But continuing with your analogy, I'll note that the people who fight financial crimes and fraud have absolutely had to integrate responses to cryptocurrencies. That didn't prove that there was lasting value to the cryptocurrency hype wave.
I will note that you didn't use it for most of the coding. And that looking at your code, you also seem perfectly capable of figuring out the AppleScript if you had wanted to. In an amount of time that would not have wildly increased the total project time.
So overall, this looks very "stone soup" to me. Did it get you over a hump? Surely. Did it actually contribute much in terms of code? Not from what I see in your script.
I do support work myself, and our developers have a function that is broken and shoving too much stuff into the database. This crap is compiled into a DLL so I can't do anything about directly. Went to GPT and said "hey, I have a table named X, when Y updates write a trigger that deletes this"
In a few seconds it dumps out a working trigger. Now, I've never written one of these before, dong SQL stuff isn't really my forte, but with a basic understanding of what I needed done I got working code with an explanation faster than I could parse the how to.
Even if AI doesn't contribute more than that, for lots of use cases it's more than enough.
This sort of reminds me of the 3d-printing hype cycle. People were very excited! We could make any object! We could soon be 3d-printing 3d printers! But in practice, 3d printing is now mostly useful for small, marginally valuable stuff. That's not nothing; I have friends who are excited about their 3d printers. But I also have friends who are excited about woodworking and accordion playing and falconry. But none of these are particularly world-changing.
My interest in it is this: given this (flawed) technology exists, what are the things we can use it for that are genuinely useful and help us build interesting things that we may not have built before.
I've been finding an increasing number of areas where it gives me a 10-30% productivity boost... and a few areas like the above one where that tips me from "not going to do that project" to "doing that project". Those are what's exciting to me, far away from the breathless hype around this stuff.
So in this case the leg-up that GPT gave me was the difference between the project existing and it not existing at all.
It would suggest an incorrect type as an input to a constructor and if you ask it to correct the error, it simply suggests another conceptually similar type to fit in, which also throws an error upon compilation.
A modified chatgpt that was bound under the hood, to a strict set of rules about the structure of c++ and it's relationships, could do much better. The bot gives a strong intuitive sense that we have only solved half or a third of a potential revolution in programming, with more to come.
You should start believing people are using this.
People are comparing LLM hype to crypto and VR hype and that’s really not accurate. Those are both always in a state of “someone is using it now, and a ton of people will use it in the future”. ChatGPT is in a state of “a ton of people are using it now”.
It would remind me of the explosion of social networking, except it doesn’t have network effects! It’s crazy how much it’s getting used by all sorts of people not because they want to interact with their friends (messaging apps) but just because they want to use it.
I was a big AI and even just generally ML doubter until I started using it as part of my workflow and reading about how other people are, too. At that point, you kinda have to update your priors.
To my shock, She started explaining to me how awesome it is and how all of her teacher friends are using it to offload a lot of the boring work like writing comments and stuff for students report cards. They are required to write numerous paragraphs for each student several times.
Something in the past I would take my wife an entire week each time.
These are people that need help setting up Wi-Fi on a laptop. And they are in love with ChatGPT.
She manages 80 students.
Alternate would be school hire a lot more staff.
If I'm understanding right, your wife is coming up with some bullet point style feedback about each of her students as part of a recurring evaluation, then using ChatGPT to generate lengthy prose from those bullet points.
Hailing ChatGPT as a savior here feels something like the tail wagging the dog. Why not just change the nature of the task so that what's required is to provide the quick bullet points, instead of using language model prose generation as a go-between for communication from teacher to student? What value is being added by the middle ChatGPT layer? More importantly, what value was being added by the 36 hours of marginal effort previously spent doing this manually?
That's one of my main hang ups with all the proposed use cases of ChatGPT - it seems we've created arbitrary problems and now have this complicated tool to solve them. Which in turn will enable us to keep making bigger and more arbitrary problems. But the problems don't need to exist in the first place
Reducing internal pushback to customer hostile bureaucratic requirements imposed by management on employees is a harmful effect of ChatGPT, not a benefit.
I agree with what you say. However she is not willing to a fight worldwide bureaucracy.
Employee pushback isn’t just individual.
What were they doing before ChatGPT?
Teachers tend to work very long hours because of having to grade work, write these kinds of reports, etc, all in their "non-teaching" time.
My entire family fully understands it's just a tool, you need to check it's output, and it's not perfect but it's saved them a bunch of time and allows them to focus on what they are best at by getting a good starting point from *GPT.
I'm not fully convinced it was a net-positive, because sometimes asking the "dumb" question shows that other people in class had that question but didn't want to ask, or didn't know they had that question until someone else voiced it.
The hype is real.