Comment by besterman23
7 hours ago
I fear there is a tendency for some people who are against the proliferation of AI to make sweeping gut-reaction comments like this that do more harm to their position than good.
Perhaps everything that you said is 100% correct, but people are easily swayed by things that “sound right” being overly aggressive opens your position (in people’s monkey brains) to “enlightened centrists” that leave the door open for bolder opposition.
Many people have been genuinely helped or feel like they have been genuinely helped by LLMs in coding, they aren’t going to change their opinions based on fiery rhetoric and accusatory tones. Much like I can’t force myself to believe in Santa Claus regardless of how aggressively my nephews claim he is real.
underrated analogy to claiming Santa Claus being real by nephew. lmao
it's hard to be swayed into skeptic camp if you see how it works and it really solves those hard problems that took hours. Genie is out of the bottle
I'm part of the skeptic camp, and I'd argue that we are never going to have LLMs solving all problems. Therefore, we need humans to know and understand how to solve problems. Unfortunately, when "the genie is out of the bottle" and most (but not all!) problems are solved, it becomes very very hard to argue for doing things the hard way.
Furthermore, it becomes practically impossible to argue for doing things the hard way for all of the solved problems, which were and are necessary for being able to learn the craft deeply. At the very least, deeply "enough" to advance the state of the art.
In short: 1) Humans are necessary. 2) Knowledge and experience are necessary. 3) Humans need time and practice to gain knowledge and experience.
When one side argues that one or all of the above points are false, I don't believe them.
Realistically, we are seeing people even in the most pro-AI groups begrudgingly concede #1 and #2, while expecting #3 to just sort itself out.
That’s the hard part with all of this, you’re either of the opinion that there should be some individuals capable of determining if the AI is doing the right thing, and therefore all the previous prerequisites are still necessary, or that we should just let it all ride on the AI, which is probably easier, but certainly more dangerous.