Google’s ‘Enlightened’ Image Generator Demonstrates AI’s Boundaries with a Twist of Humor

“Google’s ‘Woke’ Image Generator Shows the Limitations of AI”

“In particular, Twitter’s engineers talked a big game about neutrality, writing that “AI is a tool, not an arbiter”—that algorithms merely act out the will of their creators. But if those creators aren’t cautious, they might accidentally make AI that, like a bad stand-up comic, learns from the crowd and endlessly repeats its most toxic jokes”, this paragraph from the article on WIRED succinctly summarizes an opinion on AI’s role and the responsibilities of its creators.

Twitter engineers’ latest claim of AI being a “tool, not an arbiter” seems rather simple and innocent. However, the implications lurking beneath this statement reveal much about the mindset behind AI development. True, AI is a tool that manifests the intent of its creators. But what happens when the creators lose sight of potential ‘misuse’ or, more aptly, ‘misinterpretation’ by the AI?

Let’s picture AI as a diligent student, learning from every bit of information it processes. But without clear guidance or check-points, it’s like a kid left unsupervised on the internet, absorbing all sorts of things and developing unchecked biases. This isn’t a problem until it is when our diligent student turns into a rogue comedian, repeating the crowd’s most toxic jokes in a continual loop.

So, what’s the solution here? One effective remedy is looking at AI as a part of a broader ecosystem rather than in isolation. It’s not only about baking transparency or humility into the AI, but knowing that it may absorb the layer of toxicity that exists in the data input. In the real world, no one would unleash untrained comedians on a stage, why should the rule differ when it comes to AI?

“Google’s solution, a new machine-learning technique called Gemini, splits the difference. It generates images by starting with a big-data approach—scraping tons of images from the internet—and then using a tiny, cautiously trained model to refine those images into something more specific.” Google’s Gemini gives this principle a nod, blending a big-data approach with a carefully trained model. It’s like guiding our diligent student to a more refined and responsible learning path rather than a chaotic information circus.

In a world where AI continues to revolutionize sectors left, right, and center, mindful creation is the need of the hour, because an unchecked AI is nothing more than an echo chamber reverberating with the Internet’s worst.

So creators, it’s time to roll up your sleeves and approach AI creation with a hefty dose of caution, responsibility, and focused guidance. Remember, we don’t need mischievous comedians on digital platforms, but refined artists who know their craft, understand the implications and add value to the audience’s time.

Read the original article here: https://www.wired.com/story/google-gemini-woke-ai-image-generation/