ARTIFICIAL intelligence is no longer just a peripheral tool in research. The rapid accumulation, availability and accessibility of data sets has now become a structural layer of the scientific method itself. Across materials science, drug development, biology, economics, education, and business, AI systems are now generating hypotheses, designing experiments, synthesizing data, and even executing laboratory workflows. The result is not just faster science, but a fundamental shift in how knowledge is created.
One of the clearest examples of use of AI in scientific discovery is the Materials Project at the U.S. Department of Energy’s Lawrence Berkeley National Laboratory. Originally launched as a computational database of material properties, it has evolved into the world’s most widely used materials informatics platform, which as of December last year is serving more than 650,000 scientists, engineers, and companies.
Materials databases
The Materials Project does not simply store data. It provides machine-learning-ready datasets that allow AI models to predict how new compounds will behave before they are ever synthesized. These predictions are now used to guide the design of better battery cathodes, solid electrolytes, catalysts, and semiconductors. In practical terms, AI can now evaluate millions of candidate materials in silico, narrowing down a handful of promising candidates for real-world testing — a process that once took decades.
This predictive infrastructure feeds directly into Berkeley Lab’s Autonomous Laboratory (A-Lab), which began operating in 2023. A-Lab combines robotics, machine learning, and closed-loop optimization to run materials experiments without human intervention. The system chooses which compounds to make, synthesizes them, measures their properties, updates its models, and decides what to try next. In its first months of operation, A-Lab successfully synthesized dozens of novel inorganic compounds that had never existed before, demonstrating that AI-guided experimentation can outperform traditional trial-and-error research.
Battling cancer
AI-driven biology and drug discovery is now used to combat pancreatic cancer, one of the deadliest and most treatment-resistant cancers. Researchers using AI-based graph neural networks screened millions of drug combinations to identify molecules that could overcome chemotherapy resistance. The system predicted hundreds of previously unknown drug synergies, many of which were later validated in laboratory cell cultures. This kind of combinatorial search is essentially impossible using conventional screening methods, but AI made it computationally tractable.
Pharmaceutical companies are now building AI directly into their research pipelines. AstraZeneca, for example, acquired the AI company Modella to integrate foundation models into oncology drug development. These models analyze molecular structures, biological pathways, and patient data simultaneously to identify which drug candidates are most likely to succeed in clinical trials.
Another major leap comes from biological data itself. Illumina, the world’s largest DNA sequencing company, recently released what it calls the Billion Cell Atlas, a massive dataset of gene-level activity across different diseases. AI models trained on this atlas can predict how cells will respond to genetic or chemical perturbations, allowing researchers to identify new drug targets for cancer, immune disorders, and metabolic diseases. Instead of running thousands of wet-lab experiments, scientists can now simulate them computationally and focus only on the most promising leads.
AI lab assistants
AI is also changing how science itself is conducted. New generations of AI agents can now generate hypotheses, design experiments, operate instruments, and analyze results. These systems are being deployed in physics, chemistry, and biology as autonomous research assistants.
In practice, this means an AI can read thousands of papers, detect gaps in the literature, propose experiments, and then instruct robotic systems to carry them out. Researchers at multiple national laboratories have already demonstrated closed-loop AI systems that move from idea to data without human input. Scientists increasingly act not as bench technicians, but as supervisors of AI-driven discovery engines.
Personalized intelligence at scale
By 2026, more than 80 percent of students worldwide are using AI tools for tutoring, writing assistance, exam preparation, and research. Intelligent tutoring systems analyze how individual students learn, where they struggle, and how they respond to different explanations, then adjust instruction in real time.
At the National University of Singapore, machine-learning systems analyze course engagement, assessment patterns, and learning trajectories to identify students who are at risk of falling behind and automatically recommend targeted interventions, supplemental materials, or tutoring pathways. In Qatar, Hamad Bin Khalifa University uses AI-powered learning platforms that dynamically adapt reading materials, assignments, and practice exercises to each student’s progress, allowing learners to move at different speeds through the same curriculum. In Spain, Rovira i Virgili University applies AI-based assessment systems that evaluate student presentations and written work in real time, generating individualized feedback on clarity, pacing, and conceptual mastery — something that would normally require hours of instructor time per student. In the United States, Northeastern University has deployed institution-wide AI learning assistants for nearly 50,000 students, providing personalized study guides, automated quiz generation, and adaptive feedback tied directly to course objectives.
AI in economics
Forecasting and policy analysis, functions within the purview of economics have also been transformed. Traditional economic models rely on sparse, delayed data. AI models, by contrast, ingest high-frequency streams from financial markets, supply chains, labor platforms, and consumer behavior to produce near-real-time economic indicators.
One example is the Anthropic Economic Index, which tracks how AI is automating tasks across industries and regions. This index reveals which jobs are being transformed, how fast productivity is changing, and where labor displacement is occurring — information that was previously impossible to measure at scale.
Investment firms now use AI to simulate how technological change affects GDP, inflation, and employment. These models suggest that productivity gains from AI could add more than a full percentage point to economic growth in advanced economies while stabilizing labor markets through efficiency gains.
Automation to autonomy
In business, AI agents are becoming “digital coworkers.” They forecast demand, optimize logistics, negotiate supplier contracts, and run simulations of corporate strategy. Large enterprises deploy AI not just to cut costs, but to explore scenarios, test product launches, and manage risk.
Companies now use AI to retrain workers, redesign workflows, and reallocate capital in response to real-time market signals. While automation will eliminate some roles, it is also creating demand for new skills in data interpretation, AI supervision, and human-machine collaboration.
ARTIFICIAL intelligence is no longer just a peripheral tool in research. The rapid accumulation, availability and accessibility of data sets has now become a structural layer of the scientific method itself. Across materials science, drug development, biology, economics, education, and business, AI systems are now generating hypotheses, designing experiments, synthesizing data, and even executing laboratory workflows. The result is not just faster science, but a fundamental shift in how knowledge is created.
One of the clearest examples of use of AI in scientific discovery is the Materials Project at the U.S. Department of Energy’s Lawrence Berkeley National Laboratory. Originally launched as a computational database of material properties, it has evolved into the world’s most widely used materials informatics platform, which as of December last year is serving more than 650,000 scientists, engineers, and companies.
Materials databases
The Materials Project does not simply store data. It provides machine-learning-ready datasets that allow AI models to predict how new compounds will behave before they are ever synthesized. These predictions are now used to guide the design of better battery cathodes, solid electrolytes, catalysts, and semiconductors. In practical terms, AI can now evaluate millions of candidate materials in silico, narrowing down a handful of promising candidates for real-world testing — a process that once took decades.
This predictive infrastructure feeds directly into Berkeley Lab’s Autonomous Laboratory (A-Lab), which began operating in 2023. A-Lab combines robotics, machine learning, and closed-loop optimization to run materials experiments without human intervention. The system chooses which compounds to make, synthesizes them, measures their properties, updates its models, and decides what to try next. In its first months of operation, A-Lab successfully synthesized dozens of novel inorganic compounds that had never existed before, demonstrating that AI-guided experimentation can outperform traditional trial-and-error research.
Battling cancer
AI-driven biology and drug discovery is now used to combat pancreatic cancer, one of the deadliest and most treatment-resistant cancers. Researchers using AI-based graph neural networks screened millions of drug combinations to identify molecules that could overcome chemotherapy resistance. The system predicted hundreds of previously unknown drug synergies, many of which were later validated in laboratory cell cultures. This kind of combinatorial search is essentially impossible using conventional screening methods, but AI made it computationally tractable.
Pharmaceutical companies are now building AI directly into their research pipelines. AstraZeneca, for example, acquired the AI company Modella to integrate foundation models into oncology drug development. These models analyze molecular structures, biological pathways, and patient data simultaneously to identify which drug candidates are most likely to succeed in clinical trials.
Another major leap comes from biological data itself. Illumina, the world’s largest DNA sequencing company, recently released what it calls the Billion Cell Atlas, a massive dataset of gene-level activity across different diseases. AI models trained on this atlas can predict how cells will respond to genetic or chemical perturbations, allowing researchers to identify new drug targets for cancer, immune disorders, and metabolic diseases. Instead of running thousands of wet-lab experiments, scientists can now simulate them computationally and focus only on the most promising leads.
AI lab assistants
AI is also changing how science itself is conducted. New generations of AI agents can now generate hypotheses, design experiments, operate instruments, and analyze results. These systems are being deployed in physics, chemistry, and biology as autonomous research assistants.
In practice, this means an AI can read thousands of papers, detect gaps in the literature, propose experiments, and then instruct robotic systems to carry them out. Researchers at multiple national laboratories have already demonstrated closed-loop AI systems that move from idea to data without human input. Scientists increasingly act not as bench technicians, but as supervisors of AI-driven discovery engines.
Personalized intelligence at scale
By 2026, more than 80 percent of students worldwide are using AI tools for tutoring, writing assistance, exam preparation, and research. Intelligent tutoring systems analyze how individual students learn, where they struggle, and how they respond to different explanations, then adjust instruction in real time.
At the National University of Singapore, machine-learning systems analyze course engagement, assessment patterns, and learning trajectories to identify students who are at risk of falling behind and automatically recommend targeted interventions, supplemental materials, or tutoring pathways. In Qatar, Hamad Bin Khalifa University uses AI-powered learning platforms that dynamically adapt reading materials, assignments, and practice exercises to each student’s progress, allowing learners to move at different speeds through the same curriculum. In Spain, Rovira i Virgili University applies AI-based assessment systems that evaluate student presentations and written work in real time, generating individualized feedback on clarity, pacing, and conceptual mastery — something that would normally require hours of instructor time per student. In the United States, Northeastern University has deployed institution-wide AI learning assistants for nearly 50,000 students, providing personalized study guides, automated quiz generation, and adaptive feedback tied directly to course objectives.
AI in economics
Forecasting and policy analysis, functions within the purview of economics have also been transformed. Traditional economic models rely on sparse, delayed data. AI models, by contrast, ingest high-frequency streams from financial markets, supply chains, labor platforms, and consumer behavior to produce near-real-time economic indicators.
One example is the Anthropic Economic Index, which tracks how AI is automating tasks across industries and regions. This index reveals which jobs are being transformed, how fast productivity is changing, and where labor displacement is occurring — information that was previously impossible to measure at scale.
Investment firms now use AI to simulate how technological change affects GDP, inflation, and employment. These models suggest that productivity gains from AI could add more than a full percentage point to economic growth in advanced economies while stabilizing labor markets through efficiency gains.
Automation to autonomy
In business, AI agents are becoming “digital coworkers.” They forecast demand, optimize logistics, negotiate supplier contracts, and run simulations of corporate strategy. Large enterprises deploy AI not just to cut costs, but to explore scenarios, test product launches, and manage risk.
Companies now use AI to retrain workers, redesign workflows, and reallocate capital in response to real-time market signals. While automation will eliminate some roles, it is also creating demand for new skills in data interpretation, AI supervision, and human-machine collaboration.
