NinjaAI.com

OpenAI - ChatGPT Global Usage Paper - Harvard & Duke Research Podcast by Jason Wade, Founder NinjaAI

NinjaAI.com

Briefing Document: Consumer Usage of ChatGPT - Trends, Economic Value, and Evolution

Date: September 15, 2025

Source: Excerpts from "How people are using ChatGPT | OpenAI" (National Bureau of Economic Research (NBER) working paper by OpenAI’s Economic Research team and Harvard economist David Deming)

Executive Summary:

This briefing summarizes the key findings from the largest study to date on consumer ChatGPT usage, based on a privacy-preserving analysis of 1.5 million conversations since its launch three years ago. The study highlights the increasing democratization of AI, with shrinking demographic gaps and rapid global adoption. ChatGPT is primarily used for everyday tasks, providing practical guidance, information seeking, and writing support. Crucially, the technology is demonstrating significant economic value creation in both personal and professional contexts, acting as a productivity tool, a decision support system, and a driver of uncaptured value in daily life.

Main Themes and Most Important Ideas/Facts:

1. Democratization of AI and Closing Usage Gaps:

• Shrinking Gender Gap: ChatGPT's early gender disparities have "narrowed dramatically, with adoption resembling the general adult population." By July 2025, the share of users with typically feminine names rose to "more than half (52%)" from 37% in January 2024.
• Global Accessibility and Rapid Growth in Developing Nations: ChatGPT has become a "broadly accessible global tool," with "especially rapid growth in low- and middle-income countries." By May 2025, adoption growth rates in the lowest income countries were "over 4x those in the highest income countries."
• Implication: This widespread adoption reinforces OpenAI's belief that "access to AI should be treated as a basic right—a technology that people can access to unlock their potential and shape their own future."

1. Primary Use Cases: Everyday Tasks and Practical Guidance:

• Focus on Practicality: "Three-quarters of conversations focus on practical guidance, seeking information, and writing."
• Common Work Task: Writing is identified as "the most common work task."
• Niche Activities: Coding and self-expression "remain niche activities."
• Usage Categorization (Asking, Doing, Expressing):Asking (49%): A growing and highly-rated category where "people value ChatGPT most as an advisor rather than only for task completion." This reflects users seeking advice and insights.
• Doing (40%): Task-oriented interactions, including drafting text, planning, or programming, where the model generates outputs or completes practical work. Approximately one-third of "Doing" usage is for work.
• Expressing (11%): Uses involving "personal reflection, exploration, and play."

1. Creation of Economic Value in Both Work and Personal Life:

• Dual Role as Productivity and Value Driver: Approximately 30% of consumer usage is "work-related," and 70% is "non-work," with both categories "continuing to grow over time." This underscores ChatGPT's "dual role as both a productivity tool and a driver of value for consumers in daily life."
• Decision Support and Productivity: ChatGPT "helps improve judgment and productivity, especially in knowledge-intensive jobs." This highlights its role beyond simple task automation, assisting in complex decision-making processes.
• Uncaptured Value: In some cases, ChatGPT is "generating value that traditional measures like GDP fail to capture," suggesting a broader economic impact beyond conventional metrics.
• Deepening Engagement: As users discover benefits, "usage deepens—with user cohorts increasing their activity over time through improved models and new use-case discovery."

NinjaAI.com

1. Executive Summary

Google has recently released Gemini 2.5 Flash Image, codenamed Nano Banana, a powerful new AI model designed for state-of-the-art image generation and editing. This briefing provides a comprehensive overview for developers looking to integrate Nano Banana into their applications using the Gemini Developer API. Key functionalities include image creation from text, image editing with text and image inputs, photo restoration, multi-image inputs, and conversational image editing. The tutorial emphasizes practical implementation steps, including API key generation, billing setup, SDK installation, and best practices for prompting.

2. Key Themes and Concepts

2.1. Introduction to Nano Banana (Gemini 2.5 Flash Image)

• Definition: Nano Banana is Google's latest model for image generation and editing, offering "state-of-the-art capabilities for creating and manipulating images."
• Purpose: It unlocks "a wide range of new applications" for developers.
• Access: While end-users can access it via the Gemini app, developers are encouraged to prototype and test prompts in Google AI Studio (aistudio.google.com).
• Model ID: For all API requests, the model ID to use is gemini-2.5-flash-image-preview.

2.2. Development Environment and Setup

• Google AI Studio: This is the primary playground for experimenting with AI models before coding, and the entry point for building with the Gemini API. Developers can use Nano Banana "free of charge within AI Studio." A direct link for a new session is ai.studio/banana.
• Required Tools:An API key from Google AI Studio.
• Billing set up for your Google Cloud project.
• The Google Gen AI SDK for Python or JavaScript/TypeScript.
• API Key Generation: Available in Google AI Studio by clicking "Get API key" and then "Create API key." This requires selecting or creating a Google Cloud project.
• Billing: While prototyping in AI Studio is free, using the model via the API is a "paid service." Billing must be enabled on the Google Cloud project associated with the API key.
• Cost: Image generation with Nano Banana costs $0.039 per image. This is based on an official pricing of "$0.30/1M input tokens and $30/1M output tokens," where "A standard 1024x1024px output image consumes 1290 tokens, which equates to $0.039 per image."
• SDK Installation:Python: pip install -U google-genai (and pip install Pillow for image manipulation).
• JavaScript/TypeScript: npm install @google/genai.

2.3. Core Capabilities and Functionalities

• Image Generation from Text: Users can "generate one or more images from a descriptive text prompt." The example provided creates a "photorealistic image of an orange cat with green eyes, sitting on a couch."
• Image Editing with Text and Image Inputs: The model allows users to "provide an existing image along with a text prompt to perform edits," noting its excellence "at maintaining character and content consistency from the input image." An example transforms a cat image into a "street-level view of the cat walking along a sidewalk in a New York City neighborhood."
• Photo Restoration: A "powerful application" of the model, enabling restoration and colorization of old photographs "with impressive results" using a simple prompt like "Restore and colorize this image from 1932."
• Multiple Input Images: The model can handle "multiple images as input for more complex editing tasks," demonstrated by applying a T-shirt from one image onto a person in another: "Make the girl wear this t-shirt. Leave the background unchanged."
• Conversational Image Editing: For "iterative refinement," developers can use chats sessions to "maintain context across multiple requests." This enables conversational editing, such as initially changing a cat to a Bengal cat and then, in a subsequent prompt, instructing it to "wear a funny party hat."