Hacker News Reader

Comment by numbsafari | Hacker News Reader

No surprise that this doesn’t include any consideration for the profound moral, ethical, or social implications of AI and AI solutions.

Nothing about bias in models. Nothing about communication. Nothing about privacy. Nothing about security. Nothing about resiliency. Nothing about the responsibility of the individual practitioner. Nothing about the economic or ecological impact of the work.

Every undergraduate degree should include both general, and subject specific courses on ethics, morals, social, economic and ecological impacts.

View on HN

For comparison, another poster mentioned a real AI course over at https://news.ycombinator.com/item?id=23322449 which does include this.

There is a tendency for articles by extremely talented, highly educated, yet relatively immature people to completely over-optimize for technical depth. It's a failure with many articles that get posted here. I wonder how the author might change their article after 10 more years of experience.

I don't know. ...it's a bit sci-fi to add ethical instruction to AI dev. We are still way way distant from that.

Are there ethical classes for Organic Chemistry majors? Industrial engineers? Physics?

> I don't know. ...it's a bit sci-fi to add ethical instruction to AI dev. We are still way way distant from that.

I don't know what you mean. Can you elaborate?

> Are there ethical classes for Organic Chemistry majors? Industrial engineers? Physics?

Good question. We do have these, although the application varies. Engineering disciplines probably have this more explicitly embedded from direct ethics-type classes through to insertion of failure case studies and the like. These degrees tend to align with specific professional organisations for mechanical engineering, construction, etc. Membership of these organisations may require specific courses, but most often this is pushed down to the educational institutes. Computer Science and related areas have tended to follow this direction.

For someone studying chemistry or physics, there is less of a direct link, but it's somewhat unusual for an ethics requirement to be missing from a syllabus. Someone studying these might move into medicine, finance, or go to a startup. There is quite a spread of requirements there, and the destination generally governs the these in a more organic way. If you took this at the purest level and the chemist moved into a chemistry job at a pharma company, then membership of something like the American Chemistry Society might be expected. This bring a specific set of requirements.

In any case, it is only a certain segment of the computing world which eschews ethics and professionalism. A large segment interacts with all of these other professions which do place higher value ethics and enforce to a higher degree. Even working internationally, you'll find that yourself bound by a code of ethics in your country, but limited by law in others.

> Are there ethical classes for Organic Chemistry majors? Industrial engineers? Physics?

Yes, yes there are.

"a bit sci-fi"? What does that mean? We allow AI systems to decide who goes to jail in some places - I'd say that's a clear enough sign that the field needs people thinking all the way through their work.

There's some good entry level books on the topic, like "Weapons of Math Destruction" that you might enjoy.

Some universities do have a mandatory ethics course, so yes? Not all of course, or maybe not even most. But that's not a good argument "x and y don't follow the law anyways why should we"

It varies. It's definitely more explicit in my experience of UK and EU systems. In the US, it's not always as explicit, but sometimes approach this in other ways, cf. Honor Councils. Computer science and related areas can be perceived as immature, and this conversation is one indication of this.

I seem to recall humanities courses once having served the purpose.

> No surprise that this doesn’t include any consideration for the profound moral, ethical, or social implications of AI and AI solutions.

Would you say the same thing for an undergrad in SE (actually asking)? I feel like the hype train for AI has moved it into a plane where it has to be some moral arbiter but it seems to me that the same demands could be placed on most fields of study including general software engineering.

> Nothing about bias in models. Nothing about communication. Nothing about privacy. Nothing about security. Nothing about resiliency. Nothing about the responsibility of the individual practitioner. Nothing about the economic or ecological impact of the work.

Maybe it is just my bubble but I feel like very few AI jobs actually deal with hot button ethics (facial recognition for dubious purposes etc). It is just that these fringes get all the attention as if they represent the broad community (they don’t). For instance I work in AI for satellite communications, most of the things you bring up don’t really apply (any more than they do to the software engineers I work with). My former work was in AI and computer vision in marine ecosystems. Again... your moral questions don’t really apply (privacy of coral reefs?). I agree it is important to be aware of these issues but most of the real jobs in AI I feel like are more scientific in nature.. not the privacy busting, bias inducing ones we hear about a lot. Again that could be my bubble to some degree I suppose though.

