@mike I agree however the web is already broken nowadays with walled garden platform and ads everywhere. AI content is just an addition to the enshittification. People and new ideas are the key!
@mike Slightly off topic but with Google abdicating it's role in search does that leave the door open for Bing? I know Bing is still the butt of many jokes and Microsoft is racing towards everything AI along with everyone else, but will someone in Redmond sense the opportunity?
@mike I think Bing is also all-in on AI. We're going to need a totally new way to discover and navigate the web that is fundamentally better than a search engine. I believe trustworthy people recommending quality content to others who share an interest will be a fundamental part of the solution.
@uriel@mike Yes I've used it and I think it's where Google and Bing are headed. Perplexity is essentially forcing Google to move much faster. OpenAI will soon release something similar. All paths lead to the same result: subsume all the content on the web in order to satisfy the search on site and minimize the need to link out to the web.
@seeingwithsound@mike I personally wouldn’t completely discard LLMs. For example, I think self hosted LLMs could be helpful in curating the federated feed. Similarly, ML recommendation systems could help people find interesting accounts or posts among other things. I think the problem is with deception and lack of control, but it doesn’t have to be that way.
@andresmh@mike Agreed, there are many potential benefits to using LLMs, including in blind accessibility, but it remains a mixed blessing. I would for instance suggest not to never trust an LLM as an AI sighted guide in smart glasses, even though it can be very helpful to have the scene in front of you described in order to better anticipate things ahead of you. Who trusts an occasionally hallucinating sighted guide? Take anything an LLM says as at best a hint of possibilities or probabilities.
@seeingwithsound@mike for sure! I just don't get the completely anti-AI vibe sometimes. Partly because AI doesn't mean anything and means everything at the same time.
@andresmh do you have any recommendations for small, self-hosted models? I’m curious about trying out development with one, but I don’t know how to filter the fluff and choose a truly open and ethical toolkit
Add comment