I remember rolling my eyes at Technopoly a lot more than I did at Amusing Ourselves, but actually...I read Technopoly maybe 15 years ago, when the Internet had a lot more promise and less downside. Maybe if I read it now it would seem a lot more correct.
Amusing Ourselves, on the other hand, was brilliant when I first read it, but now we seem to be living in a world that is both Orwell and Huxley. The free market provides endless pap, AND governments/political figures have learned to use new media to provide endless misinformation designed to keep us fearful. So maybe if I read it today it would not seem as relevant.
This is not exactly correct. When the book was written, a lot of commentators had written off television as useless garbage -- totally bad for society. Postman (correctly) complicates that by pointing out that television is great for emotional storytelling, and he is in favor of fictional television shows that model social values for people. Television turns everything into emotional content, which is why TV news evolved to be so sensationalistic -- TV is not good at news. Printed media is good at news. So, each media has uses that play to its strengths and weaknesses. When a medium is used in weak ways, that is bad for society.
(I don't know what the redeeming argument for TikTok would be...)
TikTok is good at personal content— short form videos that capture a slice of life for other real people. TikTok is a salon.
Little bits of entertainment — music, poetry, jokes/clowning, gossip, show and tell on all kinds of topics, musings about life and interpersonal relationships, practitioners demonstrating their craft.
It presents an interesting cross-section of news that's hard to get anywhere else. It's not great or even good at the normal news format but you can read about a riot and that the mayor called in the national guard in the paper, but the video of someone on their porch getting screamed at and having warning shots fired at them for not going indoors has an impact.
That makes sense. It sounds like the good parts of TikTok are getting little slices of life from interesting people, and of course, jokes and entertaining stuff, and the misaligned parts are accounts that purport to be "informational" about news or politics. TikTok is probably just as bad a source for that as TV is.
Printed media were emotional and sensationalistic too when there was not TV yet. Journals like that still exist and they used to be common and large. Sensationalism changed form, that is it.
This feels like a question that would really benefit from talking to some musicians, instead of basing your framing on biopics and data. Some musicians are in it to be rich and famous, and become computer programmers when riches and fame leave the picture, but most of them are musicians because they like writing and playing music more than anything else. I'd bet that Of Monsters and Men is perfectly happy to make an okay living off of their remaining fanbase, because they still get to tour and make new albums. They might be even happier about it if they could get a second hit, but there are so many musicians who have never had a breakthrough hit, who make a happy living opening for other bands, meager royalties, Patreons, etc. Basically, this post assumes a causal link between chart success and life satisfaction that probably doesn't exist.
The idea that these word problems (and other LLM stumpers) are "easily solvable by humans" needs some empirical data behind it. Computer people like puzzles, and this kind of thing seems straightforward to them. I think the percentage of the general population who would get these puzzles right with the same time constraints LLMs are subjected to is much lower than the authors would expect, and that the LLMs are right in line with human-level reasoning in this case.
(Of course, I don't have a citation either, but I'm not the one writing the paper.)
Yeah, as someone with an education background I suspect GPT-4 is relatively close to the general public's performance on this problem. Many people would miss AIW, and almost all would miss AIW+. I'm about as good at this kind of thing as anyone and I'd need a minute with pencil and paper to handle AIW+; it's on par with the most difficult problems found on tests like the GRE.
I wonder if these models, trained on data from across the internet, are in some ethereal way capturing the cognitive approaches of the average person (and not picking the best approaches). If the average person does not think in these sorts of symbolic-manipulative terms, and therefore does not write in those terms, and you train a model on that writing...?
I wonder the same thing. If any academic reading this wants a paper idea:
1. Examine papers and other claims that an LLM gets something wrong that a human would have gotten wrong. How many of those claims have any citations about how many humans actually get it wrong? How many of those citations use the general population instead of the population of people who would be uniquely well-suited to answering the question correctly (i.e. people who signed up for the GRE are more likely to get GRE questions right than the general population).
