Never underestimate the military PowerPoint[1] industry!

[1] https://media.wired.com/photos/5933e578714b881cb296c6ef/mast...

well considering an element of their access is the lifting of safety guardrails, I'd assume the scope includes, to some degree, the processing or generation of nsfw/questionable content

See IDF’s Gospel AI - the goal isn’t always accuracy, it’s speed of assigning new bombing targets per hour

What you're saying just isn't true.

I can get an AI to generate an image of a bear wearing a sombrero. There are no images of this in its training data, but there are bears, and there are images of sombreros, and other things wearing sombreros. It can combine the distributions in a plausible way.

If I am trying to train a small model to fit into the optical sensor of a warhead to target bears wearing sombreros, this synthetic training set would be very useful.

Same thing with artillery in bushes. Or artillery in different lighting conditions. This stuff is useful to saturate the input space with synthetic examples.

I'm not arguing this is the purpose here but data augmentation has been done for ages. It just kind of sucks a lot of the time.

You take your images and crop, shift, etc them so that your model doesn't learn "all x are in the middle of the image". For text you might auto replace days of the week with others, there's a lot of work there.

Broadly the intent is to keep the key information and generate realistic but irrelevant noise so that you train a model that correctly ignores the noise.

You don't want to train your model identifying some class of ship to base it on how choppy the water is, just because that was the simple signal that correlated well. There was a case of radiology results that detected cancer well but actually was detecting rulers in the image because in images with tumors there was often a ruler so the tumor could be sized. (I think it was cancer, broad point applies if it was something else).

Generative models can combine different concepts from the training data. For example, the training data might contain a single image of a new missile launcher at a military parade. The model can then generate an image of that missile launcher hiding in a bush, because it has internalized the general concept of things hiding in bushes, so it can apply it to new objects it has never seen hiding in bushes.

If you're building a system to detect something, usually you need enough variations. You add noise to the images, etc.

With this, you could create a dataset that will by definition have that. You should still corroborate the data, but it's a step ahead without having to take 1000 photos and adding enough noise and variations to get to 30k.

The model they're training to perform detection/identification out in the field would presumably need to be much smaller and run locally without needing to rely on network connectivity. It makes sense, so long as the openai model produces a training/validation set that's comparable to one that their development team would otherwise need to curate by hand.