4 ms·

The AI Dividend

13 points by mikece 3y ago

simne 3y ago
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.
PeterStuer 3y ago
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.
kdamica 3y ago
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
- ilaksh 3y ago
  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.
JohnFen 3y ago
> 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.
- seanthemon 3y ago
  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?
- toss1 3y ago
  >>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?
  - JohnFen 3y ago
    > 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.
    - toss1 3y ago
      Sorry to hear that. It was nice while it lasted. Do you see any potential alternatives to achieve the same goals?
      - 3y ago
ilaksh 3y ago
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.
- barathr 3y ago
  (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.
  - ilaksh 3y ago
    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.
    - barathr 3y ago
      Open-source, research, and small companies would be exempt.
SeenNotHeard 3y ago
Previous comments from original Politico article: https://news.ycombinator.com/item?id=36527284 https://news.ycombinator.com/item?id=36527284 https://news.ycombinator.com/item?id=36519833 https://news.ycombinator.com/item?id=36519833 (includes comments from article's co-author)

13 points by mikece 3y ago

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.

PeterStuer 3y ago

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.

kdamica 3y ago

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

ilaksh 3y ago
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.

JohnFen 3y ago

> 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.

seanthemon 3y ago
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?
toss1 3y ago
>>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?
- JohnFen 3y ago
  > 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.
  - toss1 3y ago
    Sorry to hear that. It was nice while it lasted. Do you see any potential alternatives to achieve the same goals?
    - 3y ago

ilaksh 3y ago

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.

barathr 3y ago
(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.
- ilaksh 3y ago
  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.
  - barathr 3y ago
    Open-source, research, and small companies would be exempt.

SeenNotHeard 3y ago

Previous comments from original Politico article: https://news.ycombinator.com/item?id=36527284 https://news.ycombinator.com/item?id=36527284 https://news.ycombinator.com/item?id=36519833 https://news.ycombinator.com/item?id=36519833 (includes comments from article's co-author)