Comment by extr | Hacker News Reader

Comment by extr | Hacker News Reader

Seems about right. My official title at work is "AI Engineer". What does that mean exactly?

- I'm not a researcher and not fine tuning or deploying models on GPUs

- I have a math/traditional ML background, but my explanation of how transformers, tokenizers, etc work would be hand-wavy at best.

- I'm a "regular engineer" in the sense I'm following many of the standard SWE/SDLC practices in my org.

- I'm exclusively focused on building AI features for our product, I wear a PM hat too.

- I'm pretty tuned in to the latest model releases and capabilities of frontier models, and consider being able to articulate that information part of my job.

- I also use AI heavily to produce code, which is helpfully a pretty good way to get a sense for model capabilities.

Do I deserve a special job title...maybe? I think there's definitely an argument that "AI Engineering" really isn't a special thing, and considering how much of my day to day is pure integration work with the actual product, I can see that. OTOH, part of my job and my value at work is very product based. I pay a lot of attention to what other people in the industry are doing, new model releases, and how others are building things, since it's such a new area and there's no "standard playbook" yet for many things.

I actually quite enjoy it since there's a ton of opportunity to be creative. When AI first started becoming big I thought about doing the other direction - leveraging my math/ML background to get deeper into GPUs and MLOps/research-lite kind of work. Instead I went in a more producty direction, which I don't regret yet.

The author’s definitions suggest you should have “Applied” in your title, which I like, but my impression is that “applied” roles so vastly outnumber “creation of models” roles globally that it’s actually the latter that would benefit from a modifier. For now, you have to rely on context (mostly the nature of the company’s primary output) when trying to interpret something like a job posting or an acquaintance’s title.

It’s not that crazy to add a couple of domain-specific prediction heads to a BERT-family pretrained model and then do a quick fine tuning. By volume that’s less common but I would guess most people are just using things off the shelf and might not even consider themselves AI engineers. I have no frame of reference though.

We all know the AI part is largely meaningless because of the hype and nonsense, but what defines you as an engineer? When you consider that classical engineers are responsible for the correctness of their work, combining it with AI seems like a joke

Hard to tell what you're even trying to say here. I am obviously responsible for the correctness of my work. "AI Engineer" does not generally mean "AI-Assisted Engineer", thought that was clear from my post.

> "When you consider that classical engineers are responsible for the correctness of their work"

Woah hang on, I think this betrays a severe misunderstanding of what engineers do.

FWIW I was trained as a classical engineer (mechanical), but pretty much just write code these days. But I did have a past life as a not-SWE.

Most classical engineering fields deal with probabilistic system components all of the time. In fact I'd go as far as to say that inability to deal with probabilistic components is disqualifying from many engineering endeavors.

Process engineers for example have to account for human error rates. On a given production line with humans in a loop, the operators will sometimes screw up. Designing systems to detect these errors (which are highly probabilistic!), mitigate them, and reduce the occurrence rates of such errors is a huge part of the job.

Likewise even for regular mechanical engineers, there are probabilistic variances in manufacturing tolerances. Your specs are always given with confidence intervals (this metal sheet is 1mm thick +- 0.05mm) because of this. All of the designs you work on specifically account for this (hence safety margins!). The ways in which these probabilities combine and interact is a serious field of study.

Software engineering is unlike traditional engineering disciplines in that for most of its lifetime it's had the luxury of purely deterministic expectations. This is not true in nearly every other type of engineering.

If anything the advent of ML has introduced this element to software, and the ability to actually work with probabilistic outcomes is what separates those who are serious about this stuff vs. demoware hot air blowers.

Nicely said, I'm going to borrow some language here. I've talked a little to my coworkers about how it's possible the future of SWE looks more like "build a complex system with AI and test it to death to make sure it fits inside the performance envelope you require".

I will be thinking about this comment for a bit. Thanks for this perspective!

This comment is excellent.

You're right in a descriptive manner, but I also think the parent comment's point is about correctness and not determinism.

In other engineering fields correctness-related-guarantees can often be phrased in probabilistic ways, e.g. "This bridge will withstand a 10-year flood event but not a 100-year flood event", but underneath those guarantees are hard deterministic load estimates with appropriate error margins.

And I think that's where the core disagreement between you and the parent comment lies. I think they're trying to say AI generated code-pushers are often getting fuzzy on speccing out the behavior guarantees of their own software. In some ways the software industry has _always_ been bad at this, despite working with deterministic math, surprise software bugs are plentiful, but vibe-coding takes this to another level.

(This is my best-case charitable understanding of what they're saying, but also happens to be where I stand)

> "I think they're trying to say AI generated code-pushers are often getting fuzzy on speccing out the behavior guarantees of their own software."

