Published in News

Google boss Sundar Pichai says sorry for Gemini mess

by on29 February 2024

US media claims it is black and white

Google boss Sundar Pichai apologised for the company's Gemini fiasco, calling the AI app's dodgy responses about race unacceptable and promising to sort it out.

Google pulled the plug on its Gemini image creation tool last week after it upset the US Right by creating photos of women or people of colour when asked to make images of Vikings, Nazis, and the Pope.

As expected, red-faced white guys at Fox News accused the AI of being "woke" and "anti-white." Fox always has a problem with Google, which it claims has a lefty bias.

But the AI scandal worsened when Gemini made dodgy text responses, such as saying Elon Musk's impact on society was the same as Adolf Hitler's.

Most companies offering AI tools like Gemini put up barriers to stop abuse and bias, especially after other blunders. For example, image generation tools from companies like OpenAI have been slammed when they made mostly images of white people in posh jobs and showed Black people in subservient roles.

In fact, in this case, it has been suggested that Google was doing its best to reverse a bias against black people in the AI, and it backfired (particularly when you put something into a historical context). Some of the issues are that Google has been pressured to move faster in the generative AI race as Vole and OpenAI steal the lead.

Pichai ruled out anything deliberate or having a woke agenda. He knew that some of its responses were unacceptable and Google cocked it up.

Pichai said the company has already made progress in fixing Gemini's barriers.

 "Our teams have been working non-stop to sort out these issues. We're already seeing a big improvement on many requests," he said.

"No AI is perfect, especially at this new stage of the industry's development, but we know the bar is high for us, and we will keep at it for however long it takes. And we'll look into what happened and ensure we fix it for good," he said.

Pichai said that even as Google learnt from what went wrong, it should also build on the product and technical announcements we've made in AI over the last few weeks.

That includes significant advances in underlying models, such as Google's 1 million long-context window breakthrough and our open models, which have been well received.

The Gemini mistakes are a fixable technical problem, and Pichai's note to staff Tuesday night shows that the company is working on it.

Last modified on 29 February 2024
Rate this item
(0 votes)