What Is Artificial Intelligence? Research Areas, Funding, and Career Paths

Artificial intelligence (AI) is the branch of computer science concerned with building systems that perform tasks which, when done by people, are understood to require intelligence — perceiving, reasoning, learning from data, making decisions under uncertainty, and using language. It is not a single technique but a broad research program spanning symbolic reasoning, statistical learning, and, most visibly today, large-scale neural networks trained on data. As a research field it sits inside computer science while drawing heavily on mathematics (statistics, linear algebra, optimization), neuroscience-inspired ideas about learning, and increasingly its own dedicated funding, publication, and training infrastructure — which is why it is treated here as its own guide rather than a subsection of computer science generally. See CASRAI’s parent guide, What Is Computer Science?, for the wider discipline AI sits inside, and the Branches of Science hub for how AI relates to the sciences overall.

What Artificial Intelligence Actually Studies

AI research is organized around a small set of recurring questions: How can a system represent knowledge about the world and reason over it? How can a system learn a useful pattern or decision rule from examples rather than being explicitly programmed with rules? How should it act under uncertainty, with incomplete information, or in competition/cooperation with other agents? And, increasingly, how can systems that produce fluent language or images be evaluated, made reliable, and used responsibly?

• Machine learning — algorithms that improve at a task from data rather than from hand-written rules, including supervised learning (learning from labeled examples), unsupervised learning (finding structure in unlabeled data), and reinforcement learning (learning by trial and reward).
• Knowledge representation and reasoning — how a system encodes facts and rules about a domain and derives new conclusions from them; the older, symbolic tradition in AI that predates the current statistical-learning wave and still underlies areas like automated planning and formal verification.
• Perception — computer vision (interpreting images and video) and speech recognition (interpreting audio), both now dominated by learned models trained on large labeled or self-supervised datasets.
• Natural language processing (NLP) — understanding, generating, and translating human language; the research area behind the large language models that produced much of the field’s recent public visibility.
• Robotics and embodied AI — systems that perceive and act in the physical world, combining perception, planning, and control.
• Multi-agent systems and game theory — how multiple learning or reasoning agents interact, cooperate, or compete.

Major Sub-Disciplines Within AI

Within these broad questions, AI has developed identifiable sub-fields that often have their own conferences, journals, and academic communities:

• Deep learning — machine learning using multi-layered (deep) artificial neural networks; the dominant approach behind current advances in vision, language, and speech.
• Natural language processing and computational linguistics — language-specific modeling, including the transformer-based architectures underlying modern large language models.
• Computer vision — image classification, object detection, segmentation, and scene understanding.
• Reinforcement learning — learning optimal sequences of actions through reward signals, used in robotics, game-playing systems, and increasingly in fine-tuning language models.
• AI safety and alignment — a fast-growing research area studying how to make AI systems behave reliably, avoid harmful failure modes, and remain steerable as their capabilities scale.
• Explainable AI (XAI) and interpretability — methods for understanding why a model produced a given output, motivated by both scientific and regulatory demand for transparency.
• Robotics — the applied intersection of AI with mechanical and control engineering.
• AI ethics and fairness — studying bias, accountability, and the societal effects of deployed AI systems, now a standard component of major AI research programs and conferences rather than a separate afterthought.

AI also increasingly functions as a method applied inside other disciplines — AI for drug discovery and structural biology, AI for climate modeling, AI for materials science — which is one reason its funding landscape (below) runs through so many different federal agencies rather than a single home.

The Research Funding Landscape for AI

Because AI research spans basic computer science, mission-driven application areas, and national-security-relevant computing, it is funded through more federal channels than a typical single-department discipline, and understanding that landscape is part of navigating a research career in it.

• National Science Foundation (NSF) — the primary U.S. federal funder of fundamental, non-mission-specific AI research. NSF states it is a major federal funder of AI, investing several hundred million dollars a year in foundational and use-inspired AI research. The Directorate for Computer and Information Science and Engineering (CISE) is NSF’s core home for AI funding, supporting work in machine learning theory, human-AI interaction, computer vision, and human language technologies, largely through its Information and Intelligent Systems (IIS) division and related CISE programs. NSF also leads the multi-award National AI Research Institutes program, a cross-agency effort (run jointly with partner federal agencies and, for some awards, private-sector co-funders) that funds distributed research centers on specific AI application areas rather than a single physical institute.
• National Institutes of Health (NIH) — funds AI and machine learning as applied methods across many of its institutes rather than through one dedicated “AI institute.” The National Library of Medicine (NLM) has a specific, longstanding focus on biomedical informatics and computational methods, including AI applied to clinical and biomedical data, and runs biomedical informatics training programs that are a common entry route into health-focused AI research. Other institutes — including the National Institute of Biomedical Imaging and Bioengineering (NIBIB) and the National Cancer Institute (NCI) — fund AI applied to medical imaging, diagnostics, and oncology research as part of their broader mission portfolios rather than as a separate AI program.
• Defense Advanced Research Projects Agency (DARPA) — a long-running and historically significant funder of foundational AI research within the U.S. Department of Defense, dating back to some of the field’s earliest work; DARPA continues to fund AI research programs oriented toward robustness, reasoning, and next-generation capabilities.
• Department of Energy (DOE) Office of Science — funds AI applied to scientific computing at the national laboratories, including AI methods for accelerating simulation and data analysis in physics, materials science, and climate research.
• Private foundations and industry labs — a substantial share of AI research funding and output, particularly in deep learning, now comes from industry research labs rather than traditional grant-making foundations; researchers navigating this field should treat that mixed public/private funding structure as one of AI’s genuine differences from most academic disciplines.

