Skip links

Should We Embrace or Feel Threatened by Dall-E 2?

Everyday, highly innovative tech inventors: solo individuals, and friend groups of tech enthusiasts, come up with more and more impressive technologies to solve one human problem or another. But soon or later, the world is jostled awake from its enchantment to the realization or fear that a new technology actually poses as many threats as problems it is made to solve. One of the main threats that critics of these technologies outcry is their capability to replace human expertise in certain fields. Such is the case of the AI system, Dall-E 2, invented by OpenAI to “…create original, realistic images and art from a text description.” Basically, if you need a never before seen art piece to decorate your living room wall, all you have to do is type in a description of the painting to Dall-e 2 and it generates your decor piece! What’s more, if you feel unsatisfied by the initial result, you could re-enter the same search term and it will produce a different result. This is not a drill, people! It is real and we have proof because we have tried it here at Simpliexpand and the results are quite impressive. Although the image outputs were not 100% accurate, a number of them were quite close to the figure. 

 

See them below for yourself!

1. an elephant with a beard driving a taxi stuck in the middle of traffic

2. A small monkey scaring a huge gorilla

3. wide angle shot from below of a giraffe eating noodles in space

4. a car with the just wedded sign parked in front of a home

5. A portrait painting of a beachside condo almost hit by a tsunami

See them below for yourself!

1. an elephant with a beard driving a taxi stuck in the middle of traffic

2. A small monkey scaring a huge gorilla

3. wide angle shot from below of a giraffe eating noodles in space

4. a car with the just wedded sign parked in front of a home

5. A portrait painting of a beachside condo almost hit by a tsunami

Welcome back from that short, awe inspiring trip. And just to calm your curiosities, all the images that you just watched never existed until the moment their descriptions were typed into Dall-E’s system. Every single one of them. 

Now, considering the fact that this is coming from an improvement on a system that was only released for the first time in January, 2021, it is quite impressive, some of the results the system produces. Even more impressive is the fact that the version that we got our pictures from isn’t even open to the public yet! What we have now is the beta version and major modifications will be made, of course, to further improve on what we have now. Imagine the level of accuracy this thing could actually attain when the final version arrives! But this is also a cause of concern for photographers, artistes, designers, and some other techies and creatives, and for us here at Simpliexpand. Other important concerns have also been cited and they make us question if really, Dall-e is something that is worth embracing.

One of the concerns raised about Dall-e 2 is bias in its results. Experts have reported that certain image outputs contained gender, racial, and sexual biases. For instance, when you enter a command like “person crossing the road while flying”, the system would almost always produce more images of a male than a female character portraying the person mentioned in the command. In response to this, OpenAI suggests that this might be as a result of the system’s training algorithm being filtered. In order to improve output from Dall-E 2, the training data was filtered to removed sexual imagery and violence, and since much of this training data already sexualized women, search commands that like our example will produce more images of male characters that it would female ones.

Another concern that has been raised is the issue of people taking advantage of dall-e to create deepfakes. Deepfakes are media that have been modified to imitate the likeness of real people, for fraudulent and other purposes, and it is done with the help of AI algorithms such as Dall-E 2. As a result of accuracy in some of its results, critics and other experts in the tech industry worry that the existence of Dall-E might engender a wave of deepfakes as people might want to use the platform for selfish, malicious, and other negative reasons.

 

Lastly, and by many means, most crucially, there is the concern of the system completely replacing millions of people at their jobs. Photographers, artists, graphics designers, stand the chance of losing their jobs to dall-e because of its popularity and accuracy. This, coupled with the fact the system creates images in seconds that would take the human artiste/photographer/designer hours, even days to create, makes it look like a threat to industries.

 

Now, back to answering that question of whether or not to embrace this world-changing technology. Fact: Dall-E 2 is, without doubt, one of the most impressive AI technologies the world has experienced yet. Fact also: all concerns that have been raised are based on solid reasoning. Now, while OpenAI has power over some of these concerns such as the issue of deepfakes and bias toward women, some other concerns, like the chance of Dall-E 2 putting artists and photographers out of work are very much matters of concern. So, are we going to embrace the wonder of Dall-E 2 with its probable consequences, or just stitch it in time and save nine?

 

The enthusiasts and lovers of all things tech over here at Simpliexpand are excited for a technology like Dall-E 2, and are even more excited to see how far it comes, how improved it gets in the future. Someone even imagines that future image outputs will be as accurate as looking into your own mind! As for the creatives, their excitement doesn’t measure up to a quarter of the techies’ because who wants to be taken out from their job by an algorithm? They have spent years and money nurturing their talents and skills, and so being excited about a technology that reduces all of that dedication and effort to almost nothing is quite impossible. But then again, even us humans misinterpret each other’s words every time, let alone a computer algorithm. So maybe the threat would be 100% real if the future version of Dalle-E 2 had deeper learning abilities than humans.

We would like to have your response too! Share your comments with us here on what you think of Dall-E 2. Should we embrace it or feel threatened by it? And why should we or should we not embrace it?  

Leave a comment