4 ms·

Well, right now, a manager can’t just write a vague description of what they want into a text field and get back working software twenty seconds later that they

Well, right now, a manager can’t just write a vague description of what they want into a text field and get back working software twenty seconds later that they can deploy by clicking a button. That may change in the coming years, but the situation right now is that you need a human developer for any nontrivial production use case. That’s not the situation for copywriters. Marketing copy doesn’t have an objective threshold where it either works or doesn’t, and AI can easily take “here are some points to cover, give me three paragraphs” and spit out something that won’t be embarrassing to paste into an ad. In short, the difference is that AI can’t do our whole job yet, but there are some writing jobs where it basically can.

SilverBirch 4y ago
Well there's two points to address there. The first is the question of "Are LLMs 60% of the way to replacing a developer or can they do 60% of what a developer can do?" Because much like Tesla found out with Autopilot, just because you have a system that can do 60% of the things human drivers can do, doesn't mean you're 60% of the way to having a system that can do everything - in fact, you may be 0% of the way, your approach may literally be incapable of solving 100% of the problem. Secondly, people aren't really thinking through the chain of events that follow from being able to get AI to write low quality marketing fluff. It reduces the value of marketing fluff to zero and will flood the market with it, the world isn't just going to stay stagnant, it's going to change. Remember how email used to be just full of spam? Like, unreasonably terrible? We fixed that! That changed. So I think it's important to think about what we're going to change about the world to respond to this - and it may well be these writers have to take a whole new approach to their jobs, but let's not just pretend they're all just going to walk into the ocean.
jb3689 4y ago
The difference is really in the importance and stability of accuracy. Ad text can be wishy-washy. Computer programs cannot. An out of place word is okay. An out of place big can compromise the whole solution. We can apply this to other domains to get an idea as to what AI might be good at or might struggle with.

SilverBirch 4y ago

Well there's two points to address there. The first is the question of "Are LLMs 60% of the way to replacing a developer or can they do 60% of what a developer can do?" Because much like Tesla found out with Autopilot, just because you have a system that can do 60% of the things human drivers can do, doesn't mean you're 60% of the way to having a system that can do everything - in fact, you may be 0% of the way, your approach may literally be incapable of solving 100% of the problem. Secondly, people aren't really thinking through the chain of events that follow from being able to get AI to write low quality marketing fluff. It reduces the value of marketing fluff to zero and will flood the market with it, the world isn't just going to stay stagnant, it's going to change. Remember how email used to be just full of spam? Like, unreasonably terrible? We fixed that! That changed. So I think it's important to think about what we're going to change about the world to respond to this - and it may well be these writers have to take a whole new approach to their jobs, but let's not just pretend they're all just going to walk into the ocean.

jb3689 4y ago

The difference is really in the importance and stability of accuracy. Ad text can be wishy-washy. Computer programs cannot. An out of place word is okay. An out of place big can compromise the whole solution. We can apply this to other domains to get an idea as to what AI might be good at or might struggle with.