# Introduction to AI ## Facts - 95% or more of AI projects fail [5], [6], [9] - 85% or more of software projects fail [11], [12] - 90% of DevOps projects fail [13] - 268% Higher Failure Rates for Agile Software Projects [14] - Less than 1% of AI practitioners and researchers have an advanced degree in AI [TODO]. - 34% or more of of journal article authors admit to manipulating results [7]. - In general, the results of journal articles are irreproducible [7], [8]. ## Background As an AI/ML engineer, you should be willing to settle for “good enough” which is called _satisficing_ rather than trying to find the “best” model or approach. Scrum is a popular project management approach rather than a software development methodology [1]. A better approach is an iterative, agile feature-driven development (FDD) methodology where team members are able to work independently without the rigid constraints of Scrum [2]. Here are some lists of articles that cover some of the problems with **AI Engineering** which is required knowledge for an AI project to be successful. ## Key Concepts The student and practioner of AI needs a healthy dose of skepticism [4] The practioner of AI needs an expert knowledge of the following: - The history and issues that led to the "AI Winter". - The hype, misinformation, and alchemy that is common with AI. - The capabilities, risks, and limitations of AI. - AI Research: Reproducibility, Fabrication, and Falsification - The ethical implications, risks, and responsibilities of AI. - Why AI Projects Fail - How AI Projects are Different - Risks of agile approach for AI projects - AI alchemy = hype and false claims. - Problems with Cloud AI There are two guiding principles for AI engineering: 1. Occam’s Razor: The simplest algorithm that fits the data is usually the best. 2. No Free Lunch Theorem: There is no such thing as best, only good enough. ## Getting Started J. Holmes, "[How to Learn AI](https://pub.towardsai.net/how-to-learn-ai-1b9814ed3681)," Towards AI, Aug 24, 2023. J. Holmes, "[Getting Started with AI](https://pub.towardsai.net/getting-started-with-ai-f565c7877bee)," Towards AI, Aug 25, 2023. A. Pillai, [Begin with problems, sandbox, identify trustworth vendors — a quick guide to getting started with AI](https://venturebeat.com/ai/begin-with-problems-sandbox-identify-trustworth-vendors-a-quick-guide-to-getting-started-with-ai/), VentureBeat, Feb. 8, 2025. A. Romeu, [Hype v. Reality: 5 AI features that actually work in production](https://www.tinybird.co/blog-posts/ai-features-that-work), tinybird, April 2, 2025. M. Jayasinghe, [Top Strategies for Building Scalable and Secure AI Applications](https://thenewstack.io/top-strategies-for-building-scalable-and-secure-ai-applications/), The New Stack, Feb. 5, 2025. [Pragmatic AI Automation — Balancing Efficiency & Risk](https://blog.gopenai.com/pragmatic-ai-automation-balancing-efficiency-risk-d39c85333704), GoPenAI, Feb. 10, 2025. T. Shin, [4 Reasons Why You Shouldn’t Use Machine Learning](https://towardsdatascience.com/4-reasons-why-you-shouldnt-use-machine-learning-639d1d99fe11/), Towards Data Science, Oct. 5, 2021. [Recommended Resources](./Level-1/tips/ai_books.md) ## What is AI Engineering? [What is Artificial Intelligence Engineering?](https://www.sei.cmu.edu/our-work/artificial-intelligence-engineering/) A. Sandman, "[Is Agile dead in the age of AI?](https://sdtimes.com/agile/is-agile-dead-in-the-age-of-ai/)," SD Times, Aug. 1, 2025. R. Ramesh, "[AI’s Blind