I absolutely do. It's why I was so quick to jump on that with this.

My personal belief is that every undergraduate degree should include these subjects. That's what separates an undergraduate degree from a bootcamp--in theory.

But I think it's especially important for engineers and scientists, because it is very often overlooked or discounted by virtue of the fact that there's so much "technical" stuff to get into.

I think schools should try to counteract that by producing more team taught material. e.g., let's not just do a regularly philosophy course, let's do a philosophy for engineering and science course that takes you through the same general philosophy, but with readings that focus on engineering and science related ideas and themes.

But we should also have courses, and woven into our courses, discussions of these larger issues. Engineers, especially, have such a profound impact on society. It's important for us to be trained to think about these things. Not what to think, but that it is necessary to do so, and perhaps some tools and reference points for being able to do so in a cogent manner.

Lastly, I think it's important for undergraduates to get some exposure to the ethical issues related to work: the relationship between the employee and employer, the intellectual property, integrity in science and methods, etc.

"It's important for us to be trained to think about these things."

This doesn't seem self evident to me.

I think courses covering professional ethics are now fairly common in CS degrees. For example, here is the current course at Edinburgh: http://www.drps.ed.ac.uk/20-21/dpt/cxinfr10022.htm

I got my BS in CS 13 years ago and I had to take a dedicated ethics course and also several of my other courses had ethics components. I don't know how common it is but certainly many colleges/universities have been doing it for quite some time.

I happen to have taken an ethics course but it wasn't at all part of my CS/AI education. It was just a humanities course in my undergrad that fulfilled some philosophy requirement. Did your CS program actually require ethics directly? If so that is rather interesting.

I've taken multiple ethics class as part of my degree, and to be honest, if you don't already understand morality, you need the course. But I don't believe that those courses really have the desired impact, cheaters will cheat. It is more like these courses are created to cover the institution against the risk of appearing to not be graduating ethical individuals, but they can't actually make moral people beyond catching people who break the honor code and expelling them.

Applied-ethics discussions and courses are not about making people ethical who might otherwise cheat; they're about making people aware of what the generally-acknowledged ethical concerns are in their field, so that they can reach better decisions.

(By way of example, the topic of bias in AI models would in fact be covered in any introduction to stats, including an extensive discussion of the well-known tradeoff between bias and variance.)

It should apply to every degree. AI isn't special in that regard.

A post that is of this level of quality shouldn’t be degraded because it left out a category that you find profoundly interesting.

This is a goldmine for people interested in the field of AI, what you do with your talents as an engineer is up to you - build weapons, build medical systems, build whatever you want.

Your moral compass isn’t going to be set by an extra class in your curriculum, it’s set by your life experience.

I'm really skeptical as to the value of these courses. I agree absolutely that ethics/morals/ etc. are essential things to learn and develop. I don't agree they can really be learnt in the classroom. Many of these things (e.g. ethics) are emotional intelligence skills that don't really respond to standard study. In engineering school I did all these courses and didn't get any benefit from them, other than some academic knowledge of Kantian vs Utilitarian ethics.

CS grads should be required to take a year of liberal arts at least.

wrt ethics/morals; Is any business gonna pay a SWE for that, or are they gonna pay a lawyer?

University isnt about only learning what is useful for your employer. Its to you get people who hopefully do not end up building software with racist biases build in, or hardware that refuses to work if you do not have light skin.

Are you talking about any software, or computer-vision specifically? What does a web dev need to learn about racism in software?

Possibly that people from underprivileged backgrounds are unlikely to have high speed and reliable network services, that using image mapping makes a website unusable by the vision impaired, and that making a website such that it's layout changes after loading can prevent navigation by systems for people with physical disabilities?

