Kevin Kelly’s 2016 AI Predictions Reviewed Against Current Industry Reality
A Hacker News discussion examines Kevin Kelly’s 2016 book The Inevitable and its specific forecasts regarding Google’s AI trajectory.
Key points
- Kelly predicted Google’s main product would shift from search to artificial intelligence by 2026.
- The author foresaw an oligarchy of two or three large, general-purpose cloud-based commercial intelligences.
- Kelly argued that Google uses search data to train AI models rather than using AI to improve search.
A discussion on the Hacker News front page recently analyzed Kevin Kelly’s 2016 book The Inevitable, specifically reviewing his predictions about artificial intelligence a decade later. The thread highlights excerpts where Kelly forecasted that Google’s primary product would transition from search to AI by 2026, a timeline that aligns with current industry shifts.
Kelly’s writing anticipated an oligarchy consisting of two or three large, general-purpose cloud-based commercial intelligences. He suggested that future AI systems would resemble infrastructure services like Amazon Web Services rather than standalone consumer applications. The author also predicted that the bulk of artificial intelligence would consist of special-purpose software brains interacting with users in conversational tones.
Historical Context and Google’s Strategy
Kelly recounted a conversation with Google co-founder Larry Page around 2002, before the company’s initial public offering. At the time, Kelly questioned the sustainability of free web search, but Page replied that the company was actually building an artificial intelligence. This perspective contrasted with the common view that search was merely a utility for finding information on the early internet.
The author noted that Google has since acquired thirteen AI and robotics companies, including DeepMind. Kelly argued that these acquisitions are not primarily intended to improve search capabilities, which constitute 80 percent of Google’s revenue. Instead, he posited that Google uses search interactions to train its underlying AI models, reversing the conventional assumption about the relationship between the two technologies.
Data as Training Mechanism
According to Kelly, every user query and click serves as a training signal for Google’s deep-learning algorithms. When a user searches for an image and selects a specific result, such as an Easter Bunny, they are teaching the system visual recognition patterns. This continuous feedback loop allows the company to refine its artificial intelligence with each interaction, turning billions of daily queries into educational data points.
Kelly referenced a 2015 quarterly earnings call where Google CEO Sundar Pichai described AI as a core transformative force. Pichai stated that the company was rethinking everything it did through the lens of artificial intelligence. This executive statement from 2015 supports Kelly’s earlier thesis that search infrastructure was being leveraged to build a broader AI capability rather than just optimizing search results.
What should you do?
Security and technology professionals should monitor how search behaviors influence AI model training data. Understanding that user interactions serve as training signals helps clarify potential biases in large language models. Organizations should review their data usage policies to ensure compliance with evolving AI regulations.
Background: Large language models
Large language models are a specific type of neural network trained on text. Generative AI is the broader umbrella term for any system that creates new content, including images and code. Not all generative AI uses large language models, but most modern text generators do.
What to do and how to stay safe: Kevin Kelly
- Review internal data handling policies to understand how user interactions may train third-party AI models.
- Monitor vendor disclosures regarding data usage for artificial intelligence training purposes.
- Assess organizational exposure to potential biases in AI systems trained on public search data.
- Stay informed about regulatory developments concerning AI training data and user privacy.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What did Kevin Kelly predict about Google's main product in 2016?
Kelly predicted that by 2026, Google’s main product would not be search but artificial intelligence.
How many AI companies has Google acquired according to the source?
According to the source material, Google has bought 13 other AI and robotics companies in addition to DeepMind.
What did Larry Page say to Kevin Kelly in 2002?
Larry Page told Kelly, “Oh, we’re really making an AI,” when questioned about the sustainability of free search.
