Bias, internal, external & my eye-opening talk with Google Gemini

Alchemy For Life

Google Gemini

Gemini is Google’s version of ChatGPT. It’s a large language model (LLM) that thinks and responds naturally. You can have conversations, ask for advice, ask it to play a part, and even ask it to generate images.

That’s where my journey began.

I asked it to generate a few images. This led to a discussion about Bias.

Bias

You see, Gemini’s handlers have you covered on bias, because they decided to remove all of it. The problem? They have their own bias.

Let’s talk about just how dangerous that is, and how it affects you, and why “removing all bias” is a lofty, impossible, and potentially harmful goal.

Check out AI, Language Models and Puppies for a background on AI (this time ChatGPT).

Here is the entire raw conversation. You can scroll past the images to when we discuss bias. For those familiar with AI image generation, Gemini’s generation is rather 1st generation, and for images ready for prime time, I use Midjourney.

Note – the original reason the word “blonde” came up is that, on a whim, I wanted to see if it could generate an image of the protagonist in Upside Down.