Search engines are basically 90% blogspam promotion machines at this point, and the blogspam churning engine is going to become completely autonomous. The tools we're relied on for nearly three decades now are going to become a big virtual tug of war between machine-learning-guided SEO systems and machine-learning-guided ad revenue systems.
Every single anonymous interaction is going to be suspect. The next dogwhistling fucking cryptonazi groyper you run into is probably not even going to be a human being. Machine learning is probably going to elect the next US president. Eventually the medium will become so polluted that we will have to go back to doing everything in person.
Thank you for reading my doompost
I tend to agree. The amount of auto-generated websites with "answers" you can tell are already using generative language models based on what was initially genuine search results is already insane; it's particularly bad when it comes to searches about tech stuff, or video games. Those off-the-shelf models are only going to get better. And it'll be all across the board: product reviews, fake positive posts on hobbyist's forums, and so on and so on.
I also wonder the impact it'll have on future language models; in order for the models to generate a wide range of content - on all topics - even remotely linked to the truth, they have to plug it in with internet data in the training set. What happens when said internet data itself becomes progressively polluted by the output of other ML models ? you could imagine falsehoods spreading exponentially this way.
It is a problem, and the aiweirdness tumblr person (who is something of an expert on machine learning) has speculated that machine learning free internet data will start to command a premium, so you don't get garbage in your data inputs. Like Google or the Internet Archive might be sitting on a goldmine treasure trove because they've been scrapping the web forever.
Low background steel but instead it's low AI internet data