The hatred towards AI is staggering. I created possibly the greatest personalized music algorithm of all time and I'm getting nothing but hate.

[Mattson]

Listen.  I am not a programmer and I realize using ChatGPT to write code is a bad idea but the results have gone far and above everything I could have hoped for.  Here is the story of what I did.  And in full disclosure I have not used AI to help me write this post.

 

I started using AI maybe three weeks ago after years of reservations.  I've used it for a number of things to fruitful ends but today's story relates to getting ChatGPT to give personalized music recommendations.

 

If you chat with SHODAN there are many ways to get it to recommend music that usually entail it asking you a few probing questions and telling it your favorite albums.  What I have done has gone so far and above this I genuinely feel people think I'm just telling it albums I like(which technically I am just the scale is larger than they can comprehend).

 

You see I am a last.fm user(Mattson402) have been for 20 years.  I have been meticulously logging everything I ever listened to in that time and have a data set with over a quarter million data points of listening data that is personalized and unique to me.  Its not just my favorite albums but everything—conveniently time-stamped and dated.

 

The issue I ran into is getting that data off last.fm and into an LLM.  Its not as easy as just giving it your username and a hyperlink.  It kept hallucinating hard when I would provide the link.  It kept telling me my account was private when it wasn't.

 

What did work are screen shots.  But after doing the math I worked out it would take 4000 screenshots on PC and 12000 on mobile.  That's just untenable.  So I asked ChatGPT what I should do and it told me I could write a script and export the data as a CSV and send it over.

 

Now I don't know really how to program.  I had a C++ class back in 11th grade and people would pay me to edit the HTML on their MySpace pages but outside of that I have no information on coding and while I was good at editing HTML its been over 20 years and I've forgotten it all.

 

So here is the journey ChatGPT sent me down to create this script to pull this data.

 

First it had me download python and install something called a library.  Then it had me get an API key from lastfm which is free if only for personal use.  Which is appreciated because I'm going to be making alot of calls.

 

After all that it gave me a script and had me run it.  It failed so I copied the failure into the LLM and it told me the file path for python is wrong.  Finally it runs but it keeps crashing at 130 pages(for reference I needed 4000 pages and a page is equal to one PC screenshot).  ChatGPT suggested I was getting rate limited.

 

So I had it take another pass and this time it added a 0.2 second delay.  This time we crashed on page 1480 of 4000.  I sent the error back to ChatGPT and it told me since the file didn't save I probably ran out of ram.

 

So we ran it back and this time ChatGPT added a feature that causes it to save every 100 pages.  Bingo!  It worked!  Took a half hour but it worked.  I had ChatGPT change it one last time so it ammends the csv rather than regenerates it so future pulls take less time and hammer last fm less though I haven't tested this.( Lastfm)

 

So I have the data, what can I do with it?  Well I can save 60 bucks a year and cancel last fm pro because this does everything last fm pro does and so so much more.  It has exceeded my expectations.  I'm not going to cancel last fm pro though as matter of principle.

 

Last fm pro generates weekly listening reports and fancy graphs characterizing your listening behavior.  If you're a music and stats nerd you love it.  The script can do all that and you can tell if its referencing the csv because it will cite it in its response.

 

What can I do beyond last fm pro?  I can look up the first time I ever listened to Radiohead.  Granted you can do this in about 30 seconds if you know how to navigate a profile but SHODAN is only limited by your WPM.

 

But one request you can't get in 30 seconds is "how many scrobbles do I have on Tuesdays between 3 and 4am". Granted a very niche ask but one that would be possibly hours if you just used site navigation.

 

But perhaps best of all is its capacity to make laser focused music recommendations that are so far up your alley they come out your mouth like that one South Park episode.

 

They're better than Spotify's too because they will make pulls from things not even on Spotify.  Also Spotify weights recommendations towards artists it has favorable deals with.(this is conjecture but I say it with like 90% certainty).  And then you have the problem of AI generated trash slipping into your recs... ChatGPT ignores this.

 

I am personally over the moon with how great these recommendations are.  Sure ChatGPT is doing the digging but the albums it pops out at the end are real artists.

 

I am so amazed by this so I tried sharing it with the music nerd communities because I figured they'd appreciate it but instead I amassed negative down votes and received ao much criticism you'd think I'm occupying the Rhineland

 

It is absolutely staggering the amount of hate out there for AI.  Is this even an unethical use case?  How do you get through to people on the issue?  I know you can lead a horse to water and you can't make it drink but if they'd just take a sip their minds will be blown.

 

Anyway... Thats it, thank you for coming to my TedTalk. Chat with my listening history uploaded. There is a link to the Chat with all my listening data if you want to put it through its paces.

 

Some of you may not have last fm accounts and thats a bummer but if you're in Europe you don't even need to code to try this.  Just make a GDPR privacy request with Spotify or whatever streamer you use and they will email you a CSV with your entire spotify history.  Just up that and you're set.

 

And for the record I didn't create anything here... I just found out how to get my data off a website, ChatGPT is doing all the heavy lifting... I just know what it likes to hear in this one particular instance.

 

PS: oh and another thing it can do that is a pretty big deal that I forgot to mention: it can sanitize the dataset

[CasualExtremist]

I assume you are familiar with the hallucinations LLMs are prone to have? You have to be very careful in how you place your trust throughout the interactions.

 

37 minutes ago, Mattson said:

received so much criticism you'd think I'm occupying the Rhineland

You never elaborated, so if you find it worth your time, it might also be useful to cover the main counter points brought up by others. If you find it hard to summarize the negativity then you may not understand it enough to fully claim you disagree with it.

[Lurick]

14 minutes ago, CasualExtremist said:

I assume you are familiar with the hallucinations LLMs are prone to have? You have to be very careful in how you place your trust throughout the interactions.

I mean, people hallucinate too and a large portion are willing to ignore reality to live in a specific bubble of their own reality.

 

To your point though, yes, you 1000% need to at minimum cross check the output of the AI at least a few times against reputable/known sources. Trusting any AI output at face value is a terrible thing, even more so today.

[Timme]

Triple this. And props for using the thing as intended, and not as a life-coach speechwriting guru. ChatGPT is great at analysis and transformation, but terrible at recognizing and processing specific and niche contexts in an overall, generic-looking prompt. Also, it'll never be critical and tell you "you are super wrong and not even close", because "pleasant service experience". So caution is advised.

 

ChatGPT is my go-to anime advisor. 
It can work with an XML export of the page, given the context. If the platform of interest doesn't have a native "export list" feature, batch downloaders can handle the task. d
 

 

[emosun]

1 hour ago, Mattson said:

It is absolutely staggