Selling with Data #49 - What does a decline in ChatGPT users mean?
Last week's Washington Post article, ChatGPT loses users for first time, shaking faith in AI revolution, starts off with this opening paragraph.
"The number of people visiting AI chatbot ChatGPT’s website and downloading its app fell for the first time since its launch in November, a sign that consumer interest in artificial intelligence chatbots and image-generators may be beginning to wane."
It seems a little early to extrapolate a single month demand decline to mean that interest in AI is declining. That said, June did see the first dip in users and it’s worth paying attention to what might be happening.
The article sites a few reasons for the decline:
Another data point is Google Trends, a source of user search data. Google Trends shows that while the interest in AI continues to grow, the search interest in "ChatGPT" started to fall around April and continues to decline.
Fortune had a more rational explanation, in their article ChatGPT suddenly ‘isn’t booming anymore,’ Google A.I. researcher says—and kids are the big problem. Fortune noted that while the traffic to ChatGPT's website fell by 9.7% in June, the decline was steeper in the U.S., with a 10.3% month-on-month decline. The number of unique visitors to ChatGPT also fell by 5.7% from the previous month.
Fortune suggests that the drop is based on students being on summer vacation in many parts of the world, and they are replacing homework with playing outside or on gaming consoles. The article says, "There’s only one explanation: a significant portion of students using ChatGPT to do their homework. It’s one of the most common uses for ChatGPT, according to Sam Gilbert, a data scientist and author. He found queries for things like “ChatGPT essay,” “ChatGPT math,” and “ChatGPT history” are the second most prominent type of Google Search, aside from job application-related searches."
As someone tweeted- "producing homework" remains the #1 application of LLMs. That is funny.
My view is that the decline in ChatGPT users is probably a conflation of a few things including student summer vacation. I also think that there is a bigger trend that a single large model is going to be replaced with purpose built smaller models. This is already happening and as confident and knowledge about models in communities like Hugging Face continues to grow, the interest in ChatGPT will continue to decline. ChatGPT has proven to be a great general model, which is great for consumer language. For many use cases, though, having the best model for that function will ultimately win.
ChatGPT will go down in the history books as the catalyst that started the AI Revolution, the same way that AOL and Netscape are credited as the catalysts that started the consumer cloud revolution. Similar to the cloud revolution, AOL and Netscape were replaced with cheaper, better, faster alternatives that emerged as AI.
Gartner has outlined a model of technological maturity to understand in the different AI phases and how to adopt AI in processes and activities in a business. (source).
ChatGPT did amazing things for Level 1 Awareness and creating initial active projects, for both consumer and enterprise use cases. I believe that just like the Gartner AI Maturity Model suggests, we are shifting to Level 3 and operationalizing AI to solve real business use cases through governed and responsible AI.
Please leave a comment on what you believe is causing the decline in interest in ChatGPT and how you see AI evolving through the rest of this year.
Good selling.
AI Strategist for Financial Services Companies | Keynote Speaker | Digital Transformation Leader | Salesforce Optimization Expert
1 年60% of the people I speak with have tried ChatGPT as a novelty. They asked it a few questions, were not impressed with the results and then wonder what all the hype is about. I firmly believe that those who refuse to learn prompt engineering will never find out what I know and that is that ChatGPT can do anything if…big if…you learn how to prompt correctly. What does prompting correctly mean? It means that you get the output you expected from the AI engine.
Senior Managing Director
1 年Ayal Steinberg Very insightful. Thank you for sharing.?
Executive Resource
1 年I think it's a combination of hallucinations (some of which are hysterical, by-the-way) and students' seasonal decline of usage.
IBM Worldwide Sales Enablement, Mergers & Acquisitions (M&A)
1 年Agree with the suggestion that usage may be down because kids are on summer break. That’s what I’m seeing in my little world.
Sales Leader | A.I. Strategic Partnerships
1 年Gartner's 2022 hype cycle - https://www.gartner.com/en/articles/what-s-new-in-the-2022-gartner-hype-cycle-for-emerging-technologies Foundation models sit at the top of Gartner's hype cycle. As we inevitably enter the trough of disillusionment, partners and clients will focus on launching enterprise-ready AI in a trusted and transparent way. Here are two examples of IBM research innovations that accelerate enterprise AI adoption: - CODA-prompt: https://research.ibm.com/publications/coda-prompt-continual-decomposed-attention-based-prompting-for-rehearsal-free-continual-learning - FairIJ: https://research.ibm.com/publications/fair-infinitesimal-jackknife-mitigating-the-influence-of-biased-training-data-points-without-refitting