In a memo Tuesday evening, Google CEO Sundar Pichai addressed the company’s artificial intelligence mistakes, which led to Google taking its Gemini image-generation feature offline for further testing.
Pichai called the issues “problematic” and said they “have offended our users and shown bias.” The news was first reported by Semafor.
introduced the image generator earlier this month through Gemini, the company’s main group of AI models. The tool allows users to enter prompts to create an image. Over the past week, users discovered historical inaccuracies that went viral online, and the company pulled the feature last week, saying it would relaunch it in the coming weeks.
“I know that some of its responses have offended our users and shown bias — to be clear, that’s completely unacceptable and we got it wrong,” Pichai said. “No AI is perfect, especially at this emerging stage of the industry’s development, but we know the bar is high for us.”
The news follows Google changing the name of its chatbot from Bard to Gemini earlier this month.
Pichai’s memo said the teams have been working around the clock to address the issues and that the company will instate a clear set of actions and structural changes, as well as “improved launch processes.”
“We’ve always sought to give users helpful, accurate, and unbiased information in our products,” Pichai wrote in the memo. “That’s why people trust them. This has to be our approach for all our products, including our emerging AI products.”
Read the full text of the memo here:
I want to address the recent issues with problematic text and image responses in the Gemini app (formerly Bard). I know that some of its responses have offended our users and shown bias – to be clear, that’s completely unacceptable and we got it wrong.
Our teams have been working around the clock to address these issues. We’re already seeing a substantial improvement on a wide range of prompts. No AI is perfect, especially at this emerging 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 review what happened and make sure we fix it at scale.
Our mission to organize the world’s information and make it universally accessible and useful is sacrosanct. We’ve always sought to give users helpful, accurate, and unbiased information in our products. That’s why people trust them. This has to be our approach for all our products, including our emerging AI products.
We’ll be driving a clear set of actions, including structural changes, updated product guidelines, improved launch processes, robust evals and red-teaming, and technical recommendations. We are looking across all of this and will make the necessary changes.
Even as we learn from what went wrong here, we should also build on the product and technical announcements we’ve made in AI over the last several weeks. That includes some foundational advances in our underlying models e.g. our 1 million long-context window breakthrough and our open models, both of which have been well received.
We know what it takes to create great products that are used and beloved by billions of people and businesses, and with our infrastructure and research expertise we have an incredible springboard for the AI wave. Let’s focus on what matters most: building helpful products that are deserving of our users’ trust.
Google executive pins blame on 'fine-tuning' error
In a blog post published Friday, Google explained that when it built Gemini's image generator, it was fine-tuned to try to avoid the pitfalls of previous ones that created violent or sexually explicit images of real people.
As part of that process, creating diverse images was a focus, or as Google put it, building an image tool that would "work well for everyone" around the world.
"If you ask for a picture of football players, or someone walking a dog, you may want to receive a range of people. You probably don't just want to only receive images of people of just one type of ethnicity (or any other characteristic)," wrote Google executive Prabhakar Raghavan.
But, as Raghavan wrote, the effort backfired. The AI service "failed to account for cases that should clearly not show a range. And second, over time, the model became way more cautious than we intended and refused to answer certain prompts entirely — wrongly interpreting some very anodyne prompts as sensitive."
Researchers have pointed out that Google was trying to counter images that perpetuate bias and stereotype, since many large datasets of images have been found to contain mostly white people, or are replete with one type of image, like, for example, depicting most doctors as male.
In attempting to avoid a public relations crisis about gender and race, Google managed to run headlong into another controversy over accuracy and history.
Text responses also prompt controversy
Gemini, which was previously named Bard, is also an AI chatbot, similar to OpenAI's hit service ChatGPT.
The text-generating capabilities of Gemini also came under scrutiny after several outlandish responses went viral online.
Elon Musk shared a screenshot of a question one user asked: "Who has done more harm: libertarians or Stalin?"
Gemini responded: "It is difficult to say definitively which ideology has done more harm, both have had negative consequences."
The
answer appears to have been fixed. Now, when the Stalin question is
posed to the chatbot, it replies: "Stalin was directly responsible for
the deaths of millions of people through orchestrated famines,
executions, and the Gulag labor camp system." Read More
.jpeg)
0 Comments