Understanding Careers and Research in AI

I think creativity is an important skill to develop - AI as a field moves quickly, and being able to look at new developments and ask questions about them, brainstorm new approaches or ways to use them, is very helpful. This could be true in domains or in core AI development. After that, building new AI tools is easier when you have some programming and math skills. I would encourage people to especially improve the skills they're interested in developing and pursuing.

A career in AI involves both creating models and applying them, and the balance between these activities can vary greatly depending on your specific role and interests. Some insights that the the experts shared:

[Interconnectedness] Kai emphasizes that model creation and application are closely linked. Models are often motivated by specific applications, and applications can drive modifications to models. This cyclical relationship highlights that working in AI often involves both aspects.

[Theoretical vs. Applied Focus] Mercy identifies three areas: theoretical model development (focused on mathematical and statistical methods), model building, and model application. Your career path can focus more on theory or application based on what you enjoy and where you want to create impact. She highlights the importance of seeing real-world outcomes, which aligns with a more applied focus, while theoretical work often feeds into building models used in practical settings.

[Varied Proportions of Work] Sindhu suggests that within AI careers, the proportion of time spent on modeling versus applying those models can differ significantly between individuals and roles.  Your career can evolve over time, as exemplified by her transition from focusing solely on theoretical models during her Ph.D. to incorporating more practical applications later.

[Nature of Work] Research-focused AI work often emphasizes speculative, long-term projects, which can occur in both academic and industry settings. Applied AI work typically focuses on pragmatic, short-term problem-solving and is primarily found in industry.

[Academic vs. Industry Paths] Sindhu recommends evaluating whether you enjoy academic life, as pursuing a Ph.D. involves a long-term academic commitment, while a master's degree is usually sufficient for industry roles. Mercy notes that while industry roles can offer research-focused opportunities, especially in entities like Google Research, roles vary significantly across companies.

[Career Trajectory and Passion] All panelists emphasized the importance of considering where your passion lies. Sindhu mentioned you should explore roles that excite you by looking at people who are 5-10 years ahead in their careers.

[Getting Experience] Internships are universally recommended as they provide insight into both industry and research environments. They help determine if industry culture suits you, as Kai points out.

[Research in Industry] Mercy shares that industry research can be similar to academia, offering opportunities to publish and engage with the academic community, depending on the company. Kai adds that industry research can sometimes allow for ambitious projects but typically centers around tangible problem-solving with resources.

The current landscape of machine learning (ML) jobs in industry tends to favor applied roles over research roles, but both paths have distinct characteristics and requirements:

[Applied vs. Research Roles] Kai indicates that there are generally more opportunities in applied roles because the industry is heavily application-driven. These roles focus on implementing models, developing products, and addressing practical challenges. Sindhu and Kai both note that the boundary between research and applied roles isn't clear-cut. Applied roles often benefit from research skills to understand and adapt underlying algorithms, and vice versa.

[Career Path and Education] Kai highlights that pursuing a Ph.D. equips individuals to tackle ambiguous or undefined problems, which can be valuable in both research and applied contexts. Meanwhile, undergraduate and master's degrees often focus on execution skills, such as coding and implementing solutions. Mercy and Kai both emphasize that the choice between these educational paths should align with your career interests—whether you’re passionate about foundational research or driven by application and product development.

[Grad Degrees and Applied Roles] A graduate degree isn't strictly necessary for many applied ML roles but can provide a competitive edge and enhance problem-solving skills. Applied roles can greatly benefit from the ability to bring research insights into practical implementations. Mercy underscores that practical experience and applied skills are highly valued in the industry, especially for positions focused on bringing products to market.

[Job Market and Opportunities] The demand for applied skills is strong due to the industry's focus on product development and deployment. Mercy notes that entering the industry right after a bachelor's degree offers a straightforward path to gaining experience and advancing quickly in applied roles.

To navigate the job market successfully, it's beneficial to:

• Identify your interests and strengths, considering whether you're drawn to research or practical applications.

• Gain a blend of research and applied experience to enhance your versatility.

