AI Has A Giant, Expensive Flaw

🌈 Abstract

The article discusses the author's concerns about the impact of AI on creative jobs, particularly the writing profession. It then summarizes the findings of a recent MIT study that suggests AI is currently too expensive and not cost-effective enough to replace most jobs, including those involving computer vision and self-driving technology.

🙋 Q&A

[01] AI's Impact on Creative Jobs

1. What are the author's concerns about AI's impact on creative jobs?

• The author is worried about AI threatening their career as a writer, as writing is one of the few creative jobs already being impacted by AI.
• The author has also taken a negative stance against certain AI applications, such as Tesla's self-driving system, and is concerned about being proven wrong if the technology works well.

2. How does the MIT study address the author's AI-related concerns?

• The MIT study has calmed the author's concerns about AI, both in terms of job-taking and potential "Skynet" scenarios.
• The study found that AI is too expensive and will remain so for a long time, meaning most jobs are safe from being replaced by AI, including the author's writing job.

[02] MIT Study on AI's Commercial Viability

1. What was the focus of the MIT study?

• The MIT researchers wanted to assess the commercial viability of AI, specifically computer vision AI combined with robotics and computer systems.
• They calculated whether AI was more cost-effective than humans in 1,000 visually assisted tasks across 800 occupations.

2. What were the key findings of the MIT study?

• Only 23% of workers' wages could be replaced by AI for less expenditure than hiring humans.
• The vast majority of jobs are safe from computer vision AI, as only the most simple and constrained tasks could be partially replaced.
• Even with a 20% annual decline in AI costs, it would take decades for AI to become cost-effective for businesses to automate.

3. How does the study address the potential for AI to replace jobs like taxi and truck driving?

• The study explicitly excluded jobs that require "prohibitively complex supplementary systems," such as piloting aircraft or driving vehicles.
• The author argues that the AI needed to replace these jobs is exponentially more complex and expensive than the AI assessed in the study, making them commercially unviable.

4. What are the potential development issues with AI for complex tasks like self-driving?

• The author cites Elon Musk's comments about the need for massive amounts of training data to make self-driving AI work reliably.
• However, the author argues that computer vision AI and deep learning algorithms don't necessarily get predictably better with more data, and can even pick up incorrect trends, making the development process unpredictable and costly.

5. How does the MIT study's findings impact the author's views on the AI revolution?

• The study vindicates the author's opinion that Tesla's claims about imminent fully autonomous vehicles are "shilling snake oil."
• It also indicates that the AI revolution is still far in the future, and the costs associated with AI are unlikely to drop significantly anytime soon.