I wish I could read this, and I wish even more it had been open access or otherwise published in an open way — archive.today can't get behind the firewall, and I can't access it at my university (an R1 institution) either because they don't subscribe to it. It would have been better to post it on a blog or on an archive server, which is a bit ironic maybe given the apparent point of the commentary.
I think some of the results (e.g., in math) that have been achieved using LLMs point to the value of LLMs but the basic idea of this commentary stub is worth taking more seriously. The pressure to publish, not just by hiring and promotion committees, but also by the din of communicative noise due to the number of academics involved, leads to a lot of unsurprising and predictable output.
There's a position that LLMs will supercede humans in academic work and render them unnecessary, but also a position that maybe some of that work, the work replicated by LLMs, is superfluous anyway, something we should be reducing for everyone's benefit.
It depends on the program, and even more so, the student and the mentor. It can also vary over time, with more direction early on in a graduate program, and less direction later. Some mentors are very directive, and basically treat students as labor executing tasks they don't have time or want to do. Other times, the student is coming up with all the ideas and the mentor is facilitating it with resources or even nothing but uncertain advice or permissions now and then.
This can lead to a lot of problems as I think in some fields, by some academics, the default assumption is the former, when it's really the latter. This leads to a kind of overattribution of contribution by senior faculty, or conversely, an underappreciation of less senior individuals. The tendency for senior faculty be listed last on papers, and therefore, for the first and last authors to accumulate credit, is a good example of how twisted this logic has become.
It's one tiny example of enormous problems with credit in academics (but also maybe far afield from your question).
Or maybe they are? I'm not an expert in this and reading through some of the government literature there's no mention of this.
Then at least you would know that a given price marker is a good empirical index of how other prices are changing also, at least for a given dimension/component.
The thing about arts and AI is that it's one place where the stochastic parrot criticism has never really stopped being accurate to me.
I think backlash against the stochastic parrot idea has always been based on a bit of a strawman — what's being parroted can be pretty abstract and broad in scope, down to reasoning strategies and personality. But with art in particular, I always feel like it's broadly imitating something from elsewhere.
In those AB tests in the linked essay for example, the options are all pretty prototypical genre music samples being compared against one another. Even the fact that they have pretty accurate prototypical labels says something about how prototypical they are, and you can always find something pretty stereotypical of a genre to compare against to play "gotcha" or make it more difficult. It's just not interesting or relevant to me to test whether you can tell the difference between cliched music samples from a human or cliched music samples from AI.
Real art is often like that — you have writers and musicians who follow genre conventions for a variety of reasons — but the most interesting art happens in the margins in ways that are unexpected.
The essay is worth reading for its argument that music is fundamentally a participatory activity — this is something else on my mind lately about live performance art in general, and the implications of that for understanding certain societal trends. But sometimes I think the discussion of AI and art is really missing the boat in other ways as well, ways that apply to AI in general.
I agree that there's a lack of awareness of what happens in other countries with ID, but I think it is also a different situation in the US.
States in the US in a lot of ways are more comparable to countries in the EU. It's not exactly like that but in many ways it is. So it would be like requiring an EU ID on top of a national ID.
I also don't think privacy per se is the real issue of concern, it's concern about consolidation of federalized power. Privacy is one criterion by which you judge the extent to which power has been consolidated or can be consolidated.
The question isn't "can this be federalized safely in theory", it's "is it necessary to federalize this" or "what is the worse possible outcome of this if abused?"
As we are seeing recently, whatever can be abused in terms of consolidated power will be eventually, given enough time.
I guess discussions of whether or not you can have cryptographic verification with anonymity kind of miss the point at some level. It's good to be mindful of in case we go down the dystopian surveillance route, but it ignores the bigger picture issues about freedom of speech, government control over access (cryptographic guarantees of credential verfication don't guarantee issuance of the id appropriately, nor do they guarantee that the card will be issued with that cryptographic system implemented in good faith), and so forth.
The rationalization aspect of their model can't be overstated enough in my opinion. It never starts with something clearly unethical; it starts with something more complicated. Something that is uncomfortable and morally suboptimal, but has some justification being appealed to — it benefits the group, it's otherwise unfair for some members to have to bear some small temporary cost for the benefits of others, or something of that sort. The level of the corrupt behavior becomes more and more extreme though, such that justification becomes more and more questionable, until you're left with something more seriously problematic. In the meantime, the people who questioned the slippery slope might have left, and you're left with people who aren't in a position of power for whatever reason (they're junior, or in small numbers) to question what has become clearly unethical cultural norms.
I think the article is overlooking an important category of corruption where social norms treat certain acts as theoretically immoral but in practice impose little to no social sanction for such acts.
In places like India, for example, taking bribes is just standard practice. It carries so little social sanction that it’s like jaywalking here in the US. People acknowledge it’s technically illegal, but it carries so little social sanction that people don’t consciously need to rationalize it. The same thing with cheating in schools, which is normalized in India and has become almost as normalized in the U.S.
The “power” focus for corruption is not useful. Corruption by people without power is more harmful to society than high level corruption. People skimming off the top is undesirable, but survivable. But ordinary people taking bribes and cheating grinds society to a halt and makes it impossible to develop societal wealth.
This is well evidenced in the development of asian countries in the 20th century. South Korea has plenty of corruption at a high level. But it’s clean at the lower level, and as a result it’s been able to become a rich society. By contrast, India has crippling corruption at lowest levels of society that imposes a huge drag on routine transactions and daily life.
That's a good point, although I think it depends on how you define "power". I think in a lot of ethical discourse, power is defined more broadly in terms of who has more leverage in a situation. It doesn't have to correspond to a political position or position of physical authority necessarily.
In the US at least I think the power distinction is dangerously becoming lost. Traditionally, there was a trade that was made for power: you have more power, you give up privacy to maintain transparency, and you are held to a higher standard. Frequently it seems arguments are made that people in power should enjoy the same standards as others with less power, which was not part of the trade.
I agree with you that there is an importance of cultural norms at every level, but I do think power matters, as it is key to understanding conflicts of interest and permeates every level of society.
I enjoyed this paper, and there's innumerable things that could be said about ingroup-outgroup dynamics and corruption.
In my personal experiences with corruption with organizations, ingroup membership often becomes increasingly narrowly defined, and defined in such a way as to benefit a certain group of individuals at the expense of others. The underlying rationale is a narcissistic entitlement or rationalization for why one person or small group of people is deserving of disproportiate benefits or flexibility at the expense of others. It starts with some kind of distorted egocentric schema about others in a more distal way, and then becomes increasingly strict and more proximal. Narcissistic egocentrism is at the core; it only manifests more weakly at first, and then becomes stronger. The ingroup boundaries never stop shrinking, because there always has to be some justification for why that particular group — which was never really defined by the initial ingroup boundaries, the ingroup was only a proxy for themselves — is more deserving than others.
I think some of the results (e.g., in math) that have been achieved using LLMs point to the value of LLMs but the basic idea of this commentary stub is worth taking more seriously. The pressure to publish, not just by hiring and promotion committees, but also by the din of communicative noise due to the number of academics involved, leads to a lot of unsurprising and predictable output.
There's a position that LLMs will supercede humans in academic work and render them unnecessary, but also a position that maybe some of that work, the work replicated by LLMs, is superfluous anyway, something we should be reducing for everyone's benefit.
reply