Oof.
Those "AI Browsers" are a real privacy hell
Personally, I strongly recommend NOT to use these
Oof.
Those "AI Browsers" are a real privacy hell
Personally, I strongly recommend NOT to use these
Aside from the privacy issues that should be enough to convice you not to use these
Here is another small simple example:
AI Browsers can be exploited super duper mega easy; for example, add code/instructions to an image that is too small for humans to read but can be read by the AI Tool
That alone gives endless possiblities when the users uses AI on that specific website..
💡
@stux
There are extensions that block them.
I'm no longer getting search results that direct me to Rob's facebook posts in Pittsburg.
@CatHerder @stux I'm afraid they're not talking about "ai" search results: several companies (including openAI) are launching whole "ai web browsers" where you ask for information and the software provides response (supposedly collecting and summarising actual websites, but we know how this shit actually works).
They purposefully do NOT send you to the actual websites, BUT they DO parse them (so malicious instructions WILL reach them).
There is no way to separate commands and data in LLMs.
A privacy hell, and they don't even work -- by all accounts rather than returning a page about what you ask them, they feed up slop and gibberish, just like any OTHER mass-marketed LLM program.
Which makes them a lie in two parts - they're not intelligent, and they're not browsing the web, they're just pretending to. The 'artificial' bit is probably true.