At work I do not think their use is ever justifiable because the rapidly increase the amount of satisficing behavior from my colleagues. I have had many experiences where obviously bad work was submitted that was clearly llm generated and it was clear that the person submitting it just generated the output and handed it over. People turn their brains off.

The other thing I have noticed about their use is, once you start caring about the quality of your work, their value plummets. If I were to use one for my work I would need to check its output by experimenting with code, doing research, thoroughly considering both sides of an argument, etc. But if I were not going to use one I would do my own work by experimenting with code, doing research, thoroughly considering both sides of an argument, etc. So what is the advantage to using one? Either way I am still putting forth the effort to ensure my work-product is high quality. Going clackity-clack on the keyboard is not the hard part of my job, all the other stuff is.

A lot of the lyrics to tool songs focus on transcendental, spiritual, or psychedelic experiences. They also feature complicated compositions and a high degree of technical skill. As a consequence, some folks that have a “I’m on a higher plane of existence compared to you sheeple” attitude are attracted to the band. Which I think is what the meme is poking fun of.

There is some truth to it I suppose. When pusicfer (whose lead is the singer from tool) released Apocalyptical during the pandemic (a song that was clearly criticizing the stupidity of covid deniers/misinformation consumers) a lot of tool fans got upset because the song was not consistent with their conspiratorial world-view

I got my current job a few years back. I made an account thinking it would help. It was basically useless for finding a job. The folks on there that were hiring would all demand you engage with their posts (I guess as a way of increasing “influence”) but would not actually hire.

It is possible that prospective employers might look at your account during a job interview process which is why I have kept my account. But it did not help me find a job.

I just want to share my thought process here in the vain hope that someone else might see the light of reading past the headline.

This is what went through my head as I was reading:

1. 10 times more? That seems really implausible. Where did they get the figure from?
2. Ok so the subtitle is about capitalism. That seems really tendentious. That does not really inspire a lot of confidence.
3. The first figure is a bar “chart” that presents two numbers, about 3600 serial killers from the US and about 196 from England. Where did those numbers come from?
4. Scrolling down, the first link is to a report from the “Serial Killer information center” which gives the overall figure from the US as 3204, not 3600. So where did they get the 3600 number from?

Immediately, several problems jump out at the use of this database for the conclusions the substack draws.

5. First, the definition of serial killer given in the report is “The unlawful killing of two or more victims by the same offender(s), in separate events.” That is pretty broad, and would include things like a family murder (a man kills his wife, then goes to their kid’s school and kill their kid before committing suicide.), or murder for instrumental reasons (e.g. robbery). That is not usually what people think of when they think of serial killers.

6. Second the number of killers in the report shoots up dramatically in 1960. That coupled with the fact that the sources for the data are a hodge-podge of administrative records and reporting would make me very cautious about the database. This is what the webstite the report comes from says:

The database was created using information collected by Radford University students from a variety of sources including prison records, court transcripts, media sources, true crime books, and the Internet. Great care has been taken to ensure the accuracy of the information.

So, my guess is that the data are mostly from reporting. If that is true then the data are going to be biased towards “serial killers” that show up in news events.

7. So the data sources are not clear, we don’t know the distribution of legal records to news, we don’t know how records are initially identified. What is clear is that the database is not a sample of news or administrative records about violent crime. For example, the data were not collected by randomly sampling a set of judicial systems across countries then estimating counts of convictions where the offender fit the definition of “serial killer.” Rather, the data (according to a slide show on the website) “began with student serial killer timelines.” That same slide deck reports that one of the goals of the database is to provide accurate information for a forensic psychology course. That purpose suggests a focus on case studies rather than national estimates.

8. Since those students, along with the course and school, are in the US and since the data were collected in an ad-hoc manner relying in part (I suspect heavily) on news reporting, it is a safe bet that the reason the database has so many more US killers in it is because the folks who compiled it focused on collecting data from the US.

To wit: Why does the US have so many more serial killers? Because we spent more time measuring serial killers in the US.

9. The rest of the substack article drifts away from discussion of serial killers and more towards homicides in general. Some of it is less objectionable. Some of it is contradictory and obviously wrong for example:

The United States criminalizes poverty in ways that peer nations do not. Sex work is illegal across most of the country.

Sex work is actually illegal in many countries.

Among peer nations, the U.S. is an outlier on inequality by essentially the same margin it is an outlier on serial killing.

What? What does that even mean? How are they getting that figure?

Overall, I don’t think this is particularly credible. I hope now, that you too will be at least skeptical of the arguments put forth here.

There is a lot to hate about AI. A lot of dangers and valid criticism. But AI chatbots convincing people to kill themselves isn’t a problem with chatbots, it’s a problem with the user.

To me this seems like an obvious problem with the chat bots. These things are marketed as “PhD level experts” and so advanced that they are about to change the nature or work as we know it.

I don’t think the companies or their supporters can make these claims, then turn around and say “well obviously you shouldn’t take its output seriously” when a delusional person is tricked by one into doing something bad.

I wish they’d offer an llm free version with no cap on searches. Their products are too expensive and it feels like it is mostly to pay for the llms. I can’t justify paying that much for a product I am never going to use.