How should regulators think about "AI"?
Emily M. Bender・1 minute read
AI is essentially a marketing term for automation created in the 1950s to attract funding, with automation being a more accurate replacement term for clearer discussions on its implications and consequences for society. Various types of automation, from decision-making to language translation, can exhibit biases, promote mass surveillance, and serve corporate interests, emphasizing the need for accountability and ethical considerations.
Insights
- The term "AI" was initially a marketing strategy in the 1950s to secure research funding, emphasizing the need to replace it with "Automation" for clearer discussions on its impact, accountability, and beneficiaries.
- Automation encompasses various systems like decision-making, classification, recommendation, and translation, built on data and algorithms that may introduce biases, surveillance, and automation bias, ultimately serving corporate agendas.
Get key ideas from YouTube videos. It’s free
Recent questions
What is the origin of AI?
Marketing term in the 1950s
How can discussions on AI be clarified?
Replace with Automation
What are some types of automation?
Automatic decision systems, recommender systems, etc.
How are automation systems built?
Using training data and algorithms
What are the potential risks of automation?
Biases, mass surveillance, automation bias, corporate interests
Related videos
Summary
00:00
AI vs Automation: Marketing Term Evolution and Impact
- AI is a marketing term used to make automation seem advanced and autonomous, but it originated as a marketing term in the 1950s to attract research funding.
- The term AI can be replaced with Automation to clarify discussions on what is being automated, who benefits, how well it works, who is harmed, and who is accountable.
- Different types of automation include automatic decision systems, classification automation, recommender systems, access to human labor, translation automation, and synthetic media machines like chat GPT.
- Automation systems are built using training data and algorithms to reproduce patterns, but they can exhibit biases and promote mass surveillance, tapping into automation bias and serving corporate interests.