While in most forms of engineering, a lack of ethics can demonstrably lead to death (and AI is no exception when you look at proposed use cases in military and law enforcement) even web developers work on things that are for the public good.

Accessibility isn't an ethics topic. Planes falling out of the sky is bad ethically, but when engineers study aerodynamics, they are not studying ethics. No one is suggesting engineers must attend an ethics course to prevent planes falling out of the sky - that's what we train lawyers for - to sue the shit out of those responsible for letting that happen.

A web dev isnt' going to think about "underprivileged backgrounds" etc - that's up to whoever decides who their demographic is. If you are building a website to sell Teslas, do you expect the dev to stamp their feet and demand a low-fi version on behalf of the underprivileged? Whoever is paying for development, it's their decision what to pay for.

"learning what is useful for your employer" is exactly what university is about, is exactly what you pay for, and is exactly what is expected.

It used to be the case that a "professional" was someone who cared about things at industry scope, not just their employer. SWEs are still considered cost centres, have no licence to revoke if they behave "unethically", and no bar, or association to make such a judgement anyway. Without a licence as a barrier to the profession, any uncooperative dev would find themselves outsourced for the next dev who doesn't give a toss.

Doctor and lawyers where always the benchmark for these kind of things, and it's arguable that even those professions have become corrupted by corporate influence.

First off, Ethics course requirements are incredibly common in engineering studies and I've never heard someone accuse lawyers of enforcing ethics. They enforce law and contracts, and are inherently reactive. The existence of lawyers has certainly not prevented a variety of engineering disasters that could have been avoided if the engineer had chosen to act differently.

Now, is it an ethical dilemma if you are asked to write firmware that disables a redundant safety sensor unless the customer pays additional fees? Is it a ethical to use a dataset that you may know is not fully representative for a crime estimation algorithm in order to achieve cost targets for your contract? Is it ethical to write a software that changes device behavior when being examined by regulators? Is it ethical to design an interface for a public service that will be inaccessible to a portion of the population to meet contract deadlines?

>learning what is useful for your employer" is exactly what university is about, is exactly what you pay for, and is exactly what is expected This is a warped viewpoint that assumes you have no agency. I studied engineering because I wanted to learn how it worked and how to build things. Employers be damned, I have no issue working for myself if I cannot find an employer who I can come to terms with.

If you mean to argue that the standing of engineering as a profession has greatly fallen then I agree with you. Especially in software. However, even if there is no license to be revoked, and no personal feeling of responsibility, the engineer who makes these design choices is not shielded from the legal repercussion of his actions no matter how many times his supervisor told him to even assuming he is willing to be responsible. No matter how much you may wish otherwise, engineers are still considered to hold the public trust and expected to behave accordingly. And you should consider it because when you comply with your employer to do something unethical and shit hits the fan, they are going to put you immediately in front of the firing squad to shield themselves from as many bullets as possible just the same as they are going to take the reward for as long as it flies under the radar.

How else do you enforce ethic other than reactively? Until we have minority-report style pre-crime reporting, lawyers decide who is punished on ethical grounds.

> Now, is it an ethical dilemma if you are asked to write firmware that disables a redundant safety sensor unless the customer pays additional fees?

Not for the engineers. They get a spec, they don't write it. Is it unethical to stab someone on the street? yes, but not for the knife manufacturer. Beyond that, it's for lawyers to decide the laws, and how they are enforced e.g. limits of who knifes can be sold to, and punishments for non compliance. It's likely the above scenario doesn't, because it's illegal, not because SWEs refused to do it on any other grounds. If it isn't illegal, they'll just find someone else to do it.

> engineer who makes these design choices is not shielded from the legal repercussion of his actions no matter how many times his supervisor told him to even assuming he is willing to be responsible

There has to be a law to break. If you are assuming everything unethical is illegal, why bother with ethics - just give engineers training in the law. Or, don't bother, and instead run everything by a lawyer.

Reply on news.ycombinator.com