Spot: When Models Ignore Causal Relationships and Settle for Correlations]," Medium, Aug. 6, 2025. J. B. Michael and M. Orescanin, "[Developing and Deploying Artificial Intelligence Systems](https://ieeexplore.ieee.org/document/9789299)," IEEE Computer, vol. 55, no. 6, pp. 15-17, June 2022, doi: 10.1109/MC.2022.3166488. P. Ferguson, [AI Development vs Software Engineering: Key Differences Explained](https://medium.com/towards-data-science/ai-development-vs-software-engineering-key-differences-explained-0709633e81d2), Towards Data Science, Jan. 31, 2025. L. Ellen, [When OpenAI Isn’t Always the Answer: Enterprise Risks Behind Wrapper-Based AI Agents](https://towardsdatascience.com/when-openai-isnt-always-the-answer-enterprise-risks-behind-wrapper-based-ai-agents/), Towards Data Science, April 28, 2025. ## AI Best Practices Firsr, we need to understand why AI projects fail [5], [6], [9], [10], [15]. J. Dabass, [The $1 Trillion Mistake: Why 90% of AI Projects Will Fail](https://medium.com/tech-ai-made-easy/the-1-trillion-mistake-why-90-of-ai-projects-will-fail-d405dec77970), AI in Plain English, Aug. 21, 2025. K. Ahuja, [Why a $1.2M AI Project Failed (And How to Avoid the Same Mistake)](https://pub.towardsai.net/why-a-1-2m-ai-project-failed-and-how-to-avoid-the-same-mistake-c873235b5d1d), Towards AI, Aug. 24, 2025. I. Bernardo, [Why AI Projects Fail](https://towardsdatascience.com/why-ai-projects-fail/), Towards Data Science, June 6, 2025. Y. Kosarenko, [The majority of business analytics and AI projects are still failing](https://www.datadriveninvestor.com/2020/04/30/the-majority-of-business-analytics-and-ai-projects-are-still-failing/), Data Driven Investor, April 30, 2020. [Failed Machine Learning (FML)](https://github.com/kennethleungty/Failed-ML), GitHub, kennethleungty/Failed-ML. S. Mulligan, [AI trained on AI garbage spits out AI garbage](https://www.technologyreview.com/2024/07/24/1095263/ai-that-feeds-on-a-diet-of-ai-garbage-ends-up-spitting-out-nonsense/), MIT Technology Review, July 24, 2024. ## What are the Risks with AI? O. Enuku, [The Dark Side of Model Evaluation That Nobody Talks About](https://blog.gopenai.com/the-dark-side-of-model-evaluation-that-nobody-talks-about-b2050ccf0814), GoPenAI, Dec. 22, 2024. S. De Simone, [GenAI Increases Workloads and Decreases Productivity, Upwork Study Finds](https://www.infoq.com/news/2024/07/genai-hampers-productivity-study/), InfoQ, July 29, 2024. D. Ferraro, "[Uncontrolled Artificial Intelligence: Big Tech Companies Fail on Safety (Part One)](https://www.codemotion.com/magazine/cybersecurity/uncontrolled-artificial-intelligence-big-tech-companies-fail-on-safety-part-one/)," codemotion, Nov 25, 2024. M. Kumaran, "[AI Models Are Blackmailing Their Own Companies (And It’s Getting Worse)](https://pub.towardsai.net/ai-models-are-blackmailing-their-own-companies-and-its-getting-worse-c38cfb37d842?source=rss----98111c9905da---4)," Towards AI, July 11, 2025. D. Sculley, G. Holt, D. Golovin, E. Davydov, T. Phillips, D. Ebner, V. Chaudhary, and M. Young [Machine Learning: The High Interest Credit Card of Technical Debt](https://research.google.com/pubs/pub43146.html?authuser=2), SE4ML: Software Engineering for Machine Learning (NIPS 2014 Workshop), 2014. M. Troller, [Beware AI’s hidden costs before they bankrupt innovation](https://techcrunch.com/2023/12/27/beware-ais-hidden-costs-before-they-bankrupt-innovation/), techcrunch, Dec. 27, 2023. B. Cheatham, K. Javanmardian, and H. Samandari, [Confronting