<img attachment="i_for_one_welcome_our_new_ai_overlords.webp" align="right" caption="I, for one, welcome our new AI overlords.">A friend sent me the article <a href="https://www.20min.ch/story/ki-security-finger-weg-von-den-neuen-ki-browsern-103443168" author="Michael Andai" source="20min">Finger weg von den neuen KI-Browsern</a> ("Hands off of the new AI-browsers"). The article largely focuses on the grievous security holes in these browsers, making them not browsers but data-exfiltration apps. In an age of unprecedented scammery, it is an affront that these tools even exist. But that's not even the worst of it. With a web browser, you type in an address and see the content hosted for that address. You <i>trust</i> your browser to deliver---unfiltered and unchanged---what you asked for. This <i>implicit trust is extremely important</i>, as the data your browser returns <i>informs your worldview.</i><fn> These aren't web browsers. They don't find content; they produce content. You don't actually see any web pages themselves when you "browse" with these tools. Instead, you see summaries generated on-the-fly that serve as a "response" for your "request". To be clear: you type in a prompt and see what the LLM generated as a response for that prompt. I would imagine that a lot of the pictures and short videos included in these responses are also generated. You will not see anything that anyone actually produced, unfiltered. <i>You are implicitly trusting that tool</i>---and the company that produces it as well as the laws of the country where that tool's infrastructure "lives"---to deliver a reliable worldview. For those of who use the web without an algorithmic feed, this feels like a significant change. It feels like <i>madness</i> to even <i>think</i> of using a tool like this. For people who have already been trained to simply look at what they're shown, this is more of an increase in the level of control that platforms have already had over what their users see and hear. They've been trained to not give it a second thought. Although it's not <i>technically</i> a significant difference over what a Facebook, TikTok, or Instagram feed already did, it is a big step in the wrong direction down a road these people shouldn't even have been on in the first place. If you can train people to become accustomed to this, then it opens the door to further great leaps forward for controlling what they see and hear. The level of control over what people see and hear is already too high for comfort. However, while your ability to directly access content is sometimes <i>impeded</i> with a web browser, it has, until now, never been <i>transformed</i> or <i>interpreted</i>. This is very, very different. <hr> <ft>Similarly, you trust your newsfeed (RSS) reader to reliably return <i>everything in every feed, exactly as it was published.</i> We assume that this is how it works because we aren't trained to think like criminals. There are a lot of bad people running things who don't have that limitation.</ft>