ChatGPT takes AI warfare to the next level with plugins

The war of the chatbots is gaining momentum with the artificial intelligence (AI) arena buzzing with activity like never before. If OpenAI unleashed nothing short of a tech revolution with its GPT-3-powered AI bot, ChatGPT, as a free public beta on November 30, 2022, and notched up 100mn users in two months, the pressure prompted rival Google to unveil its chatbot Bard on February 6, 2023. Other companies such as Microsoft to Adobe, Snapchat and Grammarly have rushed to show off and release similar generative AI capabilities in their own products.
ChatGPT is now taking the next leap, expanding its capabilities. If it was restricted to drawing information from its training data, which ran until 2021, last Thursday OpenAI announced a gradual roll out of plugins, in a move that significantly expands the chatbot’s functionality. The first wave of plugins, now available in alpha to select ChatGPT users and developers, allow ChatGPT to tap new sources of live data from the web, including third-party sources such as Expedia, Kayak and Instacart.
“Though not a perfect analogy, plugins can be ‘eyes and ears’ for language models, giving them access to information that is too recent, too personal, or too specific to be included in the training data,” OpenAI said on its website. For instance, ChatGPT can now pull up answers to questions such as how the box office sales of this year’s Oscar winners compare to those of other movies released recently. This new functionality is served up thanks to the browser plugin, which shows the sources the generative AI service is drawing information from before it spits out an answer.
“Plugins are very experimental still but we think there’s something great in this direction,” OpenAI co-founder Sam Altman wrote in a tweet Thursday. “It’s been a heavily requested feature.” Indeed there are marked imperfections in the results that services like ChatGPT produce. OpenAI’s own research has shown that a chatbot with access to the internet is a risky prospect. For instance, it can have a tendency to quote unreliable sources or, as OpenAI points out, “increase safety challenges by taking harmful or unintended actions, increasing the capabilities of bad actors who would defraud, mislead, or abuse others.”
But the proponents of these AI services have been focusing on the benefits. A video posted to Twitter by OpenAI co-founder Greg Brockman on Thursday demonstrates how to use ChatGPT’s Instacart plugin to assist with meal planning. The video shows ChatGPT recommending a chickpea salad recipe and then ultimately adds the required ingredients to Instacart for purchase with just a few prompts. A video posted on Expedia’s Twitter account shows how to leverage the Expedia plugin to essentially turn ChatGPT into your AI travel agent, helping travellers book flights and hotels. This is something ChatGPT previously couldn’t do, although it could identify places and create an itinerary. “You can install plugins to help with a wide variety of tasks. We are excited to see what developers create!” Altman wrote on Twitter.
Google is still lagging behind in this round of the AI war, as indicated last week by CEO Sundar Pichai when he said that Bard, despite testing by thousands of people, might still have a lot of problems. “As more people start to use Bard and test its capabilities, they’ll surprise us. Things will go wrong,” he wrote in a memo to employees, as reported by CNBC. Until now, Google has kept interactions with Bard limited to hand-picked testers, including about 80,000 Google employees, according to Pichai. But on Tuesday the company announced it was rolling out beta testing to thousands of users in the US and the UK who join the waitlist, with languages other than English to be added over time. The battle continues.