Fediverse & AI Coding Tools & Vibe Coding

[smallcircles@social.coop]

@julian @reiver @fedicat @johannab

> The problem is that some people use AI as a shortcut to no longer think.

The issue is bigger than that, I think. One needs expertise to work with AI responsibly, and recognize where it goes off the rails. In our own expert areas we invested the time to be able to correct AI output.

In all other areas, where we'll have to use hastily-introduced AI, we too are unable to detect its flaws. While AI poses as expert on anything, and is integrated everywhere.

[smallcircles@social.coop]

@julian @reiver @fedicat @johannab

Furthermore I guess the killer app for AI is "Autoritarianism". Allow a small group of bad actors to control entire populations, by letting AI wade through the wealth of information collected via surveillance capitalism, weeding out patterns of resistance well before it becomes any serious threat. And a dictator doesn't care about that 5% error margin, where the AI got it wrong.

[smallcircles@social.coop]

@julian @reiver @fedicat @johannab

Need only look at Palantir porting the operations of the US government to AI systems, and similar corporations (entire supply lines), to see this worrying trend.

[tchambers@indieweb.social]

@reiver @dansup
We should embrace any tools that respect creators privacy and produce safe code and that can enable new creative “minimal viable products” apps to ship even if they are likely to need to mature using more traditional programming later.

[khzimmer2@literatur.social]

@reiver
Your poll is biased.

You're asking people to explain why they do NOT want to use such (immoral and energy-intensive) tools.

But you're NOT asking for reasons why people WANT to use such (immoral and energy-intensive) tools.

Nice try! 😎

#no2ai

[oddpixel@mas.to]

@reiver nope. I have been cleaning up sloppy "vibe coding" for a month straight on a contract. Security, scalability, and code clarity are impacted by this approach.

[thenexusofprivacy@infosec.exchange]

Thanks, I hadn't read that ... but I wasn't thinking about any one particular incident, I've heard similar things from a lot of people.

I know where I am on this, but still curious to see the results of the poll!

@onepict

[almino@ursal.zone]

@reiver @josemurilo I voted Yes! But let me explain my point of view.

I have been programming for over 15 years… but I am also an #AuDHD who was diagnosed less than five years ago. Also I've been involved in an accident that affected my whole left arm. I can still type with it, but too much of it results in long periods of pain.

Also, besides working with computers, I am a full-time nurse student.

There are so many things I want to code, but I simply don't have the time or energy to do it. Telling an AI to create a simple loop, a class or even a module boilerplate really helps me a lot!

So, please, do help me!

[mirth@mastodon.sdf.org]

@reiver Even if you want to help I would avoid doing anything specific to whatever practice happens to be the fad this week. Clear documentation, example code, conformance tests, etc, all benefit every developer.

[zotheca@mementomori.social]

@smallcircles @fedicat @reiver

The question is: Do people even know what they are doing? Can they guarantee and ensure reliability, security, data protection and smooth operation in a network?

[aka_quant_noir@hcommons.social]

@reiver

The answer is vigorous code review. Anything that connects to an API, anything that is federated, shared, made free, or made at all should be reviewed for intent, for mistakes, for data integrity, for documentation and for competence. I don't care who or what creates it. Without code review, it's so untrustworthy that it poisons the well.