OpenAI connects ChatGPT to the Internet

Image credits: Open AI

OpenAI’s AI-powered viral chatbot, ChatGPT, can now browse the Internet in some cases.

OpenAI today launched plugins for ChatGPT, which extend the capabilities of bots by granting them access to third-party knowledge sources and databases, including the web. Available in alpha to waitlisted ChatGPT users and developers, OpenAI says it will initially prioritize a small number of developers and subscribers to its premium ChatGPT Plus plan before rolling out API access and at scale.

The most intriguing plugin is easily OpenAI’s first-party web browsing plugin, which allows ChatGPT to pull data from across the web to answer the various questions it is asked. (Previously, ChatGPT’s knowledge was limited to dates, events, and people before approximately September 2021.) The plugin fetches content from the web using the Bing Search API and shows all visited websites in the creation of an answer, citing its sources in the ChatGPT answers.

A web-enabled chatbot is a risky prospect, as OpenAI’s research has found. An experimental system built in 2021 by the AI ​​startup, called WebGPT, sometimes cited unreliable sources and was incentivized to handpick data from sites it expected users to find compelling even if those sources weren’t objectively the strongest. The since-disbanded BlenderBot 3.0 Metas also got access to the web and quickly went off the rails, delving into conspiracy theories and offensive content when prompted with certain text.

Image credits: Open AI

The live web is less curated than a static training dataset, and as a result, less filtered, of course. Search engines like Google and Bing use their own security mechanisms to reduce the chances of unreliable content making it to the top of the results, but these results can be fooled. They are also not necessarily representative of the entirety of the web. As an article in The New Yorker notes, Google’s algorithm prioritizes websites that use modern web technologies such as encryption, mobile support, and schema markup. As a result, many websites with otherwise quality content get lost in the shuffle.

This gives search engines a lot of power over the data that could inform responses of web-connected language patterns. Google has been found to prioritize its own services in Search, for example, answering a travel question with data from Google Places instead of a richer, more social source like TripAdvisor. At the same time, the algorithmic approach to search opens the door to bad actors. In 2020, Pinterest took advantage of a quirk of Google’s image search algorithm to surface more content in Google image searches, according to The New Yorker.

OpenAI admits that a web-enabled ChatGPT could perform all kinds of unwanted behavior, such as sending fraudulent and spam emails, bypassing security restrictions, and generally increasing the capabilities of bad actors who would defraud, mislead, or abuse others . But the company also says it has implemented several safeguards informed by internal and external red teams to prevent this. Time will tell if they are enough.

In addition to the web plugin, OpenAI has released a code interpreter for ChatGPT that provides the chatbot with a Python interpreter that works ina firewalled and sandboxed environment along with disk space. It supports upload files to ChatGPT and download results; OpenAI says it’s especially useful for solving math problems, performing data analysis and visualization, and converting files between formats.

Image credits: Open AI

A number of early contributors have built plugins for ChatGPT to join OpenAI, including Expedia, FiscalNote, Instacart, Kayak, Klarna, Milo, OpenTable, Shopify, Slack, Speak, Wolfram, and Zapier.

They are largely self-explanatory. The OpenTable plugin allows the chatbot to search restaurants for available reservations, for example, while the Instacart plugin allows ChatGPT to place orders from local shops. By far the most extensible of the bunch, Zapier connects with apps like Google Sheets, Trello, and Gmail to trigger a variety of productivity tasks.

To promote the creation of new plugins, OpenAI has open sourced a fetch plugin that allows ChatGPT to access snippets of documents from data sources such as files, notes, emails, or public records by asking questions in natural language.

We are working to develop plugins and bring them to a wider audience, OpenAI wrote in a blog post. We have a lot to learn and, with everyone’s help, we hope to build something that is both useful and safe.

Plugins are a curious addition to the ChatGPT development timeline. Once limited to the information within its training data, ChatGPT is, with plug-ins, suddenly much more capable and perhaps less at legal risk. Some experts accuse OpenAI of profiting from the unlicensed work ChatGPT was trained on; The ChatGPT dataset contains a wide variety of public websites. But plug-ins potentially solve this problem by allowing businesses to maintain full control over their data.