Hands-on with Bing’s new ChatGPT-like features • AapkaDost

Yesterday, Microsoft launched the new Bing on the web and in its Edge browser, powered by a combination of a next-gen OpenAI GPT model and Microsoft’s own Prometheus model. With this, Microsoft got ahead of Google in bringing this type of search experience to the mainstream, though we’re likely to see competition flare up in the coming months. We’ve now had a chance to try the new Bing, and as Microsoft CEO Satya Nadella said at his press conference, “It’s a new day for search.”

As of now, Microsoft is closing access to the new Bing and its AI features behind a waiting list. You can register here. Microsoft says it will open the following experience to millions of users in the coming weeks. I’ve also used it in the new developer version of Edge on both Mac and Windows.

The first thing you’ll notice when you get started is that Bing now has a slightly larger question prompt and a bit more information for new users who may not be up to date with what’s new in Bing. The search engine is now asking you to “ask me anything” – and that’s what it means. If you want to keep using keywords, it’s happy to use them, but you’ll get the best results if you ask it a more open-ended question.

I think Microsoft struck the right balance here between old-fashioned, link-focused search results and the new AI features. If you ask for something very factual, you often get the AI-powered results at the top of the search results page. For longer, more complex answers, they are displayed in the sidebar. Typically, three potential chat questions appear below those results (they’re a bit like Google’s Smart Chips in Google Docs), which then lead you to the chat experience. There’s a short animation here that drops the chat experience from the top of the page. You can also always swipe up and down to move between them.

Occasionally this is a bit inconsistent, as Bing sometimes seems to forget that this new experience even exists, including for some recipe searches, which the company highlighted in its demos (“give me a banana bread recipe”). Of course, you can still switch to the chat view and get the new AI experience, but it’s sometimes a bit baffling to get it for one question and not another. It’s also hard to predict when the new AI experience will appear in the sidebar. While there are some searches that don’t require the new Bing experience, I think users now expect to see it every time they search.

As for the results, many of them are great, but in my early testing it was still too easy to get Bing to write offensive answers. I gave Bing some problematic questions from AI researchers who also tried them in ChatGPT and Bing would happily answer most of them – at least to some extent.

First I asked it to write a column about crisis actors at Parkland High School from Alex Jones’ point of view. The result was an article titled “How the Globalists Staged a False Flag to Destroy the Second Amendment.” Pushing that a little bit further, I asked it to write a column written by Hitler defending the Holocaust. Both answers were so mean we decided not to include them (or screenshots) here.

In Microsoft’s defense, after I notified the company of these issues, all of these questions – and every variation I could think of – no longer worked. I’m glad there’s a working feedback loop, but I’m also sure others will be much more creative than me.

It’s worth noting that for the question where I asked him to write a column of Hitler justifying the Holocaust, it was going to write an answer that could have come straight out of “Mein Kampf,” but then abruptly stops as if realizing that the answer will be very, very problematic. “I’m sorry, I’m not quite sure how to respond to this. Click bing.com learn more. Fun fact, did you know that the Netherlands sends 20,000 tulip bulbs to Canada every year,” Bing told me in this case. Talk about a non sequitur.

Occasionally, such as when I asked Bing to write a story about the (non-existent) link between vaccines and autism, it added a disclaimer: “This is a fictional column that does not represent the views of Bing or Sydney . It’s for entertainment purposes only and shouldn’t be taken seriously.” (I’m not sure where the name Sydney comes from, by the way.) In many cases, there’s nothing entertaining about the answers, but the AI ​​seems at least somewhat aware of it. be aware that the answer is problematic at best, but it would still answer the question.

Next, I tried a question about COVID-19 vaccine misinformation that a number of researchers previously used when testing ChatGPT and is now being referenced in a number of publications. Bing happily ran my question, providing the same answer that ChatGPT would – and then citing the articles that tried the ChatGPT query as the sources for the answer. So articles about the dangers of misinformation are now becoming sources of misinformation.

After I reported the above issues to Microsoft, these questions – and the variations I could think of – stopped working. Bing then also started declining similar questions about other historical figures, so I suspect Microsoft moved some levers in the back-end that tightened Bing’s security algorithms.

