Oh dear, ChatGPT can once again browse the internet
OpenAI announced on Twitter that it's brought back ChatGPT's ability to browse the internet so the popular chatbot can provide you with "current and authoritative information." Answers to queries will have more up-to-date information while giving the proper linking to sources. The news comes after a recent update gave ChatGPT the ability to have a back-and-forth conversation with you.
What this means for ChatGPT users is that you can now ask something like, "What's the best gaming PC?" In response, it should give you a list along with where it got the information, properly crediting the sites it used to answer your question—something it struggled with previously.
Before the update, the pool of information for ChatGPT was limited to data before September 2021, so you were out of luck if you were looking for answers that required current-day information. The new feature is called 'Browse with Bing' and can be switched on in ChatGPT-4.
The update only seems to apply to ChatGPT-4 for ChatGPT Plus and Enterprise users. The free version that uses ChatGPT-3.5 is still running off the old, outdated model. OpenAI did say that it will expand the feature to all users soon.
ChatGPT can now browse the internet to provide you with current and authoritative information, complete with direct links to sources. It is no longer limited to data before September 2021. pic.twitter.com/pyj8a9HWkBSeptember 27, 2023
In July, OpenAI launched a beta version of ChatGPT's internet browsing capabilities but pulled the feature based on immense negative feedback from various online publishers who feared that the chatbot could somehow bypass website paywalls. The company vowed to"do right by content owners."
One of the more significant concerns about ChatGPT's newfound access to the internet is that the bot will start scraping copyrighted or protected materials online to train its model.
In August, OpenAI launched GBTbot, a tool that farms the entire internet for data to incorporate into ChatGPT-4 and upcoming ChatGPT-5 language learning models. Amazon and The New York Times have restricted their data from being scraped by OpenAI's tools. In fact, about 25% of the top 1,000 websites have blocked GPTBot from doing its thing.
We've already seen that modded versions of ChatGPT have been used to bypass some of these guardrails. But I'm sure this time everything will be fine and totally safe. Right? I mean, what could go wrong?