Will Generative AI Follow Development Path of the Internet?

In many ways, the development of the internet provides a model for understanding how artificial intelligence will develop and create value.

For example, the internet’s evolution from a platform primarily for finding and learning to a platform for doing and creating is a path AI could follow. Also, the internet flipped technology adoption on its head.

Where pre-internet technology adoption typically was "enterprise first, followed by small business; followed by consumer adoption," in the internet era we often saw the reverse: consumers first, then adoption of those tools by businesses.

The "information at first; transactions later" pattern should also hold.

Initially, the internet served mostly as a repository of information, where users would search for data, read articles, or view static content. Over time, it shifted into a space where people could interact, transact, create, and collaborate in real-time.

That is in many ways an apt description of the state of generative AI as well. Early search (pre-Google) really was not very useful, in large part because the range of sources was highly limited, with output pretty much limited to text and static images.

Social media had not been invented; payment mechanisms were clumsy and limited; support for video and audio input and output was limited as well. The ability to personalize was rudimentary. 

In contrast, even early generative AI already provides lots of value, but has far to go. The natural language interface is rapidly expanding to include multimedia input and output; highly-personalized content creation with important degrees of contextual awareness. 

But we still are at a relatively early stage of development.

The arrival of AI “agents” and functions provides an example. Right now, generative AI is mostly “ask a question, get an answer.” In the future it will expand to include “do things on my behalf, without an active prompt on my part.”

Feature	Advancements	Examples and Models	Impact on Agent Capabilities
Contextual Awareness	Improved ability to retain, understand, and apply context across interactions; models can remember conversation flow or details over time.	Google Gemini 2, OpenAI ChatGPT, Anthropic Claude 3	Enables smoother, more human-like conversations that adapt based on user history and context.
Multimodal Input Handling	Capability to process and respond to multiple input types, including text, images, video, and even voice.	OpenAI GPT-4 with vision, Google Gemini 2, Meta’s LLaMA 2	Supports richer interactions where users can input questions or prompts across media types.
Enhanced Code Execution	Models can run code to calculate and retrieve specific outputs, especially helpful for technical users.	OpenAI’s Codex, Google Gemini 2’s Python execution capabilities	Useful for problem-solving in technical domains; increases AI’s utility in engineering and data science.
Task-Specific Agents	Models can act as task-specific agents (e.g., research assistants, customer service agents) with focused functions.	Anthropic Claude, Microsoft Copilot, Jasper AI	Tailored functionality helps automate specialized workflows and repetitive tasks across business applications.
Real-Time Interaction and Responsiveness	Models have improved speed and conversational fluidity, supporting real-time, hands-free interactions.	Google Gemini Live, OpenAI ChatGPT with voice capabilities	Facilitates conversational agents that can handle continuous, free-flowing dialogue more naturally.
Voice and Speech Recognition	Integration of natural language processing with voice recognition for spoken commands and responses.	OpenAI Whisper, Google Assistant with Gemini 2	Enhances accessibility and allows for hands-free operation, useful for on-the-go or mobile settings.
Complex Problem Solving	Ability to handle complex problem-solving using mathematical reasoning, programming, and logical deduction.	Google DeepMind's AlphaCode, OpenAI’s GPT-4 with expanded logic	Enables advanced technical applications, from coding to scientific research support.
Interactive and Connected App Ecosystem	Deep integration within app ecosystems allows AI to manage tasks across multiple apps seamlessly.	Google Gemini 2 with Google apps, Microsoft Copilot	Helps users complete cross-app workflows, such as scheduling meetings and drafting emails, autonomously.