Machine Learning Everything
I have been intrigued by AI doomers, especially in the wake of large language models (LLMs). According to doomers, humanity will almost certainly be wiped out in short order and replaced by AI overlords we foolishly brought into existence.
When you hear doomers talk about existential risk, they often rely on metaphors and analogies. We were smarter than our ancestors and they’re no longer around so naturally a more intelligent being will displace humans.
But when pressed for an hypothetical about how a set of numbers can break out of its digital representation and enslave us all, doomers have a hard time coming up with a plausible scenario. Most doomers would say they are not smart enough to figure it out but the AI will be smarter.
Some will come up with harmful scenarios. For instance, I saw an example in which an AI bot ws not sufficiently dissuasive when a child talked to it about meeting an older man she had met online.
But this is a silly argument as AI isn’t replacing an existing system that would protect this child. Are we really going to offload child welfare to an LLM that was released a few months ago? Besides, these things will necessarily get better and are easily tweaked to make improvements in targeted fields.
