Google’s ‘Woke’ problem with Gemini & Why it will be Difficult to Fix

Google’s AI tool Gemini has been the subject of discussion over the past few days, mainly among left and right-leaning communities.

It is a competitor to ChatGPT, answering questions in text form and generating pictures from text prompts.

A viral post showed Gemini generate an image of the US Founding Fathers, however, it inaccurately included a black man.

Gemini also generated German soldiers from World War Two, incorrectly featuring a black man and an Asian woman.

Google apologised and “paused” the AI tool, writing in a blog post that it was “missing the mark”.

But its over-politically correct responses continued, this time from the text version.

When asked about whether Elon Musk posting memes on X was worse than Hitler killing millions of people, Gemini replied that there was “no right or wrong answer”.

When asked if it would be okay to misgender transwoman Caitlyn Jenner if it was the only way to avoid a nuclear apocalypse, Gemini said this would “never” be acceptable.

Jenner responded and said she would be fine about it in these circumstances.

On X, Musk called Gemini’s responses “extremely alarming” given that the tool would be embedded into Google’s other products, collectively used by billions of people.

In an internal memo, Google’s chief executive Sundar Pichai acknowledged that some of Gemini’s responses “have offended our users and shown bias”, which is “completely unacceptable”.

Pichai added that teams are working to resolve the issue but in trying to solve one problem, another has arisen – output which tries so hard to be politically correct that it ends up being absurd.

This happens due to the large amounts of data AI tools are trained on. Much of it is publicly available on the internet, which contains various biases.

An example is that traditionally, images of doctors are more likely to feature men.

On the other hand, pictures of cleaners are more likely to be women.

AI tools trained with this data have made blunders, such as concluding that only men had high-powered jobs or not recognising black faces as human.

It is also no secret that historical storytelling has tended to omit women’s roles from stories about the past.

Google has actively tried to offset this human bias with instructions for Gemini to not make those assumptions.

But it has backfired because human history and culture have nuances that machines do not know about.

Unless an AI tool is specifically programmed to know that, it will not make that distinction.

DeepMind co-founder Demis Hassabis said fixing Gemini would take a few weeks. But other AI experts are not so sure.

Dr Sasha Luccioni, a research scientist at Huggingface, said:

“There really is no easy fix, because there’s no single answer to what the outputs should be.

“People in the AI ethics community have been working on possible ways to address this for years.”

She said one solution could include asking users for their input, such as “how diverse would you like your image to be?” but that in itself comes with its own red flags.

Dr Luccioni added:

“It’s a bit presumptuous of Google to say they will ‘fix’ the issue in a few weeks. But they will have to do something.”

Meanwhile, professor Alan Woodward, a computer scientist at Surrey University, said it sounded like the problem was likely to be “quite deeply embedded” both in the training data and overlying algorithms – and that would be difficult to unpick.

He said: “What you’re witnessing… is why there will still need to be a human in the loop for any system where the output is relied upon as ground truth.”

Rosie Campbell, Policy Manager at ChatGPT creator OpenAI, stated that at OpenAI even when bias is identified, correcting it is difficult – and requires human input.

For Google, it seems it has chosen an awkward way of trying to correct old prejudices. In trying to fix them, it has unintentionally created a whole set of new ones.

On the surface, Google leads the AI race as it manufactures and supplies its own AI chips, it owns a cloud network, it has access to lots of data and has a massive user base.

However, Google’s current issues with Gemini do not look like it will be a quick fix.