[ad_1]
What you need to know
- Google has officially stated that what went wrong with Gemini’s AI image generation led the company to disable the tool on February 22nd.
- The company says it has tuned Gemini to be more diverse and avoid AI generation issues, but says it has gone too far and made the bot “cautious.”
- Google said it had taken back the image generation part of Gemini for internal testing, and said the AI was prone to hallucinations and could still produce “inaccurate” results.
Google has released an official follow-up on what happened with the Gemini model’s inaccurate AI image generation for certain prompts.
Google says that when a user prompts the bot for a series of images related to a particular culture or historical timeline, it should “absolutely get a response that accurately reflects” the user’s intent. But that’s not actually the case, and Google says it was its “tuning” measures that caused the problem in the first place.
The post added: “We’ve made adjustments to avoid falling into some of the traps we’ve seen with image generation technology, including abusive and explicit images.” Unfortunately, as Google explains, Google’s adjustments did not take into account situations such as: wide The scope of diversity is not appropriate.
The second drawback is that Gemini has become “cautious” and, while treating others very sensitively, has begun to refuse prompting. Both issues are said to have triggered inaccuracies as Gemini attempted to “overcompensate.”
The company said this was “not what we intended.” Google also took a stand against intentionally creating inaccurate information related to historical content. To fix the issue, Google plans to return Gemini’s AI image generation to a testing period to resolve the issue.
In the post, Google adds that it can’t promise that Gemini won’t hallucinate or produce “embarrassing, inaccurate, or offensive results,” even after the job is done. However, the company insists that it will take appropriate action if problems arise.
In the meantime, Google is suggesting users use AI image generation for search, as the system it uses leverages “fresh, high-quality information” from the web.
Yesterday (February 22), Google officially disabled Gemini’s ability to generate AI images based on user prompts. This was in light of a series of reports and criticisms from users about bots inaccurately representing historical figures and groups of people. Google tried to encourage Gemini to create images of diverse people, but it quickly became clear that this idea didn’t apply to all prompts.
The company said yesterday that it had “missed the mark” and is currently considering next steps to fix the problem. Google has not yet said when Gemini’s AI image generation will resume.
[ad_2]
Source link
