Abstract:
Adam Fry, head of consumer products at OpenAI, recently disclosed a series of ChatGPT’s latest operating data in an interview with technology media Big Technology. According to his introduction, ChatGPT’s weekly active users have reached 1.2 billion, with about 200 million new users added in the past two months; its search query volume is currently about one-fifth of Google’s. In addition, ChatGPT generates 1.5 billion pictures every week, and the total number of weekly active users of voice input and voice dialogue has reached 250 million.
The data was disclosed during an OpenAI event in New York on November 2. Adam Fry said that the growth rate of ChatGPT is further accelerating. According to him, ChatGPT’s weekly active users have grown from 1 billion to 1.2 billion in the past 60 days, which is equivalent to adding about 100 million users every month.
One of the most noteworthy figures is the growth of ChatGPT’s search business. Fry revealed that OpenAI conducted statistics on search queries within the platform and compared them with Google. It was found that the number of search queries currently processed by ChatGPT has reached about 20% of Google.
He explained that the search here is not limited to keyword retrieval in the traditional sense, but refers to the situation where after the user raises a question, ChatGPT needs to call a network search to obtain the latest information. Every time a user asks a question that requires an Internet search, it is counted as a search.
This change also reflects a shift in the way users use ChatGPT. Initially, users primarily used it for writing assistance and programming, tasks that were not typical of traditional search engines. However, in the past year, more and more users have begun to use ChatGPT to query people, locations, products, and other content that requires real-time information. Issues in the financial and health fields have gradually become important usage scenarios.
Fry believes that one of the biggest differences between ChatGPT and traditional search engines is its ability to retain users' long-term context. For example, when a user asks about the height of the Eiffel Tower in Paris, the system can not only answer the specific number, but also provide information that is more in line with actual needs based on the user's Paris travel plans previously mentioned. This ability to continuously understand the user's context and goals enables ChatGPT to gradually transform from a one-time question and answer tool into an assistant capable of participating in long-term tasks.
Voice capabilities have also seen significant growth. Currently, the total number of weekly active users of ChatGPT voice input and voice dialogue reaches 250 million, accounting for about one-fifth of the total weekly active users. Fry said that voice is no longer just a novel feature that users occasionally try, but has gradually become a daily use method. Some users simply ask a quick question via voice, while others engage in longer conversations.
He attributed this growth to the improved interactive experience brought by the new generation of voice models, especially the ability to have more natural two-way communication. The system can respond appropriately during the conversation and can handle interruptions and additional information from the user. Additionally, background voice mode allows users to continue conversations even when they close the app or move around their home.
In terms of personal intelligent assistants, Fry also talked about Dots, launched by OpenAI on September 29. Unlike traditional chatbots, Dots is positioned as a personal assistant that can run continuously and proactively provide help. For example, users can ask it to remind them at the end of each day whether they have missed something important, and the system can combine information such as emails and messages to help users review the day's to-do tasks.
However, OpenAI is still very cautious in its permission management of Dots. Fry said that personal assistants need to handle matters of practical significance on behalf of users, so users must build enough trust in it. Allowing AI to perform actions on the user's behalf involves a higher degree of responsibility and risk than simply answering questions.
He also admitted that Dots is currently a product that consumes high computing resources. OpenAI hopes to first ensure that it can provide good enough services, and then gradually optimize operating costs. As new generation models continue to improve efficiency, it is expected that computing requirements will be reduced while maintaining service quality in the future.
As for the business model of Dots, Fry said that OpenAI is establishing revenue sources through various methods such as subscriptions, advertising and additional usage credits. He revealed that ChatGPT had previously announced more than 50 million paid subscribers, and this number has grown significantly since then. The advertising business achieved over $1 billion in annual recurring revenue (ARR) within 200 days of launch.
In terms of product strategy, Fry also confirmed that OpenAI plans to integrate ChatGPT’s current Chat and Work modes into a unified interface.
He admitted that requiring users to choose between Chat and Work has become a point of confusion during product use. Users often don’t know when to use normal chat and when to switch to Work mode that can perform complex tasks.
OpenAI hopes to use the model itself to determine user intentions and automatically distinguish between simple questions and answers and complex tasks that require long execution, thereby canceling this mode switch. Fry said the adjustment is progressing and is expected to be rolled out soon.
Image generation is also an important growth area for ChatGPT. Currently, users generate 1.5 billion pictures through ChatGPT every week. Fry said that initially users mainly used the image generation function for entertainment, such as making retro-style personal photos and family posters, but now the scope of applications has expanded to work scenarios, including making restaurant menus, promotional posters, birthday posters and various visual charts.
OpenAI is also continuing to reduce the response time of image generation, hoping to further expand the use scenarios by increasing the speed. Fry agreed with users who reported that the styles of restaurant menus and signage generated by AI were too similar, and promised to feed relevant opinions back to the product team.
He is particularly optimistic about the development direction of interactive interfaces. Fry believes that in the past few years, AI chat products have mainly relied on text for interaction, but in the future they will gradually shift to a richer visual experience. For example, when a user asks about the offside rule via voice, the system should be able to immediately generate a diagram; if the user still cannot understand it, the diagram can be readjusted based on subsequent feedback.
He believes that this kind of interactive content that can be continuously updated along with conversations will play an important role in the field of education and knowledge learning.
Finally, when talking about public concerns about artificial intelligence, Fry admitted that AI is changing the way work is organized, and this change is indeed disturbing to many people. But he believes that as more people use AI personally and understand the help it can provide in daily life, the public's understanding of this technology will gradually change.
Judging from these data and product plans, OpenAI is promoting the transformation of ChatGPT from a traditional chat robot to a comprehensive personal assistant. Search, voice, image generation and automated tasks are gradually integrated. In the future, users may no longer need to distinguish whether they are asking questions, searching for information, or asking AI to perform a complex task. Instead, they only need to express the final goal and let the system decide how to complete it.
Comments