By Mia Chitishvili
On November 6th of 2023, OpenAI announced the launch of their product, GPTs, a brand new feature where users can create their own adaptable versions of ChatGPT, the AI chatbot many of us know.[1] Tech support chatbots, scientific research, financial and medical advice, game guides, pet care advice and more - with GPTs, the opportunities are seemingly endless. The introduction of GPTs is an incredible way for people to build their own virtual assistants and share these creations with others. But this new feature may also create a new face in the digital economy, and open new methods to more efficient research.
How do custom GPTs work? What purpose do they serve for both creators and users?
A GPT - generative pre-trained transformer - is a type of artificial intelligence program that is able to identify, understand and as a result, produce response content close to that possible for an intelligent human. GPTs are trained on Large Language Models (LLM), an AI model trained on great amounts of data[2]. The very first GPT, GPT-1, was developed and introduced in 2018 by OpenAI, the organisation that owns ChatGPT. Through more developments of their models, the GPT has become the most eminent blueprint for generative artificial intelligence.
Custom GPTs were introduced as a way for users to design their own chatbots to their preferences, and can train them to develop expertise on a subject the customer needs. In contrast with the ChatGPT models (ChatGPT, Google Bard, Copilot, etc.) on the market today, in a custom GPT, users benefit from the following functionalities: supplying the documentation of the subject they are developing the chatbot on, and defining the conversation style of it.
Users do not need to have any coding experience beforehand when making their chatbot. The creation of chatbots solely depends on what data they are supplied with.
Not long after the announcement was posted on their blog, OpenAI’s Youtube channel posted a video summarising how custom GPTs look and work like.[3]Here we are given the adorable example of a chatbot specialising in puppy care.
Once the custom GPT is tailored, the creator will be allowed to feature it in an online marketplace, which is currently undergoing development. They can either be priced or free to use. In its introductory blog, OpenAI mentioned that there will be leaderboards displaying the most popular chatbots. They will also highlight which chatbots they find the most helpful across different categories.
How will this new feature influence the digital economy?
One of the main takeaways from this feature is that users do not need any coding experience to make their own custom GPT, as mentioned before. Training the chatbot will only require the provision of data on the subject of interest. Data can be any of the following, or a combination thereof: websites, documents, images. If custom GPTs development needed coding knowledge, it would certainly be less appealing for a larger audience.
Users who create their custom GPTs will have the option to price their chatbots when they place it in the marketplace. With people collaborating over custom GPTs through a global online marketplace has a potential to maximise the benefits of AI for humans.
We can compare custom GPTs with another great invention - Apple Inc.’s App Store. When the App Store was released in 2008, it gave them a new, global way to reach out to customers and make profits across borders. Early iPhones only had a few standard applications developed by Apple themselves. The launching of the App Store enabled individual developers of the smartphone to develop and deploy their applications[4] with many benefits. The stark difference between these two revolutionary products, though, is that developers who wished to sell their app in this store were required to have coding experience - not in the case of custom GPTs, as said above. Converse with the chatbot, supply data, give it instructions, and you have just possibly found a way to create a value and perhaps a profit out of it. This is a much bigger opportunity and brings down the barrier for those who would like to participate, but may not have technical expertise.
The introduction of custom GPTs may also open a new pathway into further development of human intelligence and research. The freedom to communicate with these highly developed chatbots will potentially increase the speed and decrease the time necessary for research.
But one should stay vigilant when using this kind of advice. Services and products harnessing the power of AI are still in their infancy, and have a long way to gain the trust. If a user acts on a faulty response generated by the chatbot, its creator may face consequences. In an interview published in a CNBC article with Hod Lipson, a robotics professor at Columbia University, he gave his own caution about working with custom GPTS[5]:
“If you’re creating an AI that’s giving counselling for people in distress and it says the wrong thing, things could go badly very quickly. Additionally, you should be careful if you’re building an AI chatbot that gives medical or tax advice, for instance, since users could face severe real-life consequences if they act on incorrect advice your chatbot has given them.”
- Hod Lipson, CNBC.
To sum up, the introduction of the custom GPTs feature is a huge leap forward in the history of artificial intelligence. It is thanks to the genius of OpenAI’s innovators that helped define this pioneering moment in the tech field. But note that we are only witnessing the early years of AI, and we will have to see what new developments will come in the coming years.
[1] OpenAI, “Introducing GPTs”, OpenAI blog, published Nov. 6, 2023, <https://openai.com/blog/introducing-gpts> accessed Dec. 25, 2023
[2] “Generative Pre-Trained Transformer”, Encord, <https://encord.com/glossary/gpt-definition/> accessed Dec. 26, 2023
[3] OpenAI, “Introducing GPTs in ChatGPT”, Youtube, published Nov. 7, 2023, <https://www.youtube.com/watch?v=CmP3XXwKJ60> accessed Dec. 26, 2023
[4] The Unofficial AppleKeynotes Channel, “Apple Special Event 2008 - App Store Introduction”, Youtube, published Jan. 2, 2022, <https://www.youtube.com/watch?v=WUrzjLjP4UQ> accessed Dec. 28, 2023
[5] DeVon C. “You can now build your own version of ChatGPT—here’s what to know”, CNBC “Make It”, Nov. 11, 2o23, <https://www.cnbc.com/2023/11/11/openai-will-let-users-build-custom-versions-of-chatgpt.html#> accessed Dec. 28, 2023
Image bibliography:
OpenAI, "Introducing GPTs", published Nov. 6, 2023, OpenAI, <https://openai.com/blog/introducing-gpts> accessed Feb. 22, 2024