Chapter 10

yyyy

Reading Time: 4.21-5.11
Reading Task: Chapter 10
Summary of the Content:
AI is currently not competent for certain jobs. The advantages of AI are that it can reduce costs, is good at classification and predicting the future based on cumulative data. But at the same time, AI also has the drawback of being unable to "arrange" schedules. When someone sends an email to arrange a meeting time, AI still cannot fully understand the content of the email, confirm the schedule, and reply to the other party. At the same time, AI also lacks common sense. This common sense is common in human society, but because there is so much of it, it is difficult to input all of it as basic rules into AI.
Evaluation:
By comparing the advantages and limitations of AI, the author clearly Outlines the application boundaries of current AI technology. The core viewpoints focus on the technical advantages and shortcomings of AI
Meanwhile, the author adopts the method of comparative argumentation to enhance readers' understanding of the current situation of AI technology. Attribute the limitations of AI to the "tacit knowledge" characteristic of common sense rather than merely technical flaws.
Reflection:
The essence of AI's "intelligence" is statistics rather than human-like reasoning. We need to be vigilant against personifying technology. For instance, we cannot assume that AI can completely replace secretarial work. Moreover, common sense in human society is massive, dynamic and unstructured, which reveals the fundamental bottleneck of current AI's reliance on supervised learning.
When developing AI systems, human-machine collaborative design should be achieved. For example, the system can mark fuzzy requests that require human intervention. Simulate social interaction through reinforcement learning and gradually accumulate common sense instead of attempting to regularize it all at once.
For us, in the AI era, we need to enhance the capabilities that AI cannot replace, such as cross-domain association and ethical judgment, rather than competing with AI for rule-based tasks. For AI design, if human common sense is forcibly compressed into algorithmic rules, it may lead to the loss of social diversity.
In addition, the authors implicitly assume that "common sense must be input," but neuroscience points out that human common sense is not innate but develops through embodied cognition. This suggests that AI may need the ability to interact with the environment, rather than just data input.