The AI Dividend | Hacker News Reader

The AI Dividend

schneier.com

13 pointsmikece16 comments

> You are owed profits for your data that powers today’s AI, and we have a way to make that happen. We call it the AI Dividend.

I don't want an AI dividend. I want to be able to put my stuff on the web without it being used to train AI.

>>I want to be able to put my stuff on the web without it being used to train AI.

That ship has sailed quite a while ago.

In fact, like low-background-steel from before the first nuclear tests, AI now has the problem of whatever it scans on the web now has a contamination problem.

If we want our text to be invisible to AI, seems we need to keep it private, either only local, or behind a connected but gated online community. Or publish it only on paper.

Are there enough people who don't want to train an AI with their online musings that they'll put up with a members-only media? Those restrictions are ostensibly why Musk is restricting Twitter to rate limits and must-sign-in. Will it work?

> That ship has sailed quite a while ago.

I know. I'm just still coming to terms with the fact that the open web is no longer a place that I feel OK publishing on. I've had to remove all of my websites from being publicly available and I'm pretty angry about it.

Sorry to hear that. It was nice while it lasted.

Do you see any potential alternatives to achieve the same goals?

Certainly not with the same reach as the public web. But my websites are for the benefit of the public more than for my benefit. They're a way of giving back to the community. That's the loss I mourn more than anything.

So I am still publishing -- only now, it's an invite-only sort of deal. I have heard from several others that are doing the same thing.

The best-case scenario that I see right now is if others who also feel pushed off the open web get together and do some sort of private federation, resulting in a smaller, private web.

Good plan. Scaling (up) the 'Web Of Trust' may be the most difficult thing of all... I wish you well in building it out!

I want to put my stuff on the internet and I honestly don't care if it's used to train AI, it's so deeply a part of so many pieces of text, does it really have such an impact?

This idea keeps coming up but the reality is that the per-person payouts would add up to such a small amount that it wouldn’t justify the administrative costs of implementing the program, from both a payout and an enforcement perspective.

I don’t think this is a good idea in the first place, but if we wanted to implement like this, it would be much simpler just to raise existing taxes for ‘big tech’ companies

We have conclusive proof for many years that current approaches to taxing large corporations do not work. We need to make structural changes to the way that works. Something along the lines of this proposal could be more effective.

I think something like this is a great idea.

The particular numbers he used don't add up though. $0.001 per word generated by AI is more than what OpenAI charges for GPT-4 which is $0.06 per 1000 tokens or approximately $0.00008 per word.

(I'm co-author of this piece.) You're right, and that's the exact idea: it will make LLM-generated text much more expensive but still a couple of orders of magnitude cheaper than a professional writer.

That makes many applications infeasible because the user conversion cost will be too high. It means you can't provide even one free trial run unless you are VC funded.

Open-source, research, and small companies would be exempt.

I'm on the fence on this for now but I think there are legitematly different objections that could be made to the appropriation and exploitation of data besides the mere financial.

E.g. that some people are prepared to sell certain personal services does not mean those services can be taken from anyone without concent as long as some financial compensation is postfactum regulated into the non-concentual 'transaction'.

Not trying to draw a direct equivalence in any way, just drawing a deliberatly exaggerated analogy to illustrate a point.

This is not working solution.

Existed only two moderately successful examples of societies of massive share - USSR (Russia) and China.

USSR finished at war (and disappeared, when understood impossible to win that war), and China constantly approaches global war.

All other who tried, disappeared (many with agony), or ended these experiments at very limited scale.

Previous comments from original Politico article:

https://news.ycombinator.com/item?id=36527284

https://news.ycombinator.com/item?id=36519833 (includes comments from article's co-author)

Reply on news.ycombinator.com