paralel between digital photography and using llm’s
i happened to be editing scans of my last roll of film on my phone while i was taking a boring introduction into github copilot today. then a realization hit me: writing a piece of computer code or editing some text yourself vs using an llm like chatgpt is quite similar in comparison as making an analog picture vs making a digital one.
it is much easier to make a digital picture. and if millions of people make thousands upon thousands of picture with their phones each, there will surely be some great ones among them. but that won’t get you around the fact that it takes patience, time, effort and sometimes expensive equipement to make a great picture on purpose.
in fact, i would argue that it is easier for an amateur photographer to make a great picture on film than using a digital camera. mostly for one simple reason: the limitations of that setup - the exact same things that make digital photography attractive - force you to slow down and think. to be creative. how is one supposed to improve when there is no effort involved and no cost associated with mistakes?
just as digital photography opened the options for many people to do something they could not efficiently do before, llm’s will open doors for many people to do new things. but just as photographers didn’t disappear, the experts in their respective fields will also stick around.
You can’t beat the learning feedback cycle of digital. You develop your skills faster. Most of the knowledge transfers to film.
maybe you are right… if developing your skills is what you consiously do then you can do it faster with digital. that’s of course, not what most people taking pictures want.
i would still argue that the lessons stick better if you need to put more effort in and slowing down is a benefit in its own right.
Indeed, deliberate practice is a requisite to develop skills. And to transfer those skills, the equipment must be somewhat similar. A DSLR and an SLR. A camera phone and a point and shoot.