A Quick Guide to the AIpocalypse

🌈 Abstract

The article discusses the risks, harms, and threats posed by the increasing use of artificial intelligence (AI) in our daily lives. It covers topics such as AI governance, AI incidents, cybersecurity threats from AI, and the societal impacts of AI, including issues with the justice system, social media algorithms, and the environmental cost of AI.

🙋 Q&A

[01] The Risk, The Harm, and the ugly Threat

1. What are the different types of AI risks, harms, and threats discussed in the article?

• AI risks refer to everything that can go wrong with an AI system during the design, training, or production pipeline.
• AI harms are the harmful or near-harmful consequences of deploying and using an AI system.
• AI threats are threats caused by the malicious use of AI, such as the rise of AI-powered cyberattacks and deepfakes.

2. What efforts are being made to catalog and address these AI risks, harms, and threats?

• IBM has invested in AI Governance and created an AI Risk Atlas, similar to the Mitre CVE classification for cybersecurity vulnerabilities.
• The AI Incident Database is cataloging AI incidents, but its reach is limited to publicly published information.
• The author advocates for the development of open digital safety standards for AI, similar to what exists in the aviation industry and cybersecurity.

[02] The Boring AI Dystopia

1. What are the examples of harmful AI discussed in the "Unfair Justice System" section?

• The ProPublica study on Machine Bias demonstrates how AI systems can perpetuate institutional racism in the justice system, as they are only as impartial as the data they are trained on.
• AI algorithms used to calculate recidivism risk and determine sentences can reflect the biases present in the underlying data.

2. What other examples of harmful AI are discussed in the article?

• Facial recognition databases, such as the one maintained by the US State Department, raise concerns about privacy and the potential for mistaken identity issues.
• Social media algorithms designed to increase attention and retention can lead to addiction and negative effects, especially on children and teens.
• The use of AI-powered tools for propaganda and misinformation is a growing concern, especially during election periods.
• The high energy consumption and carbon footprint of training and using large language models like GPT are highlighted as an environmental concern.

[03] Terminator

1. What is the author's view on the development of autonomous weapons systems?

• The author cites a speech by the former president of the Association for the Advancement of Artificial Intelligence (AAAI), who argued that until strong robustness guarantees can be provided for human-robot systems, autonomous weapons should not be deployed on the battlefield.
• The author states that weaponized AI exists and is advancing, even if it may seem innocuous at first.