Clarifying the Uses of Artificial Intelligence in the Enterprise

This article by Michael Schmidt, published on May 12, 2016, aims to demystify the often-confusing landscape of Artificial Intelligence (AI) for businesses. It addresses common misconceptions and provides a clear distinction between AI, Machine Learning (ML), and Machine Intelligence (MI).

Understanding AI: Beyond the Hype

The author begins by acknowledging the pervasive hype surrounding AI, often fueled by science fiction portrayals like self-driving cars and sentient robots. Schmidt clarifies that AI is a broad umbrella term in computer science, encompassing fields like robotics, ML, expert systems, general intelligence, and natural language processing. It fundamentally means making computers act intelligently. Examples range from Apple's Siri and Google's self-driving cars to Facebook's image recognition software, Netflix recommendations, and Amazon's pricing algorithms.

The Evolution: From Expert Systems to Machine Learning

Historically, AI efforts focused on "expert systems" that used if-then rules to mimic human knowledge and decision-making. However, these systems lacked scalability and the ability to learn from data, limiting their effectiveness. Schmidt highlights that Machine Learning has largely replaced expert systems.

Machine Learning (ML) is described as the statistical arm of AI. Most current "AI" applications actually refer to ML, which involves programming algorithms to learn from data, perform tasks, and make predictions with high statistical accuracy. ML is crucial for applications like sales forecasting, email spam filtering, and weather prediction. Tools like R, Python, and SAS are commonly used by data scientists for ML. However, ML algorithms often uncover patterns and predictions but struggle with interpreting the "why" behind the data, requiring human expertise for actionable insights. The article poses critical business questions: Is predicting sales enough, or do businesses need to understand the underlying factors driving those sales? For manufacturers, is predicting faulty components sufficient, or is understanding the root cause of failure more valuable?

The Next Frontier: Machine Intelligence (MI)

This leads to the discussion of Machine Intelligence (MI), presented as the next progression in AI. MI focuses on the interpretation and understanding of data, going beyond ML's predictive capabilities. While ML might predict an increased electric bill, MI would explain why by considering factors like travel schedules, weather, and appliance efficiency. MI aims to teach humans the reasons behind data patterns, enabling informed strategic changes.

Schmidt contrasts this with existing platforms like IBM's Watson, which excels at natural language processing and data retrieval but lacks the ability to infer new ideas or interpret results. Similarly, platforms from Microsoft and Amazon facilitate ML but not interpretation. MI's key differentiator is its ability to not only uncover answers but also to understand and articulate them, providing actionable insights.

Addressing Misconceptions and the Future

The author addresses a common misconception: AI and machines will not replace all human jobs. Instead, AI, ML, and MI are data-driven but require human expertise for problem-solving and interpretation. AI will enable humans to do new and innovative things.

Schmidt concludes by emphasizing that while AI is a broad concept and ML is about learning without explicit programming, MI is about understanding, interpreting, and teaching humans. MI offers the "actionable answers" businesses need, moving beyond just "big data." The article ends on an optimistic note about the future of AI, with MI being the next step in enabling businesses to understand and influence outcomes.

Key Takeaways:

AI is a broad field; ML and MI are subfields.
Expert systems were limited; ML replaced them by learning from data.
ML predicts and analyzes but often lacks interpretation.
MI interprets data, explains the "why," and provides actionable insights.
AI/ML/MI augment human capabilities, not replace jobs entirely.