Are soft skills like ethics important for a career in AI?

Question

Do I actually need to worry about soft skills and ethics stuff if I just want to be a dev in AI or is that just corporate fluff? I am finishing up my masters in computer science in about 4 months here in Seattle and I have been stressing out big time about my portfolio and what recruiters actually want. I have spent all my time grinding LeetCode and learning the deep math behind transformers and neural networks because I thought that was the whole point but now I am seeing all these job postings at places like Anthropic and even smaller startups that list things like AI alignment and ethical frameworks and even communication skills for non-technical stakeholders as major requirements.

It is honestly making me super anxious because my program barely touched on any of that and I feel like I am falling behind before I even start. I read this long article on Medium that said ethics is going to be the new coding for AI because of all the bias issues and safety concerns but then I see people on Discord and Reddit saying that companies just hire specialized philosophers or ethicists for that and the actual engineers just need to make the model fast and accurate so it doesnt really matter. I am stuck in the middle and dont know who to believe. Am I supposed to be reading philosophy between my Python sessions?

I have got about $1,300 left in my savings to last me until I graduate in June and I really cannot afford to waste time or money on the wrong certifications or classes if the industry does not actually care about the soft side of things. I am looking for entry-level ML engineer roles but every single interview prep guide tells me something different. Like, will they actually ask me about trolley problems in a technical interview? I feel like I am drowning in all these different expectations and I am scared I am gonna fail my technicals if I focus too much on the vibe of the AI but then if I ignore it will they think I am some kind of robot who does not care about the consequences of what I am building? I just want to build cool stuff but everything feels so much more complicated now. Is it actually worth the pivot or should I just stick to the hard math and hope for the best...

PeakDistrictClimber · Accepted Answer

Honestly, seeing someone care about ethics this early is amazing! You have to realize that in the real world, a fast model that hallucinates or shows bias is basically useless for a serious business. It is fantastic that you are looking at this now because safety and alignment are what make AI reliable and trustworthy for actual users. I have seen so many projects fail because the devs ignored the human element and ended up with a massive PR disaster or a lawsuit. You definitely dont need to become a full-time philosopher, but you should absolutely understand how to audit your datasets for bias. It is not just fluff; it is about engineering quality and long-term stability! I would totally recommend picking up O'Reilly Ethics of AI and Data Science 1st Edition because it bridges that gap between pure math and societal impact perfectly. Also, since you are on a budget, check out the Coursera DeepLearning.ai AI For Everyone Course which is super affordable and gives a great high-level view of why these things matter to stakeholders. If you want something deeper, NYU Press Algorithms of Oppression 1st Edition is a must-read for any ML engineer who wants to build systems that actually work for everyone. Focus on showing you can document your model's limitations in your portfolio. Recruiters love that level of professionalism! It shows you are not just a code monkey but a responsible engineer who thinks about risks. Keep grinding those transformers, but definitely spend some time on the safety side too. It is such a game changer!

JamesDox · Answer

TL;DR: Ethics is basically just high-level QA. Companies care because bad models lose money. Stick to the free stuff and dont blow your savings. Jumping in here because i have seen this play out first hand. Over the years, i have worked on plenty of models where we thought the math was bulletproof, but the implementation was a disaster because we ignored the social context. I once saw a facial recognition project get scrapped after six months of dev work because the team didnt account for bias in the training set. It was a total waste of compute and salary. It wasnt that they werent smart, they just werent looking at the right metrics. In my experience, startups like Anthropic want people who understand that alignment is a technical challenge, not just a vibe. You dont need to be a philosopher, but you do need to know how to build guardrails. Since you are tight on cash, please dont spend that $1,300 on certs. You can literally just read the documentation for Google Cloud Responsible AI Practices online for free. If you want something more structured, look for a used copy of Manning Publications Machine Learning Engineering in Action or check out the Microsoft Azure AI Fundamentals Exam AI-900 material. Most of the ethical frameworks recruiters talk about are actually about data hygiene and model robustness. Stay focused on the technical side of how to measure bias and you will be fine. Dont pivot, just broaden the scope of what you consider to be engineering.

Randallgoalf · Answer

Over the years, Ive seen that ethics is basically just system safety. Get Wiley Trustworthy AI 1st Edition. Recruiters want robust code that wont break in production.