Hacker News new | ask | show | jobs
by dwohnitmok 1385 days ago
> I was against both the gatekeeping of GPT-3 and DALL-E by Open 'faux' AI. But now it seems that every-time an open-source alternative or version was released into the wild, it seems that the uses become even more dystopian;

So are you still against gatekeeping? Are you in favor of releasing AI advances to the wild?
1 comments
rvz 1385 days ago
I am still against OpenAI’s gatekeeping and gave AI itself a chance to be more used for good and significantly less dystopian.

Even with the release of GPT-3, there seems to be very little good in such a system despite it being generally underwhelming at generating convincing sentences. However with DALL-E 2, it has gotten much better for worse on digital images, to the point where even gatekeeping it would spur on an open source competitor superseding DALL-E 2 anyway.

But it was actually after the release of Stable Diffusion that done it for me when most here hyping just want to aid the race to the bottom and at the same time are screaming that it will help artists when (like NFTs) it won’t.

So looking at both DALL-E and Stable Diffusion, it is yet another contribution that advances the dystopian AI industry which will just be used for fake news, surveillance and catfishing. Worse part is that they haven't built any detectors for this.

link
dwohnitmok 1385 days ago
> So looking at both DALL-E and Stable Diffusion, it is yet another contribution that advances the dystopian AI industry

Given that there are more powerful models that have already been developed, should they be gatekeeped or released?

link
rvz 1385 days ago
Rather than ignoring the several conditions I mentioned, Read again on what I said:

> I am still against OpenAI’s gatekeeping and gave AI itself a chance to be more used for good and significantly less dystopian.

> Worse part is that they haven't built any detectors for this.

So it is neither. If a given AI project has no detectors or a straight indicator of knowing that it is generated by an AI, then the whole project should be effectively scrapped and cancelled, postponed, etc until it has one. It is that simple. And No. DALL-E 2's tiny watermark doesn't count.

'AI researchers' know the dystopian scam that they are creating and they know that they need detectors and analyzers for them to significantly reduce the risk of malicious use. So it doesn't matter if there are others that are more powerful as the conditions are still the same.

link