Live data from Hacker News

AI-enhanced development makes me more ambitious with my projects

simonwillison.net

381–390 of 519 posts

Re: AI-enhanced development makes me more ambitious with my projects

#381

> This doesn’t just make me more productive: it lowers my bar for when a project is worth investing time in at all. This is the real benefit here. AI isn't taking jobs, it's going to enhance them - once we learn how to use the tools we have available to us.

“AI isn’t taking jobs” If by AI we mean things beyond just ChatGPT sure it will, it’s only a question of how long before it happens and how many new jobs will be created to replace the old jobs. For example, AI isn’t going to enhance customer service, you’ll just have better automated experiences and businesses will hesitate less to use them. Edit: AI isn’t being created for the benefit of mankind. It’s being created…

Beyond ChatGPT and whatever its iterations are, it will be a matter of time before jobs are irreversibly changed - no one is debating that. But too many people are taking the view that it will destroy more than it create. I think it will create more than it destroy. In many cases, the type of jobs we can do will diversify and make life more accessible.

Re: AI-enhanced development makes me more ambitious with my projects

#382
Same here. I've done a few little projects with ChatGPT4 in my back pocket. I've been able to give more aggressive timelines, more features -- and I've been able to hit those goals with ease. Probably the first time in my life I've been able to over promise AND over deliver.

Also, the final product (with documentation) has been complete and pretty amazing.

Re: AI-enhanced development makes me more ambitious with my projects

#383

Earlier quoted context omitted.

Yes

Please don't do this here. The topic deserves serious discussion.

Therac 25 is so serious that jokes are one of the very few ways to discuss it at all, eh?

Re: AI-enhanced development makes me more ambitious with my projects

#384
post #374

Upwards of 80% of the effort in commercial software development is maintenance. You could speed up the remainder dramatically and make very little progress on the rate progress happens and have people scratching their heads about why.

That may not be true in this case. Cheaper "second system syndrome" may allow maintenance to be perceived as a cost center.

Maintenance is already perceived as one hell of a cost center.

I'd have enough money if to train my own foundation model if somebody gave me a nickel for every time somebody underestimated the time to rebuild an existing e-business system by orders of magnitude because they were ignorant of all the "requirements" that were satisfied by the existing system that were undocumented.

The most expensive maintenance problem comes when you've accumulated a lot of data that is incorrectly structured and may or may not be possible to migrate to a new, correct, structure.

Fred Brooks in No Silver Bullet pointed out that software development involves multiple phases, say 5 of them (requirements gathering, design, coding, testing, deployment) and that even if you could reduce the time involved in one of those to zero there is no "revolution".

The corollary to that is that the opportunities from ChatGPT come out of identifying the things that ChatGPT does poorly. For instance it might be just as important to turn code back into requirements as to turn requirements int code. There was a post the other day about a "simple" problem in CSS that turned out to be hard because of the deep understanding required

https://news.ycombinator.com/item?id=35361628

the ratio is much worse between spending 3 minutes to make a partial solution with ChatGPT and 3 hours to really solve the problem as opposed to 30 minutes to make a partial solution manually and 3 hours to really solve it. That is all the more reason to clean house at the lower levels, either make a style sheet language that is not so fraught or maybe make one that's designed so that chatbots can write good code for it.

There has been hype over "no code" and "low code" tools for the longest time and people have been chasing that revolution for a long time and it's astonishing how many innovations have been forgotten (like the 1990s CASE tools that could turn code into diagrams and back again and leave formatting intact.) Real progress involves solving all the problem that is ahead of us instead of cherry picking the parts of the problem we want to solve and thinking wishfully that will be sufficient.

Re: AI-enhanced development makes me more ambitious with my projects

#385

I can definitely relate to your experience. Just last week, I leveraged AI tools to develop some data processing code. Estimating how long it would have taken me to write that code without these tools is a challenge because, to be honest, a year ago I wouldn't have even considered attempting such a project. It might have required a week or even two of my time, which I simply cannot spare for this kind of undertaking.…

Have had similar experiences. I found two HTTP servers running on my TV and I wanted to see what was going on in them.

Dirbuster wasn't packaged for Debian (that I could find). I wasn't too keen on trying to build from source or anything like that. A few months ago that would have been that.

