This week Barry Phillips looks at the new version of Gemini Pro Flash that takes image generation to a new level.

Transcript:

Hello Humans!

My name is Barry Phillips and welcome to the podcast that aims to summarise in five minutes or less an important development in AI relevant to the world of HR

In the old days prior to the advent of GenAI if HR wanted to represent their team or their employees visually there were two options: One call in a professional photographer and ask a few colleagues to pose for the camera. This was expensive, time consuming and guess what, surprisingly few employees felt comfortable in front of the camera. The other option was to go to iStock photo and buy a few images of professional looking people. This never worked well. The images were good. In fact they were too good. The characters featured were immaculately dressed. Fabulously good looking and very well manicured. But they looked nothing like the people in the workplace. It was like Stepford Wives meets the shop window models from Topman (golly I’m showing my age now).

When ChatGPT4 eventually developed its image generation capability a third option began to materialise. It was fun to play with but not really of much practical value. Sure it could generate images that followed the user’s prompt relatively well but it was never quite good enough to use. Without the ability to edit your only option was to try to generate a new image and blindly hope that the next version came with the improvements you had asked for. It was time consuming and well we all know in HR we don’t have a lot of time and certainly not enough to be playing around with image generators for very long.

Then by mid 2024 the edit capability got better. You could highlight a part of the image you wanted to change and it changed it for you – sometimes according to your instructions but not always. It was also still struggling with digits. Fingers and hands came out notoriously badly leaving the user feeling terribly guilty that they had created four people who were destined to have all sorts of problems holding pens, sending a text message or for that matter getting out of the very room they were sitting in.

ChatGPT5 seems to have overcome the digits problem but it’s Gemini Pro Flash (AKA Nanna Bannana) released last week that may prove to be the game changer here. Early tests of this model at Legal Island have shown impressive results.

Yesterday we asked it to generate an image showing four HR people sat around a table at work discussing a chart on working time issues. It produced just this. The information on the chart was both relevant, clear to read and with no spelling mistakes. The edit facility was truly impressive. I swapped coffee cups for mobile phones, young professionals for older ones and replaced one chart featuring data analytics for another showing a pie chart.

So what does this mean for us in HR? Well, for a start we should be cancelling our subscription to Getty Images if we haven’t done so already. But we also have an opportunity to present to the outside world much better representation of what an HR team or a work force actually looks like. It means too that there’s the option of better visualization of HR and its work than before. Employees can see them through images as much as they can through plain text sent by email or via Teams.

With each evolution of GenAI comes opportunity for HR. The question is is HR ready to embrace it and optimise its value?

Until next week, bye for now.