OpenAI has formally launched its internet-browsing feature to ChatGPT, some three weeks after re-introducing the feature in beta after several months in hiatus.
ChatGPT, the generative AI chatbot that has taken the world by storm these past 12 months, has historically been limited to data up to September, 2021 — rendering it useless as a real-time search engine. However, OpenAI started bringing internet services to ChatGPT back in March, a move that always came with inherent risks, given that the live web isn’t curated in the same way a static training data set is — this potentially opened the doors to abuse by bad actors and good old-fashioned algorithmic chaos.
Then in May, OpenAI started rolling out web search via Bing, the search engine belonging to OpenAI’s corporate backer Microsoft, before extending access to the ChatGPT mobile app in late June. However, the new feature was swiftly pulled after it was discovered that ChatGPT was capable of displaying paywalled content.
Fast-forward to late September, and OpenAI started rolling the Browse with Bing feature out again, having fine-tuned how ChatGPT follows instructions laid out by content owners — essentially, it now promised to adhere to whatever a site-owner said in its Robots.txt file, similar to traditional web crawlers.
Now, Browse with Bing is officially available to all Plus and Enterprise subscribers, with no need to toggle their beta switch in settings.
In related news, OpenAI also transitioned DALL-E 3 into beta, a month after debuting the latest incarnation of the text-to-image generator.
DALL-E 3 sports integration with ChatGPT, meaning that users don’t have to think so carefully about their text-prompts when asking DALL-E to create an image — ChatGPT can do a lot of the heavylifting to ensure that the image the user gets, is closer to what they wanted.
But more than that, with DALL-E 3 embedded directly into ChatGPT, users will now be able to receive images as part of their text-based queries without having to switch between the two apps.
DALL-E 3 is available in beta now on the web and mobile, with users able to activate the feature by selecting “DALL-E 3 (Beta)” from the GPT-4 tab inside ChatGPT.
This all constitutes part of a broader expansion that is leading ChatGPT farther from a pure text-based generator, and down a path where audio and imagery are very much part of its remit.
Last month, OpenAI gave ChatGPT a mouth and ears with users able to have a verbal conversation with the chatbot, bringing together the worlds of Alexa-style voice assistants with powerful large language models (LLMs). For example, a user will be able to ask ChatGPT to invent and then narrate a bedtime story for their kid on the spot, though it’s maybe worth being on hand to see what it comes up with.
Additionally, ChatGPT will also allow users to search for answers using images, meaning that someone can upload a picture of an object and discover what it is or find similar items.
So today’s news very much fits into OpenAI broader push to make ChatGPT a fully integrated, real-time, multimedia generative search engine.