I agree, and I think that's the root of the years-long argument of whether programmers are "real" engineers, where "real engineering" implies a level of rigor about the existence of and adherence to specifications.

My take on this is though that this unseriousness really has little to with AI and entirely to do with the longstanding culture of software generally. In fact I'd go as far as to say that pre-LLM ML was better about this than the rest of the industry at-large.

I've had the good fortune to be working in this realm since before LLMs became the buzzword - most ML teams had well-quantified model behaviors! They knew their precision and recall! You kind of had to, because it was very hard to get models to do what you wanted, plus companies involved in this space generally cared about outcomes.

Then we got LLMs, when you can superficially produce really impressive results easily, and the dominance of vibes over results. I can't stand it either, and mostly am just waiting for most of these things to go bust so we can go back to probabilistic systems where we give a shit about quantification.

I agree with that.

I think part of the issue with the lack of "real" quantification in the results of LLMs is that the output and problem domain is so ill-defined. With standard neural nets (and other kinds of ML) classifiers, regression models and reinforcement models all had very narrow, domain specific problems they were solving. It was a no-brainer to measure directly how your vision classifier performs against a radiologist in determining whether an image corresponds to lung cancer.

Now we've opened up the output to a wide variety of open-ended domains: natural languages, programming languages, images and videos. Since the output domain is inherently subjective, it's hard to get a good handle on their usefulness, let alone getting people to agree on that. Hence the never-ending discourse around them.

This seems to me patently absurd, because LLMs are not part of the probabilistic environment of the domain you're engineering; rather, you're injecting new probabilistic inputs into your system. That seems to me to be a wholly different category, and wildly misrepresents how an engineer is supposed to operate and think.

> "because LLMs are not part of the probabilistic environment of the domain you're engineering; rather, you're injecting new probabilistic inputs into your system"

You do this as a process engineer also. You don't have to have a human operator inserting the stator into the motor housing, you could have a robot do it (it would cost a lot more) and be a lot more deterministic.

After the stator is in the housing you don't need to have a human operator close it using a hand tool. You could do it robotically in which case the odds of failure are much lower. That also costs a lot.

You choose to insert probabilistic components into the system because you've evaluated the tradeoffs around it and decided it's worth it.

Likewise you could do sentiment analysis of a restaurant review in a non-probabilistic manner - there are many options! But you choose a probabilistic ML model because it does a better job overall and you've evaluated the failure modes.

These things really aren't that different.

Who cares? The word "engineer" is meaningless now and anyone can be a self-proclaimed engineer whenever they feel like it. Will anyone double check or even reject you for an engineering job when you are not? Absolutely not! Take a bootcamp, submit plenty of PRs correcting typos, and pass the interview with the help of AI and you basically made it, dreams come true!!

Yeah the term "engineer" has been diluted into oblivion, and we only have ourselves to blame for not protecting it.

In German you're not even an engineer if you don't sometimes wear a hard hat or hold a screwdriver.

Agree 100%, even blue collar workers guard their profession. Hell, I was talking to a friend and they rejected her for a retail job because she had never worked in retail before. Engineering on the other hand has zero gatekeeping - it's a sign spinner job right now. Just do a few humiliation rituals like daily standup and you're the perfect candidate!

> Just do a few cleansing rituals like daily standup and you're the perfect candidate!

There, FTFY

In Dubai, the poor underpaid folks cleaning the roads and gutters late at night are called "Cleaning Engineers" and "Garden Engineers". It's honestly sad, almost a mockery.

In the US, we've had "sanitation engineer" as the euphemistic neologism for "worker paid to pick up your garbage bins" for 50(?) years.

protecting it? ha! we’re just the first group of greater fools who thought it applied to us in the first place (hell, i became an “engineer” with an Associates degree!). just because we benefited from the prestige doesn’t always mean we’re actually held to the classical standards of engineers.

> I pay a lot of attention to what other people in the industry are doing, new model releases, and how others are building things,

what do you think of recent MIT news that 95% gen ai projects don't do anything valuable at all ?

Sounds kind of aggressive but probably the number is up there.

> what do you think of recent MIT news that 95% gen ai projects don't do anything valuable at all ?

Worth noting that a project that ends up “doing nothing” isn’t the same as a project that had/created no value.

Even some projects that in hindsight were deterministic lemons.

Assuming compute resources continue scaling up, and architectures keep improving, AI change now has an everything, everywhere, all the time, scope. Failing fast is necessarily going to have a substantial dimension.

Not sure the majority of projects do anything valuable at all.

You would need to compare that to the baseline value creation from non-AI projects.

Reply on news.ycombinator.com