10 Controversial Things I Believe About AI That I Shouldn’t Say Out Loud

🌈 Abstract

The article discusses the author's controversial opinions about AI, including its potential impact, the risks and downsides of large language models, the role of open-source AI, and the hype surrounding AI.

🙋 Q&A

[01] AI's Potential Impact

1. What are the author's views on the potential impact of AI?

• The author believes that AI could be more impactful than the internet, printing press, or fire, but cautions against prematurely labeling it a "revolution" as this can raise unrealistic expectations and lead to AI winters.

2. Why does the author think premature labeling of AI as a revolution is damaging?

• The author believes it can unnecessarily raise expectations for those who don't understand how AI works and how it could fail, while leaving no room for absorption of the failures typical of scientific endeavors.

[02] Relationships with AI

1. What is the author's view on people developing relationships with AI?

• The author believes that people would be better off working on their emotional skills to improve their relationships with other humans, rather than developing relationships with AI.

[03] Downsides of Large Language Models (LLMs)

1. What are the author's concerns about large language models (LLMs)?

• The author believes that LLMs are a net cost for the world, citing social, economic, and scientific concerns:
  • Social: Where does the data come from, who does the RLHF, and who will lose their jobs?
  • Economic: Are the millions invested in LLMs better spent elsewhere?
  • Scientific: Are other approaches to AI being stifled by the LLM gold rush?
• The author argues that the purpose LLMs serve does not compensate for the toll they create.

[04] Forced Use of AI

1. What is the author's view on forcing AI into one's life?

• The author believes that if using AI doesn't come naturally to someone, they are better off without it. The author suggests that people may not have any problem where AI can help them, and that's fine.

[05] Generative AI

1. How does the author view the impact of generative AI?

• The author believes that generative AI (AI systems intended to create data) has not indisputably improved the world from the onset, and is mostly being used for trivial tasks like cheating on homework or creating marketing copy. The author thinks this is unlikely to change soon because "humans won't change."

[06] Open-source vs. Closed-source AI

1. What is the author's view on the open-source vs. closed-source debate for AI?

• The author believes the world would gain more from AI being only open-source than from it being only closed-source, including the possibility of malicious actors using it to cause harm. The author suggests that centralization and private control are better in specific but not super common situations, such as high-tech weaponry.

[07] Imbalance in AI Research Priorities

1. What is the author's concern about the focus of AI research?

• The author believes that people who are concerned with studying medium-term (e.g., disinformation) and long-term problems (e.g., existential risk) of AI do so from a privileged position, as they are not affected by the short-term problems (e.g., bias, copyright violations, job losses) that affect the majority of people studying AI.

[08] Vested Interests in AI Hype

1. What is the author's view on the motivations behind AI hype?

• The author believes that everyone who hypes AI, including companies, researchers, journalists, influencers, and bloggers, has a vested interest (personal, professional, financial, or reputational) in the hype existing as an end in itself. This applies equally, or even more, to those who are anti-hype.

[09] Potential Misuse of Advanced AI

1. How does the author think people would use advanced AI capabilities?

• The author believes that even if AI allowed humanity to reach the stars, most people would only use the technology to fulfill their most basic drives and needs, such as making money (e.g., spam sites), minimizing effort (e.g., homework cheating), and getting sexual gratification (e.g., deepfake porn).

[10] Lack of Public Interest in AI

1. What is the author's view on the public's interest in AI?

• The author believes that AI enthusiasts live in an echo chamber, as most people don't care about AI and won't ever care about it, even once it is ubiquitously integrated into their daily lives or once AGI is achieved and it redefines society's rules.