From a blog post by @wojtekpow

[wojtekpow@mastodon.social]

@rasterweb fair point, I’ll make a correction.
Separate from the article though, which talks about small bespoke apps. Everyone ends up at least using AI software in their daily life. Pretty much all of corporate software at this point has some AI-written code in it, and that percentage will only increase with time.

[tsturm@famichiki.jp]

@wojtekpow @rasterweb Not all. While some departments in my (very large) workplace might be using AI for their code, I do not. I'm working on very specific proprietary solutions and the very last thing I want to do is have an AI get as much as a whiff of how our solutions work.

[rasterweb@mastodon.social]

@wojtekpow I'm not sure everyone ends up using AI software in their daily life.

For those of us who actively avoid it that might not be the case. I have deleted and found replacements for software that added AI features and will continue to do so.

[rasterweb@mastodon.social]

@wojtekpow The small bespoke apps, yes... I am all for that. I've made many myself, they just don't have AI in any way.

[downey@floss.social]

@wojtekpow

There is no alternative - Wikipedia

(en.wikipedia.org)

@rasterweb

[wojtekpow@mastodon.social]

@rasterweb I don’t mean apps that add AI features in UX, I mean apps coded with the help of AI. Even the Linux kernel accepts AI code. I just don’t think you’ll be able to say that you don’t use any AI-written code in your life. To me, AI in code writing is simply another abstraction shift. We moved from machine language to Fortran, COBOL to C and C++ to Objective-C and Swift, and scripting languages like Python. AI in my daily coding life feels like another step in this chain.

[robinadams@mathstodon.xyz]

@wojtekpow @rasterweb I make as much effort as possible to avoid using anything with AI-written code (switching to NetBSD now that the Linux kernel is allowing AI-written code, using Vivaldi browser, etc.) In general for each app type there's at least one alternative suitable for us "AI vegans". There may not be many of us but I know I'm not unique.

[mathieui@piaille.fr]

@wojtekpow @rasterweb I won't cover the arguments against LLM-driven development since other people have already voiced their objections, but one of the weak points of the article for me is that, if LLM boosters say the truth, then instead of being secure through irrelevance, those apps will be vulnerable due to automated script kiddies, much like the LLM crawlers are hammering the open web right now.

The warnings against skimming on security also ring a bit hollow, as the promise is for normal people to develop apps, which means they are never going to bother.

[wojtekpow@mastodon.social]

@mathieui @rasterweb true but as LLMs are getting better at finding bugs they also get better at identifying bugs in own code as more of the current harnesses automate security reviews of written code

[mccovican@infosec.exchange]

@wojtekpow @rasterweb

To me, AI in code writing is simply another abstraction shift.

And therein lies the problem. It absolutely is not. But it's been successfully normalised as such. Because people keep saying things like "AI in code writing is simply another abstraction shift".

The shift from Fortran to Python plainly did not involve such obviously abhorrent long-term systemic costs, and to suggest otherwise falls somewhere between cognitive dissonance and intellectual dishonesty (dealer's choice).

We're long past the point where the costs of LLMs are some new & surprising revelation, and where folk can make statements like that then sulk behind the thin veil of ignorance when called out.

I respect Pete's measured manner in this interaction, but to be blunt, by coming out with phrases like that, you are laying the ground and carrying water for those who would extract every last drop of that same water from us (among other resources). It is not a neutral standpoint.

[cmthiede@social.vivaldi.net]

@rasterweb @wojtekpow The quintessential yeah but: Everyone is using it whether they like it or not, somewhere, somehow (banking, healthcare, customer support, etc) unless you're the Unabomber type. Is that what you are, anti-fascist? The government has rules against that now you know I'd rather die penniless helping turtles cross a busy road than work for a giant human culture eraser just to sip FU money from a tech lord's straw.

[poprox@social.lol]

@rasterweb @wojtekpow my tech stack is ai-free and I never prompt LLMs, never. Not for anything. I don't use it, at all. all of my coding is completely "ai"-free. this is in alignment with my values. I harshly judge all those who choose to use the fascist tech. and the moment a "developer" uses "ai" "tools", they become an untrustworthy plagiarizer whose slop code I would never trust running on my devices. don't trust anyone or anything trying to _sell_ "AI". It's snake oil and they are relying on sowing illiteracy in order to make the sale.