Microsoft defends Bing's AI mistakes as it faces 'our share of challenges'

Updated

A week after Microsoft (MSFT) released its much-hyped AI-powered version of Bing, the tech giant is already playing defense for its chat bot.

Despite controversy about inaccurate answers and strange behavior, the tech giant says the majority of users are reporting positive experiences.

In a post on its Bing Blogs, Microsoft said 71% of AI-powered answers have received a thumbs up from users, meaning they clicked the thumbs up button in the response box that appears when you hover over one of Bing's generative responses.

But that doesn't mean Bing is without its faults.

When I asked the chat bot to tell me about Apple's (AAPL) latest earnings statement, it offered the correct information about the company's Q1 report, followed by data about its September quarter report.

The problem? The $83.4 billion in revenue Bing said Apple reported in its Sept. quarter was actually the amount reported in Q4 2021, not 2022. In other words, the chat bot gave me the right information from the wrong year.

When I told Bing it got the second part of the answer wrong and that I wanted information about Apple's Q4 2022 report, it told me that Apple didn't report those numbers yet, and wouldn't do so until Oct. 2023.

Yusuf Mehdi, vicepresidente corporativo de Búsqueda de Microsoft, habla con los medios de comunicación sobre la integración del motor de búsqueda Bing y el navegador Edge con OpenAI el martes 7 de febrero de 2023, en Redmond. (AP Foto/Stephen Brashear)
Microsoft's Yusuf Mehdi introduces the company's new Bing chat bot and Edge browser. (AP Foto/Stephen Brashear) (ASSOCIATED PRESS)

Microsoft says that it's actively working to address Bing's responses to questions that are time sensitive.

"For queries where you are looking for a more direct and factual answers such as numbers from financial reports, we’re planning to 4x increase the grounding data we send to the model," the company said in a statement.

Those aren't the only issues Bing is having. Other users have reported Bing can respond with an aggressive tone, with a staff member from The Verge writing about how the chat bot claimed it was able to spy on Microsoft employees via their webcams.

Bing also showed inaccurate information during Microsoft's initial demo last week, giving the wrong operating margin when it pulled up data from Gap's earnings statement. In the demo, Bing said Gap's adjusted operating margin was 5.9%, but in the actual report, Gap said its adjusted operating margin was 3.9%.

Microsoft has been open about the fact that Bing is still a work in progress, and will need input and feedback from users to improve its accuracy over time.

Still, during my experience with the platform, I've been both impressed with its ability to pull up information about topics like whether a TV would fit in the relatively small trunk of my Mustang.

But it's also offered strange answers to questions like conspiracy theories. For instance, when I asked it whether 9/11 was an inside job, it said the topic was a "matter of debate and speculation."

Bing responded two different ways to a question about the 9/11 conspiracy theory. (Image: Microsoft)
Bing responded two different ways to a question about the 9/11 conspiracy theory. (Image: Microsoft) (Microsoft)

Microsoft's Bing snafu isn't the first issue we've seen pop up with this new generation of generative AI.

Alphabet's (GOOG, GOOGL) Google was roundly criticized when its own generative AI, Bard, said the James Webb Space Telescope captured the first picture of a planet outside of our solar system.

While that answer is technically correct — the James Webb telescope's first images were of the specific exoplanet — it wasn't the first time astronomers ever took a photo of an exoplanet.

The AI train is certainly hot, with investors and analysts looking for various ways to jump into the space either through investments in AI-based companies or firms that supply the hardware needed to power AI systems.

But based on Microsoft and Google's early work, it's clear the field still has a way to go before AI-powered chat bots can be fully trusted.

And AI that can think for itself? That's even further away.

Sign up for Yahoo Finance's tech newsletter.
Sign up for Yahoo Finance's tech newsletter. (Yahoo Finance)

More from Dan

Got a tip? Email Daniel Howley at dhowley@yahoofinance.com. Follow him on Twitter at @DanielHowley.

For the latest earnings reports and analysis, earnings whispers and expectations, and company earnings news, click here

Read the latest financial and business news from Yahoo Finance

Download the Yahoo Finance app for Apple or Android

Follow Yahoo Finance on Twitter, Facebook, Instagram, Flipboard, LinkedIn, and YouTube

Advertisement