It used to be a beautifully imperfect world
LLMs make a lot of tasks faster, more accurate, easier. Like writing a function in code, analyzing long text, or searching for a fuzzy statistic. Before November 2022, you would have to look through wikipedia, a blog post, or images and reason together an answer. Before the internet, you would have to look through a book or ask your family or friends.
The use of ai cuts the important part out, the friction that hones your reasoning and understanding. It stops you from arriving at a solution that isn't necessarily the best but is still yours.
LLMs have expanded the equivalent of offloading information to a Google search to offloading reasoning to a prompt. Why remember anything when you can search it? We've figured this one out: because storing knowledge in your head makes it accessible in your subconscious mind, in your dreams, in the moments where you think. Why reason through a random unimportant problem? Because most problems are unimportant, and you build your muscles slowly over time. Where you learn to break things down, recall what you already know, experiment, use feedback.
Google showed a new feature today where you can describe any problem, and a video pops up, right at the moment where you learn what to do. They showed how to teach a kid to ride a bike for the first time. The speaker reminisced about teaching his kid to ride. Instructions now provided, for anything.
One of the definitions of sapiens from homo sapiens is "understanding". We're on the way to becoming homo actio, we act, but we understand less and less.
If I didn't need to worry about my income it'd be very easy to never use AI again. It's dehumanizing. But most people do need to worry. Many people have already discussed how if everyone else is inflating their output you probably have to do the same sooner or later. You may take half an hour to read about something that's trivial for an LLM to complete in seconds. This is especially unfortunate at the beginning of your career, when it is most important to not ask an LLM, to struggle and develop a good foundation. Maybe some employers understand this and let people take their time, but the employer's competitor is using AI and just launched what your company will finish in two months.
For the sake of argument, of course, cut AI, recommendations, guides, how-to's, and regurgitated mediocre advice from the internet from your life anywhere that isn't job related. It gets interesting at the weekend project that requires skills outside of your domain. Why not finish it in a day with an LLM, have it reason for you and you just act. Well, because maybe the accomplishment is in the process, because you wouldn't be able to do it again without internet? Here's the catch, the temptation is very strong, especially when you only have a few hours of free time after a week day and four weekends in a month. We're trying to squeeze as much life out of time we don't control, but in the process, we're poisoning our souls for more life. We aren't meant to squeeze. But that's the way thing are.
You're paid to complete tasks. Over the long-term, AI is most definitely making us less capable, at least I think so. On the flip side, it can speed up some tasks so dramatically that resisting it doesn't seem like a wise choice anymore. The crux of the argument here is that you probably don't have very much time or leverage or security. If you are retired, sitting on a pile of money and code for fun, of course you don't have to use LLMs. What if you just graduated college, have loads of debt, and are competing against others who do use AI? What if your manager encourages it?
Back to the homo actio point. We're already making robots capable of sorting boxes for 24 hours straight. My theory is that as soon as it's possible, any expensive, conscious human worker will be replaced with a robot right away. We're halfway there, we're hybrids now. It doesn't matter if it poisons us, just fire and replace. I'm talking about the slow creep of in-between tasks, little details that add up, nothing more.
You are powerless to stop this progress. So, is the strategy to find work in fields that don't benefit from ai co-working right now, or more interestingly, fields that could never benefit from automation at all? Maybe the solution is just go along, make money as a hybrid and leave when you can, protecting your time away from work.
When that google speaker was teaching his kid to ride a bike, he didn't pull up an ai for optimization tips or read a blog or watch a video, he probably went outside with no clue. The kid may have fallen over or even figured it out on their own. Whatever happened on that day, it was definitely imperfect. It was human reasoning and human being. Whatever happened on that day now seems as beautiful as ever.