Where Does AI Go From Here?

🌈 Abstract

The article discusses the current state of AI adoption and the future direction of AI, focusing on the rise of AI agents and their potential impact on various industries and use cases.

🙋 Q&A

[01] State of the Union: AI Tool Adoption

1. What is the current state of AI tool adoption based on the data presented in the article?

• ChatGPT has seen rapid growth, reaching 1 million users in just 2 months, faster than other popular apps like TikTok and Instagram.
• Other AI tools, particularly Google's Gemini, are catching up to ChatGPT in terms of website traffic.
• However, the article notes that AI tools still have a small share of overall web traffic compared to dominant platforms like Google, YouTube, and Facebook.
• The article presents data showing that AI performance has improved significantly against human benchmarks, with most categories now exceeding human levels except for competition-level mathematics and multitask language understanding.
• Adoption of AI tools like ChatGPT is still relatively low, with less than 50% of 18-29 year olds having used it.
• There is also widespread consumer fear and skepticism around AI, with concerns about job loss, extinction, and social discord.

2. What insights does the article provide about the current state of AI tool adoption?

• AI adoption is happening at an exponential rate, with rapid improvements in AI performance and capabilities.
• However, AI tools still have a small share of overall web traffic and consumer adoption, suggesting we are still in the early stages of AI adoption.
• There is significant consumer fear and skepticism around AI, which the article attributes to AI being a "black box" to most people.

[02] What's Next: The Age of the Agents

1. What is the "Age of the Agents" that the article discusses?

• The article predicts we are entering the "Age of the Agents", where AI-powered software agents will anticipate our needs and execute tasks automatically on our behalf.
• This builds on a prediction made by Bill Gates in 2017 about the rise of "AI agents" that would revolutionize how we interact with computers.
• Examples given include an AI agent automatically placing supply orders for a small business owner, without the owner having to do anything.
• The article suggests this "Age of the Agents" is starting to emerge in 2024, with companies testing business models, distribution channels, and use cases for this new era.

2. What challenges do AI upstarts face in this new era?

• The article identifies two key challenges for AI upstarts:
  1. Distribution - Competing with the distribution power of incumbents like Microsoft Office and Google Workspace will be difficult.
  2. Defensibility - If an AI tool is just an API call to OpenAI, it lacks defensibility against incumbents.

3. How does the article suggest AI tools can expand human capabilities?

• The article provides the example of how Microsoft Word expanded the user base for word processors, making them accessible to everyone beyond just secretaries.
• Similarly, the article suggests AI tools can unleash latent demand by empowering people to do things they couldn't do before, like graphic design or animation.
• The article cites the example of how AI could drastically reduce the cost and time required to produce animated films, allowing many more studios to create their own content.

[03] Pricing, Distribution, and Use Cases

1. What pricing models are emerging for enterprise AI tools?

• The article notes that companies are shifting to enterprise use cases for AI and consequently increasing pricing:
  • OpenAI rolled out ChatGPT Enterprise at $60 per seat
  • Perplexity charges $40 per month or $400 per year per employee for its Pro service
  • Some companies are exploring pricing based on "minutes worked" or productivity gains rather than traditional SaaS models

2. How does distribution pose a challenge for AI upstarts?

• The article points to the example of Gemini catching up to ChatGPT in website traffic, highlighting that built-in distribution like Google's is tough to beat.
• It suggests that AI productivity tools will struggle to compete with the distribution power of incumbents like Microsoft Office and Google Workspace.

3. What new use cases for AI are emerging beyond just consumer applications?

• The article highlights enterprise use cases, such as AI agents for businesses (e.g. Sierra from former Salesforce co-CEO Bret Taylor).
• It also mentions net-new use cases like Character, where people are spending an average of 2 hours per day interacting with chatbots.
• The article suggests AI tools can expand human capabilities in areas like graphic design, animation, and content creation by making these tasks more accessible and affordable.