• Stay adaptable and be open to learning throughout your career, regardless of the path you choose. This adaptability will help you thrive in the dynamic AI landscape.

There are many exciting areas of AI research. I'm personally excited about directions in AI safety and deploying AI in collaboration with humans, in uncertain real-world settings. I would encourage people to try to explore recent conference proceedings (e.g., ICML, AAAI, NeurIPS) and look to see if any tracks, papers, tutorials, or workshops look exciting. People can also look to see who is involved in that research or tutorial and reach out to them with questions and ideas. Here are some other potential opportunities for getting involved: www.try-ai.org/resources

This is a fantastic question, with a long, sometimes complicated answer! In my own view, I believe that including people who may be impacted by AI systems from the start of the AI development process, and through to continued evaluation after deployment, is a key component of an effective strategy to ensure that their views and values are embedded within the system and to ensure that this remains the case throughout the process. A good resource to get started learning more is the Data Feminism textbook (https://data-feminism.mitpress.mit.edu/). Much research continues on these topics, and recent work can be found in conference venues such as FAccT (https://facctconference.org/), AIES (https://www.aies-conference.com/2025/), and EAAMO (https://conference.eaamo.org/).

[Diverse Teams and Roles] Socio-technical research teams are diverse, often including legal researchers, human-centered design experts, UX researchers, statisticians, and model developers. These roles focus on understanding and addressing the ethical and social implications of AI technologies. (Mercy)

[Application Beyond Tech Companies] AI has significant application potential beyond traditional tech companies. Fields such as healthcare, agriculture, and manufacturing still underuse AI and present vast opportunities for improvement and innovation. Small advancements in these sectors can lead to substantial social impacts, highlighting the importance of exploring AI's role in various societal segments. (Kai)

[Dual Perspective of AI's Impact] It’s crucial to consider both how AI can impact society positively and how existing AI models may affect society. This dual perspective allows you to either work on developing AI technologies that solve societal issues or research the societal implications of current AI models and address any negative impacts. (Sindhu)

Pathways to Consider

• Interdisciplinary Education: Consider pursuing degrees that combine technical AI skills with social sciences, ethics, law, or human-centered design.

• Specialized Roles: Explore careers that focus on AI policy, ethics, user experience research, or legal aspects related to technology.

• Industry and Academia Collaborations: Engage in partnerships or projects that involve diverse team compositions aimed at tackling socio-technical challenges.

• Impact-Driven Sectors: Seek opportunities in sectors like healthcare, agriculture, or education, where AI can significantly enhance processes and outcomes.

Choose a path that aligns with your interests in understanding or influencing the intersection of AI and society, and leverage your unique skills to contribute effectively to this evolving field.

[Networking and Conferences] Mercy emphasizes the importance of networking at conferences, engaging with affinity groups, and building connections with professionals in the field. Many opportunities arise through references and personal networks rather than formal job postings.

[Relevance to Current Trends] Familiarity and experience with generative AI and language models can make you more competitive. Given the current industry focus, demonstrating relevant skills in these areas can open up more opportunities.

[Creativity and Curiosity] Kai highlights that creativity and curiosity are crucial for roles involving ethical challenges and language models. Develop and demonstrate these traits in your projects and interviews, focusing on innovative solutions or improvements in efficiency and scalability.

[Research and Academic Involvement] Sindhu suggests gaining experience in ethics-related projects as an undergraduate. Look for research projects at your university that focus on bias, ethics, or AI's societal impact, and get involved in these initiatives to build relevant skills and knowledge.

[Academic Pathways] Engage with professors or research groups working on the intersection of AI and ethics. This can provide project experience and mentoring, which are valuable for building a career in this niche area.

[Stay Informed and Educated] Regularly read about the latest research and development in AI ethics to stay informed. This knowledge can be valuable during networking or discussions with potential employers or collaborators.

[Highlight Relevant Experience] When applying for roles, tailor your resume and cover letter to emphasize any relevant experience with projects, coursework, or independent studies related to AI ethics and model evaluation.

