Google CEO Sundar Pichai addressed the company's Gemini controversy on Tuesday night, saying the tech app's controversial response to race is unacceptable. Accept and commit to the change to resolve the issue.
Google last week removed Gemini images created after the device sparked embarrassment and violence, sometimes refusing to depict white people or leading to the creation of images of Vikings, Nazis and Popes while posting images of women or people. colourful.
The debate escalated further when Gemini's arrest sparked a shocking reaction that saw Elon Musk's influence on society equated with Adolf Hitler.
The message attracted the attention and criticism, especially from conservatives who accused Google of free attacks.
< br>Most companies that provide AI tools, such as Gemini, create guardrails to reduce abuse and prevent injustice, especially when considering other information. For example, image-making tools from companies like OpenAI have been criticized for creating images of mostly white workers and depicting black people in stereotypical roles.
- Pichai, I know that some of your answers disturbed our users and were seen as unfair. "Of course this is unacceptable and we have experienced a mistake," he said.
Pichai says the company has managed to repair Gemini's railings. - Our team works 24 hours a day to solve these problems. "We've seen a lot of improvement across many applications," he said.
Google confirmed the memo and Pichai's full statement is below.
SUNDAR PICHAI'S PERSPECTIVE
I would like to address the last problem with the problem answer and image in the Gemini application (formerly Bard). I know some of his responses annoyed our users and seemed unfair; This is clearly unacceptable and we made a mistake.
Our team works day and night to solve these problems. We have seen significant improvements in many reports. Especially in the new stage of the construction industry, no information is the best, but we know that our bar is high and we will continue with determination no matter how long it takes. We will review the results and make sure the issues are resolved at scale.
Our mission is clear: to organize the world's information and make it accessible and useful to everyone. We always strive to provide users with useful, accurate and unbiased information in our products. That's why people trust them. This should be our approach for all our products, including AI products.
We will take clear actions, including changes to the model, updating the product manual, improving the release process, solid evaluation and help red group and instructions. We are reviewing all of these and will make any necessary changes.
Even if we learn from our mistakes here, we need to build on the intellectual products and technical publications we have made in the past weeks. This includes some advancements to our core model, such as our 1 million point long burst window and our well-received open source model.
We know how to create quality products that millions of people and businesses use and love, and together with our infrastructure and research scientific knowledge, it provides a great spring for the intellectual wave. Let's focus on what matters most: creating useful products that users can trust.
REED'S VIEW
The Gemini debate comes as right-wing critics often accuse tech companies of liberal bias.
But this is not actually true. This shows that Google's artificial intelligence model made mistakes while working well. The problem is not the base model itself, but the soft railings on top of the model.
This is a challenge facing not only Google, but all companies that produce artificial intelligence products.
>My understanding of the saga is that no one at Google wanted to force Gemini to describe the Pope as a woman or the Vikings as black, and no one wanted her to move between Musk and Hitler. equal justice. A failed attempt to make it less hateful than it is.
Actually, this failure shows how fast Google can act. Like all major companies, Google's growth has slowed over the years. But the challenge facing generative AI is forcing it to accelerate production.
Gemini's bug is a fixable issue, and Pichai said in a memo to employees Tuesday night that the company was working to resolve the issue.
But solving reputation problems will not be easy.
How Google CEO calls AI tool’s controversial responses ‘completely unacceptable’
About Author
You must be logged in to post a comment.