Researchers applying for AI-related federal funding should also expect to navigate general federal grants-management requirements — see CASRAI’s grants management resources for the funding-lifecycle mechanics that apply regardless of the specific funding agency.

Research Methods, Tools, and Equipment

AI research is computational rather than laboratory-based in the traditional sense, but it has its own characteristic infrastructure:

• Compute infrastructure — GPU and increasingly specialized AI-accelerator clusters (used for training large models), which have become a defining resource constraint in the field and a major line item in AI research budgets and grant proposals.
• Datasets and benchmarks — standardized, shared datasets and benchmark tasks that let different research groups compare methods on equal footing; curating and documenting these datasets responsibly (provenance, licensing, bias auditing) is itself an active area of methods research.
• Software frameworks — open-source deep learning frameworks and libraries used to build, train, and evaluate models, alongside experiment-tracking and reproducibility tooling.
• Evaluation methodology — a growing methods literature on how to measure model performance, robustness, and failure modes rigorously, distinct from simply reporting a benchmark score.
• Human evaluation and user studies — for AI systems that interact with people (language models, recommendation systems, robotics), controlled studies with human participants are a standard complementary method to purely computational evaluation.

Researchers relying on shared compute, datasets, or trained models also increasingly need clear data management practices; CASRAI’s research data management resources cover the data-stewardship expectations that funders now commonly attach to AI and data-intensive research.

Careers and Training Pathways

Most AI researchers hold a graduate degree — typically a PhD in computer science, or increasingly a dedicated AI or machine learning graduate program, though the field also has a well-established path through applied master’s programs into industry research roles. A typical PhD track involves foundational coursework in algorithms, statistics, and machine learning theory, followed by original research culminating in a dissertation, with active publication at peer-reviewed AI conferences functioning as the field’s primary venue for disseminating new results (a norm that distinguishes AI from many other computer science sub-fields, where journal publication is more central). Postdoctoral research positions remain common for those pursuing academic careers, while many PhD and master’s graduates move directly into industry research labs.

The Association for the Advancement of Artificial Intelligence (AAAI) is the longest-established general professional society specifically for AI research, and organizations such as the Association for Computing Machinery (ACM) host major sub-field-specific conferences and special interest groups relevant to AI researchers (including SIGKDD for data mining and SIGCHI for human-computer interaction). Researchers whose AI work sits inside a domain science — biomedicine, physics, environmental science — typically also belong to that domain’s own professional societies alongside a general AI or computer science one.

How AI Relates to Neighboring Fields

AI is formally a sub-field of computer science, and shares deep intellectual roots with it in algorithms, computational complexity, and systems design. It overlaps substantially with mathematics (statistics, optimization, linear algebra underlie essentially all modern machine learning) and increasingly functions as a shared method across the natural sciences: AI techniques are now standard tools in fields such as biotechnology (protein structure prediction, genomic analysis), astronomy (classifying astronomical survey data at scale), microbiology and marine biology (image-based species and pattern classification), and nanotechnology and materials research (accelerating simulation and materials discovery). Applied and engineering fields including industrial engineering, environmental engineering, and food science increasingly use AI-based methods for process optimization, monitoring, and prediction within their own domains. See the Branches of Science hub for the full landscape of how these disciplines relate to one another.

Frequently Asked Questions

Is artificial intelligence the same as machine learning?

No — machine learning is a sub-field of AI, specifically the branch concerned with systems that learn patterns from data rather than following explicitly programmed rules. AI is the broader field, which also includes symbolic reasoning, knowledge representation, robotics, and other approaches that do not necessarily involve learning from data.

What is the difference between AI and data science?

They overlap substantially but are not the same field. Data science is broader in its subject matter (any extraction of insight from data, including largely descriptive and statistical work) and narrower in its methods; AI specifically studies building systems that perform tasks associated with intelligence, and increasingly uses data science’s statistical toolkit to do so via machine learning.

What degree do I need to do AI research?

Most independent AI research positions, in academia or industry research labs, require a PhD in computer science or a closely related field such as a dedicated machine learning or AI program. Applied AI/ML engineering roles are more accessible with a master’s degree, though the research track proper is PhD-centered.

Which federal agencies fund AI research?

In the United States, the National Science Foundation is the primary funder of fundamental AI research through its CISE directorate and the National AI Research Institutes program; NIH funds AI as an applied method across several of its institutes, particularly the National Library of Medicine; DARPA and the Department of Energy also fund significant AI research programs tied to their respective missions.