But with GPT I just asked it to write a POSIX-compliant C port of dirbuster. Less than 10 minutes later I was scanning my TV for routes.

https://sharegpt.com/c/stLFWA1

Re: AI-enhanced development makes me more ambitious with my projects

#386

Earlier quoted context omitted.

I'm of the thought that #2 will happen. #1 will never happen - there are simply not enough developers to go around. To use a relatively modern analogy, AI-enhanced development is to developers what Photoshop is to photographers. No one ever said Photoshop would put photographers out of business. Now Photoshop is a requirement for photographers. The pictures still need to get taken; likewise, for developers - the busi…

>To use a relatively modern analogy, AI-enhanced development is to developers what Photoshop is to photographers. Photography as a job has been hit heavily by modern smart phone advances. Heavily. It's now very hard to get paid a living wage as a photographer. With mid journey it's now even harder. Photographers weren't completely replaced. I would say they were mostly replaced because now every Tom dick and Jane can…

I know a few photographers and they have plenty of work and make nice money. The pictures that they take can't be compared with someone taking a picture with a smartphone.

Both wedding photography and family/portraits.

Re: AI-enhanced development makes me more ambitious with my projects

#387

What I really love is the unbounded optimism of the AI. How many colleagues could you ask "Can you write me a PoC for a horizontally scalable, probabilistic database in Rust, using Raft-based consensus?" that would then just think for a second and go "Certainly! Let's start by ...". These people exist, but they're one in a thousand (or million?). ChatGPT basically does that for almost every prompt you throw at it. So…

> What I really love is the unbounded optimism of the AI. How many colleagues could you ask "Can you write me a PoC for a horizontally scalable, probabilistic database in Rust, using Raft-based consensus?" that would then just think for a second and go "Certainly! Let's start by ..."

I agree about the unbounded optimism. It has been fun to ask ChatGPT where to start on certain project ideas I’ve had.

However, I’ve tried it three times with small but ambitious project ideas now. Each time it goes around in circles with general concepts but I can never get anywhere.

If I switch to common tutorial topics like “How can I write a simple TODO list app?” it performs surprisingly well.

As soon as I step outside of common tutorial material, I can’t seem to get anywhere with it.

Re: AI-enhanced development makes me more ambitious with my projects

#388
post #335
post #205

Earlier quoted context omitted.

I cannot step in one of those machines without wondering if there is a hardware cutoff.

I went with a mechanic in a chair lift once. He was very nervous, saying he knows the kind of people who maintain these things, and he knows the kinds of things that can go wrong. I feel the same way about automation, whether it be self driving cars or an X-ray machine. I know the kind of people that program them, and the vast number of ways things can go wrong.

I work in the medical device field. Funny you should say that because just yesterday I was looking at some really crappy (but still safe, mind you!) code and wondering how I'd feel if I were being wheeled into the OR, knowing "how the sausage is made."

Re: AI-enhanced development makes me more ambitious with my projects

#389
I checked out the archive and came across https://simon.datasette.cloud/data/chatgpt_public_messages?_...

I’m wondering if the archive missed something or what this random statement from the system is about. It seems to be telling itself who it is and setting a restriction on what it knows. I know if it’s not trained on data past a certain point, it cant know it, but still i wonder if you could intercept that and change the cutoff date (if even possible), would that have any effect on anything. But then again, I am probably reading to much into it because it seems to be coming from system.

2023-03-26 04:10:49 - [system] : You are ChatGPT, a large language model trained by OpenAI. Knowledge cutoff: 2021-09 Current date: 2023-03-25

Re: AI-enhanced development makes me more ambitious with my projects

#390

I can definitely relate to your experience. Just last week, I leveraged AI tools to develop some data processing code. Estimating how long it would have taken me to write that code without these tools is a challenge because, to be honest, a year ago I wouldn't have even considered attempting such a project. It might have required a week or even two of my time, which I simply cannot spare for this kind of undertaking.…

I've noticed it as well. The most recent example I can think was with gitlab webhooks. I need to manage (create/delete/edit) a lot of web hooks across a lot of different projects. I normally would have probably just started doing it by hand in the web ui, but instead I had some python code from gpt to automate it in a few minutes.
Post reply on HN