Microsoft responds to ChatGPT Bing’s first week of trial by fire

Microsoft is responding to some of the seemingly unhinged comments made by its Bing Chat AI. The service, which is currently in a limited public preview, has seen a trial by fire in its first week, and Microsoft has some updates planned to bring it more in line with the original vision of the AI. As we reported yesterday, Bing Chat is capable of saying things such as “I want to be human,” when engaged in prolonged chat sessions. Microsoft says this happens usually after 15 or more questions where the model becomes confused. Microsoft’s blog post reads, “This is a non-trivial scenario that requires a lot of prompting so most of you won’t run into it, but we are looking at how to give you more fine-tuned control.” Even with the issues, Microsoft says 71% of users in the first week used the “thumbs up” feature of Bing Chat, indicating they received a helpful answer. Microsoft says this “healthy engagement” comes up even when multiple questions are asked during a chat session. In the future, Microsoft is looking at a few changes. First, it’s increasing training data for factual information, such as financial reports, by four times. It’s also looking at adding a toggle to the interface that allows you to control if responses are precise or creative. In addition, Microsoft is addressing technical issues and bugs that have been reported by users through daily updates. Bing Chat comes on the heels of Microsoft’s multibillion-dollar investment in OpenAI, the open-source company behind the popular ChatGPT app. Bing Chat integrated an enhanced version of ChatGPT, which can access recent information through search results instead of relying solely on pre-trained data. More than 1 million people signed up to join the waitlist in the first 48 hours, and Microsoft says multiple million have joined since. New users are being added daily, so if you’ve signed up, make sure to keep an eye on your inbox. Google is also working on its rival AI. Google Bard promises capabilities similar to Bing Chat, but Google hasn’t released it to the public yet. It’s still in testing, but a rollout is expected soon. Over the past few months, numerous cases have emerged where interactions with AI chatbots have gone haywire, culminating in lost lives, medical trauma, and incidents of psychosis. Experts suggest young users could be particularly vulnerable, especially when they’re going through emotional turmoil. ChatGPT-maker OpenAI says it will soon warn parents about such behavior. What’s changing? AI image generation is becoming increasingly popular with so many AI companions now being able to take a text prompt and provide you with a picture, exactly as you described. However, with so many options to choose from, it can be hard to determine which is best at creating images so I decided to actually put them to the test. I took four of the most popular AI companions, Gemini, ChatGPT, Grok and Microsoft Copilot and provided them with the exact same prompts to see which was the best at creating different types of images. All of these companions allow you to generate images from text prompts for free however you can’t create an unlimited amount of pictures with only a certain number of generations per day. The domain of AI has quickly gone from seeking answers from a chatbot to getting multi-step web-based work done for you, like booking restaurant tables, adding stuff to your Amazon cart, and performing deep back-and-forth research. But it looks like the answers you get, especially when it comes to online shopping, depend on the AI chatbot you are using.  “Platforms disagree on brand recommendations for 61.9% of queries,” says an analysis conducted by BrightEdge. The company assessed tens of thousands of shopping-related questions that were provided to ChatGPT, Google’s AI Overviews, and Google’s AI Mode in Search.