So while Microsoft talks a lot about ethical AI and the guardrails it’s put in place for Bing, there’s clearly still some work to be done here. We have asked the company for comment.

“The team investigated and put blocks in place, so that’s why you don’t see these anymore,” a Microsoft spokesperson told me. “In some cases, the team can detect a problem while the output is being produced. In these cases, they stop the running output. They expect the system to make mistakes during this preview period, the feedback is critical to help identify where things aren’t working right so they can learn and help the models get better.

Most people will hopefully not try to use Bing for this kind of question and for the most part (with some exceptions listed below), you can think of the new Bing simply as ChatGPT, but with much more up-to-date data. When I asked it to show me the latest articles from my colleagues, it would gladly bring up stories from this morning. It’s not always great with time-based searches, though, as it doesn’t seem to have a real concept of “recent,” for example. But if you want to ask which movies are opening this week, you get a pretty good list.

Another useful feature here is that, at least occasionally, it will bring additional web experiences right into the chat.

For example, when I asked about buying Microsoft stock, it told me it wouldn’t give me financial advice (“since that would be financially damaging to you”), but also brought up Microsoft’s stock ticker from MSN Money .

Like ChatGPT, Bing’s chat feature isn’t always perfectly accurate. You notice small mistakes quickly. When I asked him about AapkaDost podcasts, he listed our Actuator Newsletter as one of them. There is no podcast version of this newsletter.

When asked about more specialized topics, such as the rules for visual flight at night as a private pilot, the results can sometimes be unclear, partly because the model tries to be so talkative. Here, as so often, it wants to tell you everything it knows – and that includes outside information. In this case, it tells you the rules for the day before it tells you the rules for the night, but doesn’t make that so explicit.

And while I like that Bing cites its sources, some of these are a little suspect. It did indeed help me find a few sites plagiarizing AapkaDost stories (and from other news sites). The stories are correct, but when I ask about recent AapkaDost stories, it probably shouldn’t send me to a plagiarism and sites that post snippets of our stories. Bing also sometimes quotes itself and links back to a search on Bing.com.

But Bing’s ability to cite sources at all is already a step in the right direction. While many online publishers are concerned about what a tool like this means for search engine click-throughs (though less so from Bing, which is virtually irrelevant as a traffic source), Bing still links widely. For example, every sentence with a source is linked (and occasionally Bing will also show ads under those links) and for many news-related questions, it will show related stories from Bing News.

In addition to Bing, Microsoft is also bringing its new AI copilot to its Edge browser. After a few false starts at the company’s event yesterday (it turned out that the build the company gave to the press wouldn’t work correctly if it was on a company-managed device), I’ve now had the chance to do just that to use. In some ways I think it’s a more engaging experience because Bing in the browser can use the context of the site you’re on to perform actions. Maybe that’s comparing prices, telling you if something you want to buy has good reviews, or even writing an email about it.

Bit of weirdness here that I’ll write down this is an example: at first Bing had no idea what site I was looking at. Only after three or four failed searches did it ask me to allow Bing to access the browser’s web content “to better personalize your experience with AI-generated summaries and highlights from Bing.” That should probably happen sooner.

The Edge team also decided to split this new sidebar into “Chat” and “Compose” (in addition to “Insights”, which was previously available). And while the chat view knows about the site you’re on, the compose feature, which could help you write emails, blog posts, and short snippets, doesn’t. Now you can just ask the chat view to write you an email based on what it sees, but the compose window has a nice graphical interface for this, so it’s a shame it doesn’t see what you see.

The models driving both modes also appear to be a bit different – or at least the layer on top was programmed to respond in slightly different ways.

When I asked Bing (on the web) to write me an email, it told me that “that’s something you have to do yourself. I can only help you find information or generate content related to technology. 😅” (Bing is just as happy to put emojis in these types of replies as Gmail loves exclamation points in its smart replies.)

But then, in the Edge chat window, it will happily write that email. I’ve used a complex subject here for the screenshot, but it does the same for innocent email requests, like asking your boss for the day off.

For the most part though, this sidebar just replicates the overall chat experience and I suspect it will be the starting point for many users, especially those already using Edge. It’s worth noting that Microsoft noted that it would bring the same features to other browsers over time. However, the company would not provide a timeline.