Hacker News Reader

Comment by ralusek | Hacker News Reader

Can someone help me understand the deep fake controversy? That's like making photoshop illegal.

It was always possible to modify images to produce inappropriate or insensitive content, but plugging a turbocharged state of the art image generator with virtually no guardrails into every Twitter reply and then failing to address the issue long after it was obviously being used for CSAM or deepfakes of real people against their will.. well that's worse

Notice the Wikipedia link says the problem was "put her in a bikini." The claims about "Grok just lets you undress people" were massively exaggerated because people hate Elon (perhaps for good reason) and not worse than other models.

Having clothes removed to the point of wearing a bikini is "being undressed" and I feel you're choosing not to understand the impact of being publicly sexualised in a bikini can have.

Grok is directly tied to Twitter in a way that other models don't have, so the use of Grok to do this stuff is inherently more public and traumatising for the targets.

You're right that people hate Elon and that they have good reason to do so, but you might be falling for the trap of underestimating the legitimate and unique concerns about Grok because it's easy to assign them just to "Elon hate."

Think of it this way:

If you install Photoshop locally (ignoring that it's now cloud based), and made deep fakes locally - that's probably fine. If something goes wrong as a result, only you are liable. It's a general purpose tool - the tool author isn't liable.

If you instead set up a server, and let users create deep fakes on that server, then as the operator of the server you have some level of culpability.

AI safety is a tricky topic. At some level, having it is a pain. It's a general purpose tool! Why limit me? The answer is that I don't control the tool, and am not the one running the tool - the provider is. If I don't want AI safety, then I need to run the model on my own machines (or on rented servers).

If an LLM provider is going to sell the service on the strengths of the benefits you get from it, they should take responsibility for the downsides.

Doing a good "deepfake" Photoshop requires skill.

With an AI model it requires the ability to speak or write, not much more.

Most people are against CSAM.

A lot of AI users are profoundly stupid and intently malicious. That changes perception of the tool... IMO it's because generative AI data is inherently toxic and contains elements that incite primal rage, but that's just my gut theory.

Reply on news.ycombinator.com