Skip to content

AI in Academic Research: A Comparative Analysis of ML and Generative AI Tools

1. Introduction

Artificial intelligence (AI) is rapidly transforming the landscape of academic research, offering unprecedented opportunities to accelerate discovery and enhance productivity [24]. As foundational AI concepts, including machine learning (ML) and, more recently, generative AI, become increasingly sophisticated [26, 33], their application within research workflows is expanding across diverse disciplines. This report provides a comprehensive overview of AI-assisted research tools, a comparative analysis of traditional ML and generative AI techniques, and a critical examination of the ethical considerations surrounding their implementation. The scope of this report encompasses tools designed to streamline literature review, facilitate data analysis and visualization, and augment the writing and editing process [24].

This analysis will address several key research questions. First, how do the capabilities of generative AI models compare to those of traditional machine learning approaches in the context of academic research? Second, what are the most promising current and emerging AI tools available to researchers, and what are their respective strengths and weaknesses? Third, what ethical challenges arise from the integration of AI into research, and how can these be effectively addressed to ensure responsible innovation? Finally, what are the potential future trends and implications of AI for the future of academic research?

The following sections will delve into these questions, providing a detailed exploration of the technological advancements, practical applications, and ethical considerations shaping the future of AI-assisted research. We begin by outlining foundational AI concepts to establish a common understanding of the technologies discussed.

2. Foundational AI Concepts

Foundational AI concepts underpin the recent advancements in AI-assisted research tools. Artificial intelligence (AI), broadly defined, focuses on enabling computers to perform tasks that typically require human intelligence [26]. A core branch of AI is machine learning (ML), which allows systems to learn from data without explicit programming [26]. Within ML, several key paradigms exist. Supervised learning involves training algorithms on labeled datasets, enabling predictions or classifications based on input data [3]. Common supervised learning algorithms include linear regression, logistic regression, decision trees, and support vector machines (SVMs) [26]. Unsupervised learning, conversely, identifies patterns in unlabeled data, often used for clustering or dimensionality reduction [3].

More recently, generative AI has emerged as a prominent subfield of ML, gaining significant attention for its ability to create new content [33]. Generative AI encompasses models like large language models (LLMs), diffusion models, and generative adversarial networks (GANs) [13]. While generative AI is a type of machine learning [33], it distinguishes itself by its capacity to generate novel outputs—text, images, audio, or other data—rather than simply predicting or classifying existing data. LLMs, such as GPT-4, achieve this through training on massive datasets and leveraging deep learning techniques [17]. GANs, composed of generator and discriminator networks, iteratively refine their output to produce realistic synthetic data [13].

However, it is crucial to recognize that generative AI and traditional ML are not mutually exclusive. They can be, and increasingly are, used in a complementary fashion. For example, generative AI can create synthetic data to augment training datasets for traditional ML models, particularly when real-world data is scarce [20]. Furthermore, generative AI can assist in feature engineering or provide contextual information to improve the accuracy of traditional ML algorithms [33]. While generative AI excels at tasks involving natural language or common images, traditional ML often remains superior for applications requiring high precision, privacy, or domain-specific knowledge [33].

Understanding these foundational concepts is critical for evaluating the capabilities and limitations of the AI-assisted research tools discussed in subsequent sections. Next, we will explore a more detailed comparison of traditional ML versus generative AI in the context of academic research applications.

3. Traditional ML vs. Generative AI in Research

This section provides a comparative analysis of traditional machine learning (ML) and generative artificial intelligence (AI) techniques as applied to contemporary research workflows. We will examine the distinct capabilities of each approach, focusing on common research tasks such as data analysis and interpretation, literature review, and the formulation of novel hypotheses. The subsequent discussion will detail specific applications of traditional ML methods, followed by an exploration of the emerging landscape of generative AI tools and their potential to augment and transform the research process. This comparison will highlight the strengths and weaknesses of each approach, offering researchers a nuanced understanding of their respective roles in advancing scientific inquiry.

