This article contains spoilers for the 1973 film Soylent Green.
Im talking about DALL-E 2 and OpenAIs announcement that its incredible text-to-art generator would be entering a closed-beta.
Most exciting of all: an additional one million people will gain access to DALL-E 2.
Lets do a cartwheel.
Up front:There would be no cartwheels in the Neural offices at TNW that day.
DALL-E 2 is, in this humble editors opinion, a scam.
But thats nothing new in the technology world.
Facebook is a scam.
Google is a scam.
Microsoft is a scam.
Theyre all profiting off of something that has nothing to do with what theyre telling us theyre selling.
It’s free, every week, in your inbox.
If you lay bricks, youre a bricklayer.
And if you write programs, youre a programmer.
But if you sell ads in 2022, youre probably a search giant or a social media company.
It doesnt have to make sense so long as it makes profits.
For the most part, big tech makes its profits by convincing you to be the product.
The data we generate powers the products that these companies sell.
Background:DALL-E 2 is a fancy extension of OpenAIs transformer model, currently iterated as GPT-3.
It takes text prompts and turns them into pictures.
Its mind-bogglingly good and will absolutely 100% revolutionize the world of content creation.
Let me be super clear here: I love it.
I think its amazing.
However,sellingaccess to it is dangerous.
Its not just unethical.
Spoiler alert:Soylent Green is an old movie but its a classic.
Per the movie, in a distant future humankind is on the brink of starvation.
To survive, the government gives them nutrition bars and thats all most people ever eat.
One day, they come out with a new flavor called Soylent Green.
Everybody loves it and, luckily, its produced in abundance.
Unfortunately, it turns out that Soylent Green is made of humans.
To solve the starvation problem, the government started processing dead people and feeding them to us.
Who gave OpenAI the right to sell ownership to images our data helped create?
But thats not what DALL-E 2 does.
Dont get hung up on the art aspect of it.
It can smash those eight pictures together and make them one picture.
Eventually, OpenAI gets to the point where its model is smashing millions or billions of images together.
Its still the same trick.
Wrapping it up:Data is data and output is output.
Doesnt matter if its art, text, or advertising profiles.
When OpenAI takes your data, it never had your permission.
This sets a dark and dangerous precedence for the use of scraped data going forward.
