We’re fast becoming accustomed to AI running virtual errands and completing tasks that previously needed a team of experts.
But what are the ethical implications lurking behind this powerful product?
From comical deepfakes to a Peeping Tom turned rogue – we ponder how AI will affect children’s development and if it will undermine artists’ creative spark.
Plus, is AI robbing us blind, sneaking past copyright laws to meld themes of one artist with a design from another?
Read on to find out how to navigate the ethics of AI image generation.
AI and Ethics: What Can Go Wrong?
Credit: Steve Johnson
From copyright issues to deepfakes and spreading misinformation, AI has opened up a legal and ethical can of worms.
Let’s take a look.
Does AI Respect Copyright Laws?
One of the first ethical concerns regarding AI-generated images was their potential infringement of copyright laws.
In 2023, Midjourney, DivianArt, and Stability Art were taken to court for infringing on artists’ copyright. Their AI generators had created art in the same style as the artists who were suing them.
AI is trained on millions, if not billions, of unlicensed photos and information. Many of these images are protected by copyright laws.
AI generations are made by a technique called diffusion, which creates entirely new images from the millions it has in its database.
However, even though the images are new, AI has been known to mimic an artist’s unique signature style and has spat out art with an artwork’s copyright label attached.
Oops, not a good look, back to AI school with you.
There’s no exact amount of alteration that removes a copyright from an image. By law, if a new creation is different from the original but is considered to be a derivative of an original work, it can infringe on copyright laws.
Now, AI does not regenerate the image as an exact replica, but it can use sections of the image and can use images as a style guide; therefore, it can infringe on copyright laws.
If we think about it from a different angle, humans are also trained on billions, if not trillions, of images. Is the art we create influenced by copyrighted images? Yes, for sure it is.
Do we even remember the images that triggered and inspired us? The pictures that sit in the depths of our subconscious mind? I would say, most likely not.
However, there is a difference. We usually create something completely different, an art piece that doesn’t replicate exactly the originals that inspired us. The art we create will reflect our unique taste and style.
You, as the generator, can ensure you don’t infringe on copyright laws by editing your AI generations; otherwise, you can use an AI generation website that offers indemnity.
AI generations are not considered worthy of copyright, but then the interesting aspect of this situation is that we don’t know who owns the AI generation; is it the coder, the AI company, or the person who prompts the AI? Or is it AI itself?
Does AI Support Bias and Stereotypes?
AI hasn’t been trained to be politically correct – or has it?
While AI is impartial and it won’t discriminate, it does reflect the bias of the coder and the images used to train it. AI will represent a bias towards certain stereotypes based on the images it’s been trained on.
For instance, if AI were trained only on photos of purple flowers, it would likely struggle to generate flowers of other colors.
If the data used to train an AI algorithm is not diverse, the AI’s generations will reflect this.
When prompted to make images for causes, articles, or projects, AI will often pump out a generic result.
AI is known to amplify social biases; middle-aged men will be in positions of power, women will be minding children, and the old veteran will be a beggar.
However, this is not because AI is intentionally producing this. It’s simply relaying the data it was trained on.
The user also affects AI’s bias through their interactions and by giving feedback.
This means AI will align its image creations with the expectations of the user, much like Google algorithms.
If someone who loves deep-sea diving searches for ‘currents’, the results will show ocean currents. In contrast, if a financial advisor Googles the same term, their results will be related to finances, not the sea.
The same is true of AI images: if a young girl asks for a picture of a horse, it will be a cute white horse riding a cloud, and if a man prompts the same, it will be a stallion racing through a field.
AI is giving the user what it thinks they want. It has been programmed, just like Google, to hone its results to the user’s preferences.
In this case, it’s not AI that is biased or supporting stereotypes; it is the users’ prompts or the images used to train it.
The Misuse of DeepFakes
Now we get into more interesting territory: deepfakes. Who would have thought we would be able to use a small handheld device to create doctored footage of a public figure?
Footage that’s so believable, the world can be brought to a standstill.
One of the most controversial deepfakes was of Zelensky asking his Ukrainian troops to surrender.
I am not a fan of war, so I would happily create deepfakes to tell ALL troops to surrender, but the Russians beat me to it. Misinformation has always been a weapon of war.
Deepfakes raise major concerns as it’s now impossible to know what’s real and what isn’t. Misinformation is in free fall; basically, we have to question everything.
Novices using high-end AI apps can create a believable deepfake wth nothing more than an idea and a downloaded photo.
Previously, we could detect that an image had been manipulated, but now we are left wondering.
There are telltale signs, such as incorrect proportions, a lack of shadows, or odd blurs. However, these aren’t always obvious at first glance.
Deepfakes don’t only affect our current news and political campaigns. There’s also the rise of deepfake revenge porn, which is used to blackmail or humiliate people.
It’s known that shaming can have damaging effects on people’s psyches. Once a deepfake video is released, even if it’s retracted later, the damage will be done.
The court system has to hire AI experts to discern if footage is real or fake. Even using AI deepfake detectors and an expert AI eye, it is impossible to be sure whether the footage is authentic.
Authenticating fabricated images and videos is a new challenge, and court systems need to be prepared.
And guess what, as AI advances, deepfakes will become increasingly realistic.
Misinformation and Propaganda
AI means that anyone, anywhere, can manipulate images and footage. Simply upload a photo and prompt, and the believability of the alteration will depend on the AI app being used.
If truth is told, photo manipulation has been at play since photos were first created. Misinformation is nothing new; only the tool has changed.
Think about when Photoshop was first created; it was an exclusive tool. Most people didn’t know it existed, never mind how to use it.