[Rapid Growth and Competition] Mercy highlights that the explosive growth of AI has matched the surge in interest from a wide pool of candidates. This makes the field competitive, giving employers a larger pool from which to choose, and allowing them to demand extensive experience.

[Crowded Field] Sindhu notes that as AI grows, the field becomes more crowded. Additionally, tools like generative AI coding assistants reduce the need for tasks traditionally performed by humans, shifting expectations toward a higher level of problem-solving and technical expertise.

[Industry Transition and Uncertainty] Kai explains that the emergence of generative AI has impacted hiring strategies as industries are still figuring out how to incorporate these technologies into their future plans. This transitional period adds uncertainty to the job market.

[Opportunity for Innovation] There are opportunities to innovate and explore verticals that haven’t been fully developed yet. This could be a chance to start your own initiatives or enter less saturated areas of AI.

Motivations for Employer Demands

• Skill Shortages: Employers are seeking candidates who not only have technical skills but can also adapt to new AI tools and frameworks.

• Maximized Efficiency: With AI tools capable of handling routine tasks, companies look for candidates who can contribute value beyond basic coding, such as developing advanced algorithms or integrating AI into business processes.

• Adaptation and Future-Readiness: Employers seek employees who are ready to navigate the evolving landscape of AI, especially as it becomes integral to various sectors.

I would encourage them to go for it! If someone is interested in AI, there are plenty of ways to get involved. I think someone could take at least two approaches: first, bring your current expertise and strengths to think about applications and ethics of AI; second, start to take some courses or tutorials to learn more about the fundamentals of AI science, and look for potential internship or research opportunities to practice those new skills.

[Focus on Strengths and Interests] Kai suggests leveraging your existing strengths and interests to identify areas in AI that resonate with you. Aligning your focus with areas where you have a unique edge, such as a background in math or physics, can help you carve out a niche that isn't overcrowded.

[Concentration on Specific Verticals] Mercy emphasizes the importance of concentrating on a particular vertical or area of passion. By focusing on a specific domain that interests you, such as health or multimodal digital biomarkers, you can stay updated with relevant advancements without getting overwhelmed by every new trend.

[Curiosity-Driven Learning] Sindhu advises shifting your mindset from trying to "catch up" with everything to "keeping up" with areas that genuinely pique your curiosity. This can reduce anxiety and make the learning process more enjoyable and sustainable.

[Selective Conference Focus] Choose a few reputable conferences, like ICML or NeurIPS, to follow. Explore their papers and keynotes to gain insights into emerging trends and topics that align with your interests.

[Use of Digests and Updates] Subscribing to digests from platforms like Medium can help you receive curated articles and updates tailored to your interest areas, making it easier to stay informed about relevant advancements.

[Balanced Approach] Recognize that it's impossible to follow every new development. Prioritize the areas most relevant to your career goals and personal interests, and allow yourself the flexibility to adapt your focus as your interests evolve.

I think it is difficult to say, but I think software developers will benefit from learning to use AI as a tool, as that is likely a skill employers will look for. 

It might be possible to try to prevent or discourage cheating by having more in-person interactions and assessments, such as oral exams or discussions. It might also be worthwhile to discuss why people should or should not use AI, or encourage people to honestly report when they would use AI, rather than passing AI answers off as their own.

I think it's important to keep learning objectives in mind to inform how much AI is acceptable. If a course is focused on building fundamental skills that AI may already easily automate, perhaps less AI use is preferable. In those cases, encouraging less use with more in-person learning activities and assessment may be helpful. In my upper level courses, I prefer to allow the use of AI and encourage people to learn to use it responsibly. For example, I like to ask people to use it to try to answer questions, then discuss whether it is correct or not, or whether it could lead to negative impacts if followed without question (https://www.gse.harvard.edu/ideas/usable-knowledge/23/07/embracing-artificial-intelligence-classroom). There is also some potential to get help more quickly than teaching staff can provide (https://teaching.cornell.edu/generative-artificial-intelligence), again, if used carefully.