From: allin
Recent data indicates a decline in the usage and interest levels of ChatGPT and other AI models like Bard [01:17:26]. This trend is observed in Google Trends and usage data from Similarweb [01:17:32].
Observed Usage Decline [01:17:43]
ChatGPT usage appears to be falling from a peak in May [01:17:43]. There was a modest increase from March to April, an even less modest increase from April to May, and then a decline from May to June [01:17:48].
Potential Reasons for Decline
Several factors are proposed for this declining usage:
- Educational Usage Cycle One primary theory suggests the decline is driven by educational usage, as many students used ChatGPT for essays and schoolwork [01:17:56]. School being out of session contributes to this drop [01:21:06].
- Novelty Wearing Off The initial curiosity factor may have played out, and the novelty of the product has worn off for many users who were initially testing its capabilities [01:20:51].
- Usage Vacuum Fulfillment These products launched during a “usage vacuum,” a period where consumers were “hankering for something really interesting and unique and new and novel” after the “total fart in the wind” that was VR [01:18:33]. This led to an explosion of initial usage [01:30:32].
Challenges and Skepticism about Current Interface
Some express skepticism about the current state and long-term viability of the chat interface for widespread consumer adoption:
- Limited Functionality Compared to Expectations The current system is seen as only a “50-60% copy” of what it could be, with less traffic [01:11:21].
- Lack of Novelty The main challenge for companies like Meta introducing AI is not just copying existing models but inventing a “de novo feature” that genuinely makes people want to use it [01:10:40].
- Product Difficulty and Limitations ChatGPT is described as hard to use, particularly for general consumers [01:22:00]. Most people paying $20/month for it is compared to paying for Google [01:22:07]. Plugins are a “pain in the ass,” and without them, there’s no new data past 2021, rendering it “Dead on Arrival as a product” for current information [01:22:11].
- Performance Issues One user experienced error messages when trying to use the mobile app for a simple search, forcing them to revert to Google [01:24:22]. Google is considered “much more performant” [01:25:12].
- User Interface Friction The chat interface, compared to the evolution of user interfaces (DOS, Windows, browser, scrolling feeds), requires “more input” for “less output” (mostly textual) [01:26:30]. This creates friction in user adoption [01:26:55].
“The only winner here is NVIDIA I think the loser here here are most of the VCS who pumped in hundreds of millions to billions of dollars in Rando stuff too early” [01:20:27]
Future Potential and Evolution of AI
Despite current declines, there’s optimism for the future of AI and chat interfaces, especially beyond consumer chat bots:
- Enterprise Use Cases The most exciting AI use cases are in enterprise software and healthcare [01:20:03]. These “real AI leaps” are still 18 to 36 months away from mainstream products [01:20:10].
- Shift to “Action Bots” The next “10x” feature for consumer-facing chat interfaces will be the move “from information retrieval to action” [01:22:53]. This means AI could not only provide information (e.g., best hotels) but also perform actions (e.g., book a hotel directly) [01:23:14]. Companies like Meta are expected to launch “Action Bots on WhatsApp and Instagram in the not too distant future” [01:23:33].
- Role in New Companies At the Snowflake Summit, 600 new startups applying for a contest were using an application of ChatGPT built on Snowflake data [01:23:51]. This shows “tremendous appetite for people to help Enterprises access that information and build much more seamless less discovery on top of it” [01:24:12].
- Native for New Generations Younger generations interact with their phones primarily through voice and chat, making this interface “so native to the way that they interact with the world” [01:27:32]. One speaker had a “brilliant conversation with ChatGPT about what happened with rates and inflation” using only his voice, finding it “way better than the experience I would have on Google” [01:27:52].
- Improving Performance While the product is currently hard to use, speakers expect improvements in accuracy, performance, speed, and interface, along with added features, to achieve the next level of usage [01:31:31].
Despite these future potentials, the current cost structure of ChatGPT ($20-40/month) is seen as a barrier for mass adoption, particularly for younger users [01:28:19].