3.1. Traditional ML Applications

Traditional machine learning (ML) techniques have long been integral to academic research across diverse disciplines. These methods provide researchers with tools for analyzing complex datasets, identifying patterns, and building predictive models. Classification and regression, two fundamental ML tasks, are frequently employed. For instance, in biomedical research, support vector machines (SVMs) have been utilized for cancer classification and subtyping, maximizing the margin of separation between disease states and demonstrating strong performance with high-dimensional genomic data [9]. SVMs’ ability to handle complex genomic data and identify subtle patterns makes them well-suited for this application [9]. Similarly, decision trees, and their ensemble methods like random forests, are widely used in fields like financial markets, image processing, and medical research due to their ability to handle large-scale data and their interpretability [30]. Recent advancements in decision tree algorithms focus on optimizing attribute selection, improving model performance through eigenvalue importance and adaptive entropy [30].

Regression techniques, including linear regression and regression trees, are commonly applied to model and predict continuous variables. In ecological studies, for example, linear regression might be used to model the relationship between environmental factors and species abundance [1]. More complex relationships are often captured using regression tree ensembles, which can handle non-linearities and interactions between variables [1].

Clustering algorithms, an unsupervised learning approach, are valuable for exploratory data analysis and identifying hidden groupings within datasets. Hierarchical clustering, for example, is used in biology to group species by biological features, helping to understand evolutionary relationships [25]. It is also used for categorizing populations in clinical research and customer segmentation [25]. While computationally intensive, techniques like using heaps can improve efficiency [25].

Data augmentation techniques, often integrated with traditional ML, can address limitations related to data scarcity and overfitting [23]. Geometric and color space transformations create variations of existing images, increasing dataset size and improving model robustness [23]. However, care must be taken to ensure transformations preserve label integrity [23]. The use of these techniques alongside frameworks like TensorFlow and Keras can streamline the augmentation process [23].

It’s important to note that the success of traditional ML models relies heavily on careful feature selection and pre-processing [1, 30]. Without appropriate feature engineering, models can be prone to overfitting or may fail to capture the underlying relationships in the data [30]. Furthermore, while effective, traditional ML techniques often require significant domain expertise to interpret results and validate findings. Having examined the applications of traditional ML, the following section will delve into the capabilities of generative AI models and their potential to address some of the limitations of traditional approaches.

3.2. Generative AI Applications

Generative AI models are rapidly expanding the scope of what’s possible in academic research, offering capabilities that complement and, in some instances, surpass those of traditional machine learning approaches. Unlike traditional ML, which typically focuses on prediction or classification based on existing data, generative AI excels at creating novel content – text, images, code, and more – mirroring the style, tone, or structure of its training data [17]. This capability unlocks new avenues for research across various stages of the process, from initial ideation to dissemination of findings.

A significant application lies in assisting with literature review and synthesis. Tools leveraging LLMs can summarize research papers, identify key themes, and even generate search strings combining relevant keywords and MeSH terms, accelerating the often-time-consuming process of staying current with the literature [7]. However, researchers must remain vigilant regarding potential inaccuracies and biases inherent in AI-generated summaries, validating the information against the original sources and acknowledging the assistance provided [28]. AI can also aid in identifying connections between papers, visualizing citation networks, and uncovering emerging research trends [7].

Generative AI is also proving valuable in data analysis and interpretation, though with caveats. While tools like Bard or ChatGPT can assist in summarizing online survey data or datasets, researchers must exercise caution when dealing with confidential or protected information, ensuring compliance with privacy regulations and data security protocols [28]. The ability to generate synthetic data, particularly in scenarios where real-world data is limited or unavailable, represents another promising application [20]. This synthetic data can be used to augment training datasets for traditional ML models, improving their performance and generalizability.

Furthermore, generative AI is transforming the process of manuscript drafting and editing. LLMs can assist with grammar correction, sentence restructuring, and even generating initial drafts of sections like introductions or methods descriptions [29]. However, the potential for plagiarism and the need to maintain originality necessitate careful review and substantial authorial input [28]. Image generation tools, such as Midjourney or DeepAI, can create figures and illustrations for publications, but researchers bear the responsibility of verifying their accuracy and addressing intellectual property considerations [28].

Emerging applications extend beyond these core areas. Researchers are exploring the use of LLMs for hypothesis generation, leveraging their ability to identify patterns and propose novel research directions [11]. Multi-agent systems, combining the strengths of multiple LLMs, are being developed to automate complex scientific discovery processes [11]. The development of tools specifically tailored for scientific hypothesis generation, and even laboratory validation of those hypotheses, represents a significant step toward AI-augmented research [11]. However, challenges related to knowledge integration, causal reasoning, and ensuring the reliability of AI-generated insights remain [4]. The integration of generative AI into academic research is not without its ethical considerations, and these must be carefully addressed through clear guidelines, responsible use, and ongoing evaluation [29, 32].

