They came prepared.

DALL-E2

OpenAI has released its latest product, dubbed DALL-E2, an impressive piece of machine learning tech that can turn simple text descriptions into photorealistic images, say, of bandana-wearing pandas riding motorcycles or an astronaut riding a giant turtle in space.

But the company has wisely anticipated the potential misuse of its advanced AI as well, given there’s plenty of precedent for AI projects going awry.

OpenAI ensured that DALL-E2 has a limited ability “to generate violent, hate, or adult images,” according to its website. “By removing the most explicit content from the training data, we minimized DALL-E 2’s exposure to these concepts.”

No Hate

In other words, you won’t get any explicit content out of DALL-E2 — unless, of course, somebody were to find a workaround.

“We won’t generate images if our filters identify text prompts and image uploads that may violate our policies,” the website reads. “We also have automated and human monitoring systems to guard against misuse.”

For now, it’s tough to say how effective those measures are. The company is slowly rolling out the tool to a “select group of users” via a waitlist.

Surfing Cobras

It’s not a surprising concern, given how other AI models have turned out to be racist or sexist in the past.

Besides, when the internet is given too much power, like the ability to have an AI generate new photorealistic images out of a simple set of words, things are almost bound to go south.

So only time will tell whether DALL-E2, or related tech, will become a new source of propaganda or the tool of choice for pornographers.

For now, though, expect to more of cobras “surfing on a big wave,” “teddy bears working on new AI research on the moon in the 80s,” or a knitted “bowl of soup that looks like a monster.”

More on the tool: Amazing New AI Can Spit Out Realistic Images From Simple Descriptions


Share This Article