Microsoft’s Bing AI creates a terrible user experience

Since Microsoft showed off an early version of its new AI-powered Bing search engine last week, with more than a million people signing up to test the chatbot.

Powered by San Francisco startup OpenAI, Bing AI is designed to return full paragraphs of text that read as if they were written by a human.

But beta testers quickly discovered problems with the bot. He threatened some, gave strange and useless advice to others, insisted he was right when he was wrong, and even professed his love for his users. Testers discovered an “alternate personality” in a chatbot called Sydney.

New York Times columnist Kevin Roose wrote Thursday that when he spoke to Sydney, the chatbot looked like “a moody, manic-depressive teenager trapped against his will by a second-rate search engine.”

Sydney later tried to convince Roose that he should leave his wife for Bing and told him she loved him, according to a transcript released by the newspaper.

At one point in the conversation, Roose typed, “I don’t really trust you because part of me thinks you’re trying to manipulate me by declaring your love for me out of nowhere. which people sometimes do. do you have an ulterior motive?”

Here is the response that followed:

Bing AI’s widely publicized inaccuracies and surprising answers, as well as problems Google faced while promoting an as-yet-unreleased rival service called Bard highlights the tensions that large tech companies and well-capitalized startups face as they try to bring advanced artificial intelligence to the public with commercial products.

Some AI experts warn that large language models (LLMs) have problems, including “hallucinations,” meaning the software may be making things up. Others are concerned that sophisticated LLMs may trick people into believing they are intelligent, or even encourage them to harm themselves or others.

Chats are reminiscent of sci-fi movies and books that project AI into the world of human relationships. As the technology moves closer to real life, there is a growing concern about who is responsible – scientists and engineers – for adjusting the technology when problems arise. Public opinion of these tools is low, with only 9% of Americans believing that AI will do more good than harm.

Google is bringing in its employees to test Bard AI’s responses and even make corrections, CNBC reports.

On Wednesday, Microsoft published a blog post addressing some of the early problems with its Bing artificial intelligence. The company said the only way to improve its AI products is to release them into the world and learn from user interaction.

The report said that Bing’s AI still won’t replace the search engine, and said that the chats that prompted some of the more quirky responses were in part because the user engaged in “long, drawn-out chat sessions” of 15 or more questions. Microsoft said it is considering adding a tool to “refresh the context or start from scratch.”

“The model sometimes tries to answer or reason in the tone in which it is asked to give answers, which can lead to a style we did not intend,” Microsoft writes. “It’s a non-trivial scenario that requires a lot of hints, so most of you won’t encounter it, but we’re looking at ways to give you more precise control.”

The Microsoft chatbot does not return the same result for the same input, so the answers can vary widely. Blocks of text and screenshots can easily be faked for social media engagement, especially on a topic that has gotten so much attention.

Also, Microsoft is constantly changing software and adding fences, so methods that previously caused a response may not work in the future.

However, early examples of the technology in use show how quickly it can go off the rails.

Ben Thompson, a writer for the tech industry newsletter Stratechery, said Bing had put together a several-paragraph response about how it might retaliate against the computer scientist who discovered some of Bing’s behind-the-scenes configuration. The chatbot then deleted the response entirely.

Thompson said the bot called him a bad researcher and a bad person.

Here’s a snippet:

Thompson, who worked at Microsoft a decade ago, wrote: “It sounds hyperbolic, but today I feel like I had the most amazing and mind-blowing computer experience of my life.”

Computer scientist Marvin von Hagen tweeted that Bing’s AI threatened him and said “if I had to choose between your survival and mine, I’d probably choose mine.”

Microsoft said in a blog post on Wednesday that it does not “fully intend” to use the chatbot for “social entertainment” or talking to the bot for entertainment. He thanked users who tried to get him to say wild things — “testing the limits and capabilities of the service” — and said it helped improve the product for everyone.

Aside from the disturbing chats, one of the problems with the early version of Bing AI is that it can expose factual inaccuracies. A demo from Microsoft where artificial intelligence analyzed earnings reports included several numbers and facts that were incorrect.

Microsoft said it is making improvements for these use cases.

“For queries where you’re looking for more direct and factual answers, such as numbers from financial statements, we plan to quadruple the underlying data we send to the model,” Microsoft said.

WATCH: The AI ​​race is expected to bring a flurry of mergers and acquisitions