I asked Leonardo.ai for an overweight lazy man sat watching television news being brainwashed and it interprets my prompt with the following words.
A fat lazy man lying on a sofa stuffing himself with chips and beer while watching copious amounts of television, allowing himself to be brainwashed by endless television news content
The word interpretation is absolutely fine but in the image generated it is pretty obvious who the man on the television screen is supposed to be. Yes, that sure looks like the US President Donald Trump.
I never asked for any type of political bias, party or organisation, I did not even stipulate either left wing or right wing in my prompt to Leonardo. And the result is indeed partly what I asked for, we do see an overweight man sat in an armchair stuffing his face with chips while drinking beer and he is watching the news.
The image generated is the unhealthy image I wanted to produce but as for the political messaging I certainly did not request that at all. It is quite clear from this image that Leonardo has been trained to associate the words unhealthy, brainwashing and news with Donald Trump.
So therefore am I to assume that Leonardo's creators are anti-Trump lefties? Or have the users of Leonardo contributed to their own training of Leonardo, and they are the ones who have injected this political bias to a scene that requested no political bias?
I had thought the latter suggestion might explain why I am witnessing such blatant political bias but I am assured that it is only a dedicated internal team of researchers, engineers and designers employed by Leonardo's owner who train and develop the image generator, with some input from academic partners who are university based.
I know that the global design platform Canva acquired Leonardo.ai last year, and given their close links with the university sector perhaps we all shouldn't be too surprised by this blatant anti-Trump image. But this is a very worry development. I now have to do a lot more questioning regarding the politics behind these types of AI platforms. I also have to be a lot more cautious now.
What other political messaging is Leonardo injecting into its pictures? Is it subtle, do people even realise the messaging is there or are they simply imbuing it all blissfully unaware? It's all quite concerning.
Had I asked for a picture infused with anti-Trump messaging I would have asked for it!
Peace!