The LLM users/middlemen pay. The user probably pays less than they would have to pay the author. The LMM provides information retrieval / discovery.
354 karma · joined August 26, 2015
- Paraphrasing n=7 words (and quite a few more) within a sentence can easily be fair use.
- As n gets big, the bloom filter has to also.
If/when attribution is solved for LLMs (and not fake attribution like from Bing or Perplexity) then creators can be compensated when their works are used in AI outputs. If compensation is high enough this can greatly incentivize creativity, perhaps to the point of realizing "free culture" visions from the late 90s.
However, the suit goes far beyond claiming that such copying violates their copyright: "Unauthorized copying of Times Works without payment to train LLMs is a substitutive use that is not justified by any transformative purpose."
This is a strong claim that just downloading articles into training data is what violates the copyright. That GTP outputs verbatim copies is a red herring. Hopefully the judge(s) will notice and direct focus on the interesting, high-stakes, and murky legal issues raised when we ask: What about a model can (or can't) be "transformative"?
> Nothing at all. It distracts me from my family, friends, work, and myself.
> I have no interesting insights or original thoughts about it, and if I did,
> absolutely no one would care. So why read it?
Do you vote or participate in politics in any way or do events in the news directly affect your life? One reason it's rational to read the news: If you don't read the news, then you're at risk of lacking important contextual knowledge about society and social norms. It's pragmatic to keep track of these conversations to avoid embarrassment and to not behave in ways that might offend others. I find it useful to be aware of contentious issues on which I might have a side or a stake.
So I'll still get my colonoscopy, thanks
http://www.pnas.org/lookup/doi/10.1073/pnas.1915006117
You might think that it's possible to use machine learning to predict whether people will be successful using established socio-demographic, psychological, and educational metrics. It turns out that it's very hard and simple regression models outperform the fanciest machine learning ideas for this problem.
The way this study was done is also interesting and paves the way for new kinds of collaborative scientific projects that take on big questions. It draws on communities like Kaggle, but applies it to scientific questions not just pure prediction problems.
The more interesting finding is that there are more women in high status (first or last) author positions in the open science community (when the number of authors is smaller) and there is an increasing trend in this community of women occupying such positions. This would be predicted by theories that STEM is an individualistic enterprise less likely to attract people with communal values. Women are more likely to have communal values and this is often provided as a cultural explanation for gender gaps in STEM. But the open science community is a part of STEM that sees communal practices (specifically, the publication of data and code along with information about findings) as key to improving science. This is in contrast to the reproducibility community which has legitimate criticisms of established scientific practices but does not emphasize pro-social practices in the same way as the open science community.
In sum I think the paper is useful by
1. Showing that while "open science" and "reproducibility" have some superficial similarities they are distinct communities with interesting differences.
2. Showing ways that the open science community seems more collaborative and communal and thus it seems attractive to women (and likely may be this way because women are helping to drive it).
The paper also has some shortcomings. Names are not gender and gender isn't binary. There's a lot of discussion about diversity and team science which honestly doesn't seem to have much to do with the empirical contributions of the paper.
1. Sharing economy remains strong, but investors (at least in uber/lyft possibly others) are disappointed as it isn't possible to maintain monopoly. Multi-sided markets expand into more sectors as platforms and apps become understood as infrastructure.
2. Companies providing engaging private small-group social media experiences (along the lines of signal, whatsapp groups) are really important to most people. This is also somewhat disappointing to investors as this kind of activity is difficult to monetize and local networks are easy to transport from app to app.
3. Major social network platforms become increasingly controlled spaces as the "open internet" becomes a sea of disinformation. Most people instead turn to regulated enclaves with barriers to entry and strict moderation.
4. AR / VR aren't adopted in widespread ways by consumers. VR is popular among gamers and AR has some use in specialized work environments such as military, factory, and warehouses.
5. The world succeeds at preventing some of the worst ecological consequences of climate change, and transitions to a sustainable green economy by way of economic contraction, increased efficiency, and limited carbon capture technologies (i.e. conversion to biomass) but fail to keep warming below a 2⁰C level. Mass migrations out of equatorial regions are a huge geopolitical problem and the source of much violence. Eco-fascism is a significant political ideology that limits the ability of developed nations to admit refugees.
6. (this is my stretch and my own vision) Online communities learn from the example of Wikipedia how to self-organize to create valuable goods. Many productive online communities will be relatively closed and occupy a similar niche to corporations or cooperatives and exist for the economic opportunities provided to their members.