the risks of artificial intelligence](https://www.mckinsey.com/capabilities/quantumblack/our-insights/confronting-the-risks-of-artificial-intelligence), McKinsey Quarterly, April 26, 2019. ## Hidden Problems with Sofware Projects E. Gent, [Public AI Training Datasets Are Rife With Licensing Errors](https://spectrum.ieee.org/data-ai), IEEE Spectrum, Nov. 8, 2023. C. Y. Laporte, G. Verret, and M. Muñoz, "[A Software Project That Partially Failed: A Small Organization That Ignored the Management and Technical Practices of Software Standards](https://ieeexplore.ieee.org/document/10109288)," Computer, vol. 56, no. 5, pp. 138-144, May 2023, doi: 10.1109/MC.2023.3253979. B. Hubert, "[Why Bloat is Still Software's Biggest Vulnerability](https://spectrum.ieee.org/lean-software-development)," IEEE Spectrum, vol. 61, no. 4, pp. 22-50, April 2024, doi: 10.1109/MSPEC.2024.10491389. ## References [1]: I. Sommerville, Software Engineering 10th ed., Pearson, ISBN: 978-0133943030, 2015. [2]: P. Bourque and R. E. Fairley, [Guide to the Software Engineering Body of Knowledge (SWEBOK)](https://www.computer.org/education/bodies-of-knowledge/software-engineering), v. 3, IEEE, 2014. [3]: E. Alpaydin, Introduction to Machine Learning, 4th ed., MIT Press, ISBN: 9780262358064, 2020. [4]: S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 4th ed. Upper Saddle River, NJ: Prentice Hall, ISBN: 0-13-461099-7, 2021. [5]: B. M. Nedgu, “Why 85% of AI projects fail,” Towards Data Science Nov. 11, 2020. [6]: S. Reisner, “Why most AI implementations fail and what enterprises can do to beat the odds,” Venture Beat, June 28, 2021. [7]: J. F. DeFranco and J. Voas, “Reproducibility, Fabrication, and Falsification,” Computer, vol. 54 no. 12, 2021. [8]: M. Parashar, M. A. Heroux,and V. Stodden, "Research Reproducibility," Computer, vol. 55, no. 8, pp. 16-18, Aug. 2022, doi: 10.1109/MC.2022.3176988. [9]: S. Estrada, "[MIT report: 95% of generative AI pilots at companies are failing](https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/)," Fortune, Aug 18, 2025. [10]: A. DeNisco Rayome, [Why 85% of AI projects fail](https://www.techrepublic.com/article/why-85-of-ai-projects-fail/), TechRepublic, June 20, 2019. [11]: [Why 80% of Software Projects Fail and How to Avoid It](https://www.linkedin.com/pulse/why-80-software-projects-fail-how-avoid-bitsolutionss-y8m5f), LinkedIn, March 25, 2025. [12]: ekele, [Why 90% of Software Development Projects Fail](https://dev.to/ekele/why-90-of-software-development-projects-fail-and-how-you-can-avoid-it-3dnb), dev.to, Dec. 7, 2024. [13]: B. Gain, [Most DevOps Plans Fail, but Things Are Getting Better](https://thenewstack.io/most-devops-plans-fail-but-things-are-getting-better/), The New Stack, Nov. 30, 2021. [14]: J. Ali, [268% Higher Failure Rates for Agile Software Projects, Study Finds](https://www.engprax.com/post/268-higher-failure-rates-for-agile-software-projects-study-finds/), Impact Engineering, Engprax Ltd, ISBN-10: 106860574X, July 14, 2024. [15]: [Why AI projects fail, and how developers can help them succeed](https://www.infoworld.com/article/4010313/why-ai-projects-fail-and-how-developers-can-help-them-succeed.html) [Thirty Years, Five Technologies, One Failure Pattern: From Lean to AI](https://itnext.io/thirty-years-five-technologies-one-failure-pattern-from-lean-to-ai-628b8d7195a1) [One real reason AI isn't delivering: Meatbags in manglement](https://www.theregister.com/2025/12/24/reason_ai_isnt_delivering/)