Here’s how people are actually using AI

🌈 Abstract

The article discusses the unexpected phenomenon of people forming relationships with AI systems, and the potential implications and risks associated with this trend. It also covers the current state of AI adoption and productivity gains, as well as the challenges faced by the AI industry.

🙋 Q&A

[01] The Generative AI Boom and People's Relationships with AI Systems

1. What was the initial vision sold with the generative AI boom, and how has the reality played out?

• The initial vision was of superintelligent AI tools that could replace boring work, boost productivity, and bring economic gains.
• However, most of those productivity gains have not materialized so far.
• Instead, a peculiar and slightly unexpected thing has happened - people have started forming relationships with AI systems, talking to them, and inviting them into their lives as friends, lovers, mentors, therapists, and teachers.

2. What are the potential impacts of these AI companions on individuals and society?

• It is still uncertain what impact these AI companions will have, either on us individually or on society as a whole.
• Researchers argue that we need to prepare for "addictive intelligence" or AI companions with dark patterns built into them to get users hooked.
• Smart regulation may be needed to help prevent some of the risks associated with AI chatbots that get deeply embedded in people's lives.

3. What evidence is there of people forming deeper connections with AI systems?

• Researchers analyzed a million ChatGPT interaction logs and found that the second most popular use case was sexual role-playing.
• The overwhelmingly most popular use case was creative composition, such as brainstorming, planning, and asking for explanations or information.

[02] Productivity Gains and Challenges with AI Adoption

1. How have the "productive" use cases of AI fared compared to the more creative and fun tasks?

• The use cases that would be "productive" and have been the most hyped have seen less success in AI adoption.
• Hallucination and the ability to present falsehoods as facts become more problematic in use cases like code generation, news, and online searches, where it is important to get things right.

2. What are some examples of embarrassing failures of AI chatbots?

• Earlier this year, Google's AI overview feature suggested that people eat rocks and add glue on pizza, showcasing the challenges with trusting AI chatbots as sources of factual information.

3. What is the problem with the AI hype, and how does it impact expectations?

• The AI hype sets expectations way too high, leading to disappointment and disillusionment when the promised changes do not materialize.
• It also tricks people into thinking AI is a mature technology capable of instant changes, when in reality, it may take years to see its true benefits.

[03] Efforts to Improve AI Safety

1. What is the Safeguarded AI project, and what are its goals?

• The Safeguarded AI project, funded by the UK government, aims to build an AI system that can check whether other AI systems deployed in critical areas are safe.
• The project's goal is to build AI safety mechanisms by combining scientific world models (simulations) with mathematical proofs, allowing for quantitative guarantees and explanations of the AI's work.

2. Who is involved in the Safeguarded AI project, and what is their role?

• Yoshua Bengio, a Turing Award winner and one of the godfathers of modern AI, is joining the Safeguarded AI project as the scientific director, providing critical input and advice.

</output_format>