Imagine how much propaganda and misinformation could have been pumped out by those with access to it. No one would have even questioned it.
Some even say that the moon landing photos were Photoshopped!
In the past, superior high-end fakery was limited to a few, such as the wealthy, technologically savvy, the military, and the government.
Now, it’s possible for any devious teen with a smartphone to create believable photos. Yikes! Imagine the social havoc they could wreak.
In some ways, deepfakes make the world a difficult place to navigate. Seriously, what images can we trust?
On the other hand, maybe it’s good that people are starting to question whether information, images, or footage is really real.
The media has always fabricated information to suit its bias, and advertisers warp reality to sell a product.
Time for AI to Cough Up
AI has built its platform on hard sweat, capable coding, and ingenious programmers. To train AI, they used unimaginable amounts of images from the internet.
The significance of AI using others’ work to build its impenetrable empire is compounded by the fact that it’s now stealing the livelihood of the same people.
Businesses turn to AI generators to pop out a logo, create a pretty picture, or redesign a poster. They skip the hassle of finding a human designer.
So there is one question I will ask: if AI image generators have reaped the benefits of artists, photographers, and designers’ talent, should they pay?
AI and Personal Privacy
Credit: Cottonbro Studio
Have you made an AI-generated business photo yet? Altered your age, or swapped sexes with your partner… virtually, of course.
To create a business photo, you upload photos of your face from different angles, and AI has the information it needs to recreate your face from every angle.
AI then uses algorithms to calculate your bone structure, eye size, and lip size ratio. Using these calculations, it can recreate an image of you with any expression.
You can take it to another level and use AI video creation apps to create a moving version of you.
AI’s capability to recreate you is incredible, fun, and novel, but it has more sinister repercussions if it’s used for the wrong reasons.
In a malevolent person’s hands, this technology can be used to ruin reputations and infringe on personal rights.
Did you consent to your image being used to sell Coca-Cola, or to represent the housewives’ association of New York?
No, I hear you say, so why are you there on the internet flaunting a new apron and teaching people how to cook this season’s delicious cookies?
Many AI apps will state that once your images are uploaded, they will not be shared with a third party, but others don’t.
So if privacy is important to you and you don’t want your likeness shared far and wide, make sure to check the app’s terms and conditions.
Is this AI I am Talking To?
Not all online platforms using AI images admit to it. This can be anything from an AI-enhanced image to creating an AI generation from scratch.
There have been reports of people using deepfake influencers to rake in the dosh. Using AI video generation apps, a model influencer can be fabricated with the most appealing face and body proportions, mesmerising followers.
When there is no transparency, it creates a lack of trust. Once viewers realise that they’re watching an AI person, not a real person, they lose interest.
When they realise the influencer’s appearance is enhanced to the extreme, they feel cheated. A little like a man who marries a woman, thinking her eyes were green, and then she removes her contact lenses.
Should images and footage created using AI state that they are AI?
AI and Creative Spark
How will the presence of AI affect children? Could AI dampen mankind’s creative spark?
Imagine a young child drawing a stick figure, believing they are ready to forge a career as the next Leonardo, but then they spot what AI can do.
The child sees that a machine can rapidly create any image, simply with a prompt. Will comparing their own capabilities against AI’s lead them to discard their pencil and their creative conquest?
Could the use of AI at home or in schools lead to creative defeatism, accepting and bowing down to that which is mightier than you?
There are already studies suggesting a risk of overall cognitive atrophy and loss of brain plasticity due to the use of AI.
I believe here that it is important that new generations are taught to view AI as a tool, not a superior intelligence. This will determine whether children compare their talents to those of AI or learn to utilize it to support their endeavors.
We don’t know yet how AI has impacted young people’s motivation and creativity.
They are the young, imaginative, and adaptive minds who could utilize this tool in more innovative ways than its inventor.
The determining factor will be how they are guided to approach AI.
Last but Not Least, AI’s Energy Consumption
It might seem like AI is created from thin air with no fuss, but it has taken endless hours of machine training and mountains of images.
Today, AI image generators can pump out a plethora of glamorous and authentic-looking images and footage, yay!
One little detail we tend to overlook when enjoying the novelty and benefits of AI is the data it uses.
The images AI references to create generations need to be stored somewhere – in data centres.
Running data centres takes considerable amounts of energy. Plus, creating images also consumes significant amounts of energy.
To appease adverse ramifications, AI platforms and developers can look at ways to minimise energy consumption.
Final Thoughts
As we can see, AI is definitely causing an ethical stir as it beats a path through an unknown terrain.
AI is not walking a slow strut; no, AI is running so fast it’s hard to keep up with it to check it has abided by the rule book.
Like a snowball unexpectedly tumbling down a mountain, it’s impossible for anyone to stop AI, even AI itself.
Undoubtedly, this would ruffle feathers and break some rules. It’s unavoidable when a new innovative development appears on the scene.
AI is a development that has been immediately incorporated into daily life by millions, if not billions, of people, intentionally or unintentionally.
Its impact on our lives and the development of children is still to be determined.
Handling AI’s infringement of copyright laws will be dealt with by the courts. It’s a complex issue that involves weighing up the long-term pros and cons for all those involved.
The world has always been a minefield of misinformation. AI is just another tool in the toolbox for those who wish to spread it.
If privacy is a concern for you, I would check the privacy terms of the AI apps you’re using. You could unwittingly be giving your rights away.
Aside from ethics and morals, AI is like a pen. It’s just a tool, and a tool in the hands of the good-hearted will be used for good.
But when it is used in the hands of the devious…? You’ve guessed it.
Credit : Source Post
