Deep Learning

The Automation Question

A useful way to begin studying artificial intelligence is not to ask whether a computer thinks exactly like a person. Instead, ask what kind of human intellectual work the computer is being designed to perform. Artificial intelligence is the effort to automate intellectual tasks normally performed by humans. This definition describes a broad goal, not one particular technique.

Artificial intelligence is broader than machine learning. AI includes machine learning and deep learning, but it also includes approaches that do not involve learning. Therefore, describing deep learning as if it were the whole of AI reverses the relationship between the concepts.

Two Routes to Intelligent Results

There are two useful routes for imagining how a system can produce an intelligent-looking result. In symbolic AI, programmers describe knowledge and decisions directly through explicit rules. In machine learning, the system is built for a problem where writing all the necessary rules would be extremely difficult, so machine learning provides another approach.

When Rules Stop Scaling

Explicit rules work well when the logic of a problem can be described clearly. The difficulty changes when a problem is complex and fuzzy. For image classification, speech recognition, and language translation, it proved intractable to work out all the explicit rules needed to describe the solution. Machine learning arose as an approach for these kinds of problems.

The Five-Part Rise

Deep learning's recent surge is presented as the result of several contributing factors rather than one isolated cause. The complete list named in the source is hardware, data, algorithms, a new wave of investment, and democratization of deep learning.

• Hardware
• Data
• Algorithms
• A new wave of investment
• Democratization of deep learning

From Conditions to Growth

The five factors should be read together as a broad explanation of why deep learning has recently grown. Hardware, data, algorithms, investment, and democratization are all part of the account. The source does not present them as a ranked sequence, so the safest interpretation is that they are several contributing conditions rather than five steps that must occur in a fixed order.

Common Misreadings

• Treating artificial intelligence and deep learning as synonyms.

  AI is the broader field. It includes machine learning and deep learning, as well as approaches that do not involve learning.

  Fix: Describe deep learning as an approach within machine learning and machine learning as one approach within AI.

• Assuming symbolic AI and machine learning are identical.

  Symbolic AI depends on explicit rules handcrafted by programmers. Machine learning arose as an approach for problems where writing all the rules is extremely difficult.

  Fix: Contrast handcrafted rules with the machine-learning approach used for complex, fuzzy problems.

• Claiming that explicit rules are always inferior.

  The source says symbolic AI is suitable for problems whose logic can be described clearly.

  Fix: Match explicit rules to clearly described logic, and recognize why complex, fuzzy problems create a different challenge.

• Explaining the recent rise of deep learning with only one factor.

  Hardware is named, but the source presents a list of five contributing factors.

  Fix: Include hardware, data, algorithms, a new wave of investment, and democratization.

• Inventing detailed mechanisms for the five factors.

  The source does not rank the factors or provide detailed mechanisms for each one in this section.

  Fix: Name the complete list and avoid claiming a ranking or mechanism not supplied by the source.

Practice the Classification

1. Artificial intelligence is the effort to automate intellectual tasks normally performed by humans. It is a broad field that includes machine learning, deep learning, and approaches that do not involve learning. Symbolic AI uses explicit rules and suits clearly described logic. Machine learning arose as an approach for complex, fuzzy problems such as image classification, speech recognition, and language translation, where writing all the rules proved intractable. The recent surge of deep learning is attributed to hardware, data, algorithms, a new wave of investment, and democratization. A complete explanation names all five factors without inventing unsupported rankings or mechanisms.