OpenAI, the Microsoft-backed company that developed ChatGPT, announced support on Thursday for plugins designed to allow companies to more easily embed chatbot functionality into their products.
The first plugins have already been created by Expedia.nl, TaxNote, Instacart, KAYAK, Klarna, Milo, Open Table, Shopify, limp, Speak, TungstenAnd Zapiersaid OpenAI.
“We are gradually rolling out plugins in ChatGPT so we can study their real-world usage, impact and security and alignment issues – all of which we need to get right to achieve our mission,” San Francisco-based OpenAI said in a blog post.
The OpenAI plugins are tools specially designed for language models that help ChatGPT access current information, perform calculations, or use third-party services.
Users can click and upload any number of the currently offered plugins. For example, users of grocery delivery service Instacart can upload the ChatGPT plugins and start using the natural language processor to get things like restaurant recommendations, recipes, ingredients for a meal, and the total calories of that meal.
An example of a question might be “I am looking for vegan food in San Francisco this weekend. Can you give me a great restaurant tip for Saturday, as well as an easy meal recipe for Sunday (just the ingredients)? Calculate the calories for the recipe using WolframAlpha.”
“Users have been asking for plugins since we introduced ChatGPT (en many developers are experimenting of similar ideas) because they unlock a wide range of possible use cases,” said OpenAI.
The generative AI developer said he plans to gradually roll out access more widely as he learns more from plugin developers, ChatGPT users and after an alpha period.
Arun Chandrasekaran, a leading vice president analyst at Gartner, said one of the challenges with applications like ChatGPT and their underlying AI models (such as GPT-4) is their static nature due to the gap between enterprise training cut-off dates and actual application, release dates of models.
“Through a plug-in ecosystem, ChatGPT can now process more real-time information from a range of curated sources. They are announcing both first-party plug-ins, which will allow ChatGPT to connect to Internet resources for providing more up-to-date -date information as well as external sources like Expedia, OpenTable,” said Chandrasekaran. “On the other hand, this also increases the attack surface and potentially more latency domains in the architecture.”
In its blog, OpenAI acknowledged the “significant new risks” associated with enabling third-party tools through ChatGPT.
“Plugins hold the potential to address several challenges associated with large language models, including ‘hallucinations,’ tracking recent events, and access (with permission) to proprietary information resources,” OpenAI said.
The company said it only allowed a limited number of plugin developers who had been on a waiting list documentation they could use to build a plugin for ChatGPT. That way, OpenAI can continue to monitor any adverse effects of the plugins.
Chirag Shah, a professor of data science and machine learning at the University of Washington, said that while application developers have already been able to use ChatGPT’s API to adapt it for use with products, plugins will make the task much easier.
“APIs require technical knowledge. As with social media, there are other ways to access services through subscriptions. Plugins make it easy for people to deploy ChatGPT without much effort,” said Shah. “They won’t work for every company. They are aimed at a specific target group.”
OpenAI admitted that its GPT model for major languages (LLM) is limited in what it can do today because it has not been trained with the latest up-to-date information for the myriad of applications on the Internet. For example, the LLM computer algorithm it has been trained on has billions of parameters, but they are not specific to what Expedia, for example, wants to show to its users.
Currently, the only way GPT-4 can learn is through the input of training data by a user organization. For example, if a bank wants to use ChatGPT for its internal employees and external customers, it needs to include information about the company so that when users ask questions to the chatbot, it can provide bank-specific answers.
The plugins make it easier for ChatGPT’s LLM to access company information that is specific to products, details that might otherwise be too recent, personal, or specific to be included in the training data.
“In response to a user’s explicit request, plugins can also enable language models to perform safe, constrained actions on their behalf, increasing the usability of the system overall,” according to OpenAI. “We expect open standards to emerge to unify the ways applications expose an AI-centric interface. We’re working on an early attempt at what such a standard might look like, and we’re looking for feedback from developers interested in building with us.”
Copyright © 2023 IDG Communications, Inc.