Having examined the applications of generative AI, the following section will discuss the ethical considerations and challenges associated with these technologies in greater detail.

4. Current AI Tools for Academic Research

AI-assisted tools are rapidly transforming academic research, offering solutions to streamline workflows and enhance productivity [24]. These tools fall into three main categories: literature review, data analysis, and writing & editing [24]. Literature review tools, such as Litmaps, visualize citation networks, track research progress, and identify gaps in the literature [24]. Semantic Scholar, an AI-powered search engine, offers intelligent filtering and citation analysis, while Google Scholar provides broad coverage but lacks advanced visualization features [24]. Scopus offers high-quality data but can be costly [24]. For data analysis, SPSS is user-friendly for statistical analysis but is expensive [24]. NVivo excels in qualitative data analysis, but has a steep learning curve [24]. Google AutoML allows building ML models without coding, but can be expensive for large datasets [24]. R offers high flexibility but requires coding knowledge [24]. Writing and editing tools, including Grammarly, Paperpal, and QuillBot, improve grammar, clarity, and tone [16, 24]. Paperpal is specifically designed for academic writing, improving clarity for journal preparation [16].

Several emerging tools further extend these capabilities. Elicit is an AI research assistant focused on evidence synthesis, identifying relevant papers even without precise keyword matching and organizing them into tables [2, 8]. Consensus extracts answers directly from research papers using NLP, identifying prevailing opinions [2]. Research Rabbit and Litmaps visualize connections between papers, aiding in discovery and organization [2, 22]. These tools often leverage Retrieval Augmented Generation (RAG) to improve answer accuracy, drawing on scholarly databases like Semantic Scholar and Scopus [27]. Scite.ai analyzes citation context, indicating whether citations support or contradict claims [27]. Anara combines literature discovery and analysis, providing source traceability and tools for systematic reviews [12]. While these tools offer significant advantages, it’s important to recognize that many rely on access to databases with paywalls, and the quality of results can vary [19].

However, researchers should exercise caution when using these tools. The potential for AI to generate inaccurate or biased information remains a concern [6]. Challenges include potential bias in AI systems, a lack of human insight, and data quality dependency [6]. The use of AI tools also raises ethical concerns regarding privacy and misuse [6]. Furthermore, Researchers should be aware of the limitations of these tools, particularly regarding access to the most recent publications and the potential for hallucination [8]. Careful validation of AI-generated outputs and transparency in research methods are crucial [5defb75d].

Choosing the appropriate tool depends on specific research needs, data type, and desired functionalities [24]. Factors to consider include strengths, accuracy, data privacy, user interface, and cost [24]. While AI tools can significantly accelerate research, they are best used as aids to, rather than replacements for, critical thinking and expert judgment [15]. Next, we will explore emerging applications and future trends in AI-assisted academic research.

5. Emerging Applications & Future Trends

The trajectory of AI in academic research extends beyond current applications, promising transformative changes across multiple facets of the research lifecycle. One emerging area is automated experiment design and execution. AI algorithms, particularly reinforcement learning, are being developed to optimize experimental parameters, reducing the need for extensive manual trial-and-error [11]. This is especially valuable in fields like materials science and drug discovery, where the search space for optimal conditions is vast [11]. By iteratively learning from experimental outcomes, AI can accelerate the identification of promising candidates and minimize resource expenditure. However, the ‘black box’ nature of some AI algorithms necessitates careful validation and interpretability to ensure scientific rigor.

Personalized learning represents another significant frontier. AI-powered systems can analyze student performance data to tailor educational content and learning pathways, adapting to individual needs and learning styles [9c610795]. This has implications for both formal education and researcher training, enabling more efficient knowledge acquisition and skill development. Adaptive learning platforms can identify knowledge gaps and provide targeted interventions, fostering a more effective and engaging learning experience. However, ethical considerations related to data privacy and algorithmic bias must be addressed to ensure equitable access and outcomes.

Novel data discovery and knowledge synthesis are also poised for advancement. AI algorithms can sift through massive datasets, identifying previously unknown correlations and patterns that might escape human observation [11]. This is particularly relevant in fields like genomics and astrophysics, where data volumes are constantly expanding. Graph neural networks, for example, are proving effective in representing and analyzing complex relationships between entities, facilitating the discovery of new knowledge and insights [72f9d6c4]. Furthermore, advancements in natural language processing (NLP) are enabling AI systems to extract and synthesize information from unstructured data sources, such as research reports and clinical notes, accelerating the process of knowledge discovery.

Looking ahead, several technological advancements are likely to shape the future of AI in research. The continued development of more powerful and efficient LLMs will enhance their capabilities in tasks like text generation, translation, and summarization [17]. The integration of AI with other emerging technologies, such as quantum computing and blockchain, could unlock new possibilities for data analysis, security, and reproducibility [4]. Quantum machine learning algorithms, for example, have the potential to solve complex optimization problems that are intractable for classical computers. Federated learning, which enables collaborative model training without sharing sensitive data, will become increasingly important for addressing privacy concerns in collaborative research settings. The convergence of these technologies promises to create a more efficient, collaborative, and impactful research landscape. However, ongoing attention to ethical considerations, data quality, and algorithmic transparency will be crucial to ensure the responsible and beneficial deployment of AI in academic research. Having explored the emerging applications and future trends, the following section will address the crucial ethical considerations and challenges associated with AI in academic research.

6. Ethical Considerations and Challenges

The integration of artificial intelligence (AI) into academic research introduces a complex array of ethical considerations that demand careful attention [21]. While AI offers unprecedented opportunities to accelerate discovery and enhance research quality, it also presents novel challenges related to bias, plagiarism, data privacy, and the potential for misuse [21]. A primary concern revolves around algorithmic bias, as AI models are trained on data that may reflect existing societal prejudices, leading to skewed results and perpetuating inequalities [10]. For example, biases embedded in training datasets can lead to inaccurate or unfair outcomes in applications like risk assessment or predictive modeling [10]. Mitigating this requires careful scrutiny of training data, employing techniques to detect and correct biases, and ensuring diverse representation in datasets [18].

The potential for AI-assisted plagiarism also poses a significant threat to research integrity [21]. Generative AI tools can produce text that, while seemingly original, may be derived from existing sources without proper attribution, blurring the lines between legitimate synthesis and academic dishonesty [21]. Researchers must therefore adopt rigorous practices for verifying the originality of AI-generated content and adhere to strict guidelines for citation and disclosure [31]. The Committee on Publication Ethics (COPE) explicitly states that AI tools cannot be listed as authors, and researchers remain fully responsible for the content of their publications, even when AI is used [5]. Transparency is paramount; researchers must disclose how AI tools were utilized in their work, specifying the tools used in the Materials and Methods section [5].

Data privacy is another critical ethical concern, particularly when dealing with sensitive research data [21]. The use of AI algorithms to analyze personal or confidential information raises questions about data security, informed consent, and the potential for re-identification [14]. Researchers must ensure compliance with relevant data protection regulations, such as GDPR and HIPAA, and implement robust security measures to safeguard data privacy [14]. Furthermore, the use of AI tools that require data sharing with third-party providers necessitates careful consideration of data governance policies and contractual agreements [28].

Finally, the potential for misuse of AI in research necessitates proactive ethical frameworks [21]. AI could be used to fabricate data, manipulate results, or generate misleading publications, undermining public trust in science [21]. To address these concerns, institutions should develop and enforce clear AI research integrity guidelines, emphasizing transparency, accountability, and responsible innovation [21]. Mandatory AI ethics and integrity training for researchers is crucial, fostering awareness of potential misuses and promoting ethical research practices [21]. International collaboration is also essential to establish unified ethical standards and share best practices [21]. Explainable AI (XAI) techniques, like LIME, can help increase transparency and understanding of AI decision-making processes [14]. Algorithmic impact assessments and stakeholder engagement are vital for proactively identifying and mitigating potential harms [14]. While AI offers tremendous benefits, prioritizing ethical considerations is essential for maintaining the integrity and trustworthiness of academic research. Having examined the ethical considerations, the following section will conclude the report by summarizing key findings and outlining future directions for responsible AI implementation.

