M. Grégoire4d ago

Do you worry that LLMs may take your job?

#AI

No
65.2%
Yes
17.4%
Other
17.4%
Poll ended at .
Show threadM. Grégoire13h ago

There's a great deal of opposition to #AI here on the Fediverse, to the point that there are people who don't want to run software that has had any contact with #LLMs. (For a rational set of arguments, see https://codeberg.org/small-hack/open-slopware#why-not-llms .)

In the business of persuading people, you learn that it's important to understand not just what they believe, but *why* those beliefs appeal to them, what is the underlying emotional attachment. In this matter, I suspect it's because programmers fear job loss.

1/3

open-slopware

Free/Open Source Software choosing to use and/or support LLM usage/AI, as well as alternatives and tips to requesting better policies or forking.

Codeberg.org
Show threadM. Grégoire

It appears to me that LLMs are already capable of taking the jobs of many computer programmers — if I were in that line of business, I'd be very concerned, possibly even hostile, and the reasons to oppose AI would seem compelling.

I don't think the results of this poll provide much evidence either for or against my theory.

2/3

12:06pmWeb
Show threadM. Grégoire13h ago
Personally, as a structural engineer, I worry that AI will be used in structural design and we'll have collapses; but I don't worry about it taking my job. The whole business of #engineering design is based on the idea that there is an engineer who takes personal responsibility and liability for the design. The creators of even comprehensible software always disclaim all liability, so there's no chance LLM providers will accept liability. LLMs will aid but not replace structural engineers.
3/3
Show threadDavid Megginson12h ago

@mpjgregoire wrote

Personally, as a structural engineer, I worry that AI will be used in structural design and we'll have collapses; but I don't worry about it taking my job.

AI: 😁 Here are the blueprints for your rail bridge.

You: Some crows just landed on it, and it collapsed.

…thinking…

AI: 😁 Oops, it looks like I mixed up milligrams with kilograms! Would you like me to help you find YouTube videos about Jeff Bridges?

#AI #genAI #engineering

Show threadM. Grégoire12h ago

@david_megginson 19 workers died because Vancouver's Second Narrows Bridge collapsed during construction: https://en.wikipedia.org/wiki/Ironworkers_Memorial_Second_Narrows_Crossing

The error, I've been told, is that a junior engineer calculated the shear capacity of certain supports based upon the total cross-section of the beams, rather than upon the area of the webs. Humans make errors, but at least we try to be correct, we can be punished when we err, we understand that words and equations have real meanings — they're not merely symbols.

Ironworkers Memorial Second Narrows Crossing - Wikipedia

Show threadRuth O'Day12h ago

@mpjgregoire

Not a programmer, but what I see as their fear is not job loss, but job loss to inferior product.

On the engineering side, what if some genius MBA tells you to design using AI then simply put your stamp on the final drawing? I expect that you are very good at your job but that there is no way that you can double check faulty calculations faster than AI can generate them.

We are probably not at that stage because of liability issues. But all it will take is for the MBA's and other assorted capitalists to rig the game to get enough bonus and avoid liability by ensuring that it falls on the engineers. As the MBA president of one country says: That makes them smart.

Show threadM. Grégoire12h ago
@RuthODay2 In my industry, that's not a new risk created by LLMs; there's long been a danger of engineers being told to seal work based on fancy but inadequately checked software, people you haven't supervised, etc. The code of ethics tells us not to do so, and we can be punished for breaking the code, even if there were no dire consequences in a given case.
Show threadRuth O'Day11h ago

@mpjgregoire

There are punishments for breaking the code, true, but traditionally with little upside.

If we take G Gordon Liddy's maxim ("of course crime pays, otherwise there'd be no crime") then AI changes the dynamics: the finance bros can divert some of the money "saved" by AI design to pay an engineer who will put their stamp on it.

Engineers don't have a moral purity, there has just been a decent salary ("my fair wage I openly take") to make it worth adhering to the code. The AI math changes that.

Show threadM. Grégoire10h ago
@RuthODay2 Why does the AI math change this any more than CSiBridge (for instance) does? There are
many software tools for structural engineering, and AI at present doesn't even seem like it's among the most useful.
Show threadRuth O'Day8h ago

@mpjgregoire

Traditionally there is a certain cost for engineers to envision, design, self-review, and review by someone else.

As you said, AI is not the most useful, but a fiance bro can use it, as opposed to tools designed for engineers.

So compare the cost of traditional engineering to the cost of a fin bro telling AI to do all that work at once. There's a lot of money on the table that the engineers were getting that the fin bro can pocket, at the cost of giving only one engineer a small sliver of to put a stamp on all of it.

I am sure that you have seen or learned of engineers that put their stamp on something that they shouldn't have in the traditional engineering scenario, IMO that gets worse when a fin bro frees up a bunch of money.

Show threadRob Johnston 🍁11h ago
@mpjgregoire I'm not worried about AI taking my job. I see it as making software development MORE professional, for the reason that there are a lot of decisions and trade-offs in the design and architecture of applications. One has to know this stuff. AI is taking the part of the job that involves typing.
Show threadM. Grégoire10h ago
@rjohnston Do you think many software developers share your opinion? I don't think it's very popular on the Fediverse, though of course the loudest voices are often unrepresentative.
Show threadRob Johnston 🍁10h ago

@mpjgregoire I do follow some accounts here that are pro-AI, but I agree that there are a lot more accounts that are spitting venom about it.

It's natural to fear the unknown, or to complain when you're being told that things are going to change.

Show threadJean-François Mezei12h ago
@mpjgregoire LLM can write genetic programs. But when a company’s value is in its logic (for instance insurance measuring risk, or rail operating trains, or airline planning of fleet and crée schedules , your LLM had no expertise in you industry’s specific needs. Abd the code it generates may not be maintainable by humans.