OpenAI’s GPT Shop — Broadening the Commercialization of GPT Models
By Gary Zhu
On January 10, 2024, Open AI announced the launch of the long-awaited GPT Store, a platform for users to search and upload custom versions of Chat-GPT to the public. The effect was immediate: within a few weeks, millions of models from builders around the globe flooded the market. You might be wondering how to get in on the current GPT craze, and for those with a GPT-Plus subscription, the process is surprisingly simple.
To start, the chatbot will ask what GPT model you want to create. You can suggest endless answers, from an artist, a cooking assistant, to even a smart thermostat. From there, you have several customization options, which include toggling the ability to scrape the web for information and changing between text and image-based outputs through DALLE3.
However, the real magic comes from the option to upload custom files. Users can add up to 10 files of information that Chat-GPT will use when training your new model. This extra training data is the most important step, allowing organizations like Khan Academy, Canva, and Consensus to create impressive specialized models. And, for less tech-savvy users, it’s the perfect solution for any AI-related problems they may have.
As an additional benefit to joining the budding GPT Store ecosystem, OpenAI has promised creators a share of the profit that their model generates each month, similar to other platforms like Google Play and the Apple Store. This benefits program is slated to launch in Q1.
So, now that the floodgates have opened, what does this mean for the tech world? For one, CustomGPTs and the GPT Store are welcome steps toward open-source development. Although it isn’t quite as extensive as GPT alternatives like OpenGPT, which allow you to have control over virtually every step of the training process (LLM type, prompts, retrieval, etc), OpenAI has still given developers plenty of ways to tweak native ChatGPT to their liking.
The GPT Store also represents a cost-effective way for developers to demonstrate proof-of-concepts. Instead of sinking time and money into creating more sophisticated AI models that may fall flat in the marketplace, developers can gauge interest through the GPT Store for a relatively low cost. As a result, the skills required to build AI models may become less important than the creativity to explore new AI applications, since the GPT Store is an ecosystem where novel ideas are more important than technical know-how. In the future, the store will likely push developers to create increasingly specialized models, each catering to specific audiences in fields like research, education, medicine, and more.
Where does the GPT Store go from here?
Unfortunately, as revolutionary as the GPT Store could be, many glaring issues must be addressed.
For one, the GPT Store is only available to paying customers, which limits its audience compared to similar platforms like Google Play or the App Store. However, with a much smaller target market, you might expect that OpenAI has made up for it in terms of quality.
Sadly, this isn’t the case. The GPT Store is so unrefined that many of its problems are immediately apparent. For instance, take a look at this search for a “finance advisor”.
The top result isn’t even a finance app, it is a “tech support app”. And although the rest of the results seem to be related to our query, they are all incredibly vague and unspecific. Nothing tells the user what separates “PennyPal” from applications like “WealthWise”; all we have is a title and an icon to go off of, forcing us to try each model individually until we are satisfied. Also, since most of these apps go unvetted, the market is being flooded with low-quality bots.
On most other platforms, creators have the ability to add a short description to their app, greatly improving the customer’s understanding of the product itself. Additionally, users have the option to leave feedback for others in the form of reviews and ratings; all of which are features that are sorely needed in the GPT Store.
Finally, the GPTs in the store suffer from many of the same problems as the original, namely, censorship and liability issues. Asking for financial advice results in the chatbots being unable to give a clear answer, for fear of potential consequences. If you are interested in learning more, Nick Babich’s video goes more in-depth on these issues.
In the end, the GPT Store still has a long way to go before becoming a mature product. Yet, it gives us a glimpse of the future — a world where the power of AI is accessible to all and AI innovation is pushed to its very limits. The GPT Store is only one of OpenAI’s incredible accomplishments, and we can confidently expect bigger and better things to come.