7. Conclusion

This report has explored the evolving landscape of artificial intelligence (AI) in academic research, beginning with foundational concepts and progressing through a comparative analysis of traditional machine learning (ML) and generative AI techniques [26, 33]. We examined current AI tools categorized by function – literature review, data analysis, and writing/editing – highlighting their strengths and limitations [24]. Emerging applications, including automated experiment design, personalized learning, and novel data discovery, were discussed, alongside potential technological advancements like quantum machine learning and federated learning [4, 11]. Crucially, we addressed the significant ethical considerations arising from AI integration, encompassing algorithmic bias, plagiarism, data privacy, and the potential for misuse [10, 21].

The analysis reveals a clear trend: AI is no longer a futuristic prospect but a present reality reshaping research methodologies across disciplines. Generative AI, in particular, demonstrates the potential to augment human capabilities, accelerate knowledge discovery, and facilitate innovative approaches to complex research challenges [17]. However, realizing this potential requires a nuanced understanding of the strengths and weaknesses of both traditional ML and generative AI, enabling researchers to strategically leverage each approach where it is most effective [13, 30].

Responsible implementation is paramount. Mitigating algorithmic bias through careful data curation and ongoing monitoring is essential for ensuring equitable and reliable research outcomes [18]. Maintaining research integrity demands rigorous verification of AI-generated content, transparent disclosure of AI tool usage, and adherence to established ethical guidelines [5, 31]. Protecting data privacy necessitates robust security measures and compliance with relevant regulations [14].

Looking forward, continued investment in AI ethics training, the development of explainable AI (XAI) techniques, and fostering international collaboration will be critical for navigating the evolving ethical landscape [14, 21]. As AI becomes increasingly integrated into the fabric of academic research, a proactive and responsible approach is not merely desirable—it is essential for upholding the integrity, trustworthiness, and societal benefit of scientific inquiry. The next step in this evolving field will be to assess the long-term impacts of these technologies and refine strategies for maximizing their positive contributions while mitigating potential risks.

References

  1. Modern Machine Learning Algorithms: Strengths and Weaknesses. Available at: https://elitedatascience.com/machine-learning-algorithms (Accessed: August 24, 2025)
  2. Office of Teaching, Learning, and Technology. Available at: https://teach.its.uiowa.edu/news/2024/03/ai-assisted-literature-reviews (Accessed: August 24, 2025)
  3. A Comprehensive Review on Machine Learning in Healthcare Industry: Classification, Restrictions, Opportunities and Challenges. Available at: https://pmc.ncbi.nlm.nih.gov/articles/PMC10180678/ (Accessed: August 24, 2025)
  4. https://journals.sagepub.com/doi/full/10.1177/02683962211048201. Available at: https://journals.sagepub.com/doi/full/10.1177/02683962211048201 (Accessed: August 24, 2025)
  5. https://publicationethics.org/guidance/cope-position/authorship-and-ai-tools. Available at: https://publicationethics.org/guidance/cope-position/authorship-and-ai-tools (Accessed: August 24, 2025)
  6. 12 AI Research Tools to Drive Knowledge Exploration. Available at: https://www.digitalocean.com/resources/articles/ai-research-tools (Accessed: August 24, 2025)
  7. AI-Based Literature Review Tools. Available at: https://tamu.libguides.com/c.php?g=1289555 (Accessed: August 24, 2025)
  8. Systematic Literature Review with Elicit AI: 4 Practical Use Cases + Limitations. Available at: https://medium.com/@borisnikolaev_57179/systematic-literature-review-with-elicit-ai-4-practical-use-cases-limitations-002e295caf11 (Accessed: August 24, 2025)
  9. Applications of Support Vector Machine (SVM) Learning in Cancer Genomics. Available at: https://pmc.ncbi.nlm.nih.gov/articles/PMC5822181/ (Accessed: August 24, 2025)
  10. When AI Gets It Wrong: Addressing AI Hallucinations and Bias. Available at: https://mitsloanedtech.mit.edu/ai/basics/addressing-ai-hallucinations-and-bias/ (Accessed: August 24, 2025)
  11. Paureel/LLM-SCI-GEN: Papers about scientific hypothesis generation with large language models (LLMs).. Available at: https://github.com/Paureel/LLM-SCI-GEN (Accessed: August 24, 2025)
  12. AI Tools for Literature Review: Complete Guide [2025]. Available at: https://anara.com/blog/ai-for-literature-review (Accessed: August 24, 2025)
  13. (PDF) Advancements in Generative AI: A Comprehensive Review of GANs, GPT, Autoencoders, Diffusion Model, and Transformers. Available at: https://www.researchgate.net/publication/380392586_Advancements_in_Generative_AI_A_Comprehensive_Review_of_GANs_GPT_Autoencoders_Diffusion_Model_and_Transformers (Accessed: August 24, 2025)
  14. Transparency and accountability in AI systems: safeguarding wellbeing in the age of algorithmic decision-making. Available at: https://www.frontiersin.org/journals/human-dynamics/articles/10.3389/fhumd.2024.1421273/full (Accessed: August 24, 2025)
  15. The heart of the internet. Available at: https://www.reddit.com/r/PhD/comments/17nqf59/ai_tools_for_literature_review/ (Accessed: August 24, 2025)
  16. AI Academic Writing Tool - Comprehensive AI Research Assistant. Available at: https://paperpal.com/ (Accessed: August 24, 2025)
  17. Leveraging Generative AI and Large Language Models: A Comprehensive Roadmap for Healthcare Integration. Available at: https://pmc.ncbi.nlm.nih.gov/articles/PMC10606429/ (Accessed: August 24, 2025)
  18. Editorial – The Use of Large Language Models in Science: Opportunities and Challenges. Available at: https://pmc.ncbi.nlm.nih.gov/articles/PMC10485814/ (Accessed: August 24, 2025)
  19. Guides at Oklahoma State University-Stillwater. Available at: https://info.library.okstate.edu/AI/tools (Accessed: August 24, 2025)
  20. The Synergy Between Machine Learning and Generative AI. Available at: https://medium.com/technology-nineleaps/the-synergy-between-machine-learning-and-generative-ai-90689d0b3534 (Accessed: August 24, 2025)
  21. Research integrity in the era of artificial intelligence: Challenges and responses. Available at: https://pmc.ncbi.nlm.nih.gov/articles/PMC11224801/ (Accessed: August 24, 2025)
  22. Create Visual Citation Maps for Academic Research. Available at: https://aiwisepicks.com/tools/litmaps/ (Accessed: August 24, 2025)
  23. A survey on Image Data Augmentation for Deep Learning - Journal of Big Data. Available at: https://journalofbigdata.springeropen.com/articles/10.1186/s40537-019-0197-0 (Accessed: August 24, 2025)
  24. AI research tools for for Academics and Researchers. Available at: https://www.litmaps.com/learn/best-ai-research-tools (Accessed: August 24, 2025)
  25. What is hierarchical clustering?. Available at: https://www.ibm.com/think/topics/hierarchical-clustering (Accessed: August 24, 2025)
  26. Classification vs Clustering - GeeksforGeeks. Available at: https://www.geeksforgeeks.org/ml-classification-vs-clustering/ (Accessed: August 24, 2025)
  27. Trust in AI: Evaluating Scite, Elicit, Consensus, and Scopus AI for Generating Literature Reviews. Available at: https://library.hkust.edu.hk/sc/trust-ai-lit-rev/ (Accessed: August 24, 2025)
  28. Generative AI in Academic Research: Perspectives and Cultural Norms. Available at: https://research-and-innovation.cornell.edu/generative-ai-in-academic-research/ (Accessed: August 24, 2025)
  29. Use of large language models as artificial intelligence tools in academic research and publishing among global clinical researchers. Available at: https://www.nature.com/articles/s41598-024-81370-6 (Accessed: August 24, 2025)
  30. https://www.sciencedirect.com/science/article/abs/pii/S2214579625000255. Available at: https://www.sciencedirect.com/science/article/abs/pii/S2214579625000255 (Accessed: August 24, 2025)
  31. Practical Considerations and Ethical Implications of Using Artificial Intelligence in Writing Scientific Manuscripts. Available at: https://pmc.ncbi.nlm.nih.gov/articles/PMC11838153/ (Accessed: August 24, 2025)
  32. https://www.tandfonline.com/doi/full/10.1080/15228053.2023.2233814. Available at: https://www.tandfonline.com/doi/full/10.1080/15228053.2023.2233814 (Accessed: August 24, 2025)
  33. Machine learning and generative AI: What are they good for in 2025?. Available at: https://mitsloan.mit.edu/ideas-made-to-matter/machine-learning-and-generative-ai-what-are-they-good-for (Accessed: August 24, 2025)