulas - on the edge of their abilities - to build, or text to beautify, or quirky esoteric philosophy to bounce off a robot…
Pay to be the “truth” on a fact checking tool? Fox news is very interested.
Aren’t Google and Bing and others paying to be featured in Firefox. What’s the difference?