Google has apologized for inaccuracies in historical image generation depictions produced by its Gemini AI tool. The company then disabled image generation. The tool, aimed at creating a wide range of results, has faced criticism for depicting specific white figures and groups, such as Nazi-era German soldiers, as people of color. Google acknowledges the need to improve these depictions and address the long-standing racial bias problems in AI. The controversy has mainly been promoted by right-wing figures, accusing Google of avoiding the depiction of white people. Google’s attempt to boost diversity in generative AI may have led to these inaccuracies, as image generators often amplify stereotypes.
While I’m hitting product news, Microsoft announced that Zoom will integrate Microsoft 365 apps through the new Document Collaboration Partner Program. This program allows eligible partners to integrate Microsoft 365 apps into their platforms, enabling users to share, edit, and co-author within third-party collaboration solutions. Zoom is the first partner to adopt this program.
Microsoft has added three new capabilities to its Intune Suite: Enterprise Application Management, Advanced Analytics, and Cloud PKI. The Enterprise Application Management feature allows for easy app deployment and updates, while Advanced Analytics provides AI-powered device issue examination. Cloud PKI enables the setup and management of a PKI in minutes.
Let’s play the Gemini thought exercise. You’re creating an AI image generation tool and know your data set has biases. How do you fix that? What’s your answer? Well, you put some boundaries on it… but what are those? When do you need to apply the boundaries, and when are you not? When is the query a search and not a creation?
I don’t bring these up, claiming to have all the answers – I bring them up because the questions themselves are critical. AI frameworks are such a key component to success, as they help work through some of these impossible questions.