2. For claims that are totally missing citations on human performance, run some tests with humans from the general population (or as close as you can get), and see how the LLMs compare.
You're not alone -- a lot of this resonates with me, but I (also in my 40s) might be a little further along in my journey and have a few more resources to draw on. But I feel this.
The most important thing: therapy. Don't spend too much time finding a therapist -- take the first one with an opening. If you don't get along with them, ask for a referral or use a method discussed in other comments. But the important thing is to start. Many therapists will do video calls now. Medication can also be effective.
Some more good stuff about your life:
* You're not too old for marriage or even kids. Guys are lucky in that respect -- we have an easier time finding partners and having kids when we're older.
* You took an important step: you posted this question. I am terrified to do any such thing about my own problems, so I respect you a lot for it.
* You had a network before, and professional networks are more resilient than you think. If a co-worker from 10 years ago emailed me out of the blue, I'd be happy to talk to them. I think you'd be happy to talk to your old colleagues, so why the assumption that they wouldn't respond to you? That's an example of catastrophizing thinking. A therapist can help you with that better than HN randos...but still, asking HN randos is a good step to take!
* It's not a given that the best of your life is behind you. Every part of your life can be the prime of your life if you value the things that part of your life has to offer.
There are some good points here, but the sweeping conclusion (presented with utter certainty) does not follow from them.
It seems like the author's beef is with journals, rather than peer-review. If we did away with "peer review" today, journals would still have to operate the same way -- they'd still have many more submissions than they have room for, so a team of people (ideally, peers) would need to, uh, review those submissions according to some criteria. We can discuss whether the criteria should be adjusted, but I don't see how journals survive without gatekeepers.
So, fine, he wants to do away with journals. Without a description of an alternative system, it sounds like the best researchers would just...upload their stuff to Arxiv and hope that someone reads it? Again, I'm not saying there is no alternative, but because he spends all his time arguing against "peer review", he spends no time discussing alternatives to journals that would solve more problems than they create.
He addresses the question "can we fix peer review instead of replacing it" by discussing ways that fixes have failed in the real world. So what makes him think that a replacement would be easier? The "burn it all down and rebuild it according to my preferences" approach also doesn't have a great track record!
And the certainty with which he states his conclusions gives me a sense that this is not someone who's super open to feedback.
He did make an example by publishing his paper on his own blog. So maybe every researcher can put a blog and upload papers there and expect them to appear in someone else's browser who followed the researcher because either he shared something interesting or someone else shared his paper?
I've always wondered why academia can't work like social media. Not Twitter or Facebook, which is algorithm-controlled, blogs would work where everyone maintains their own sovereignty.
Ah, but there are also systemic issues that prevent Congress from doing the things you'd like them to do. For instance, Congress is vulnerable to money. Journals have money they can use to lobby. Are there other players with money and incentive to lobby for the "burn it all down" side? Which mountain of systemic issues is easier to climb? Sounds like an empirical question.
Solving the problem by either approach is going to be really really complicated and full of compromises you won't like, and anyone who talks like their solution is super simple and obvious should be read with skepticism.
If these people create real music, "fake" seems like the wrong adjective. It's very common for musicians to be involved in a bunch of different bands, or even for the same group of people to release music under different band names, for when they want to experiment with stuff that isn't their normal band's "brand". What's being described sounds like SEO optimization to me. A recent Reply All episode interviewed someone who is very successful at Spotify SEO: https://gimletmedia.com/shows/reply-all/j4he7lv
But what if Spotify pays somebody a salary to create this music, then artificially pushes it up their search-results so people will play it which means they will play less of the music created by independent artists, which means Spotify has to pay fewer royalties to the independent artists.
It would be a bit like if Google-search for some product showed products created by Google on top of the list, but not really even revealing they were products by Google.
I have no information about if this kind of thing is happening but the article suggests there is evidence to suspect it is. Weird artists nobody knows about being on top of the list. How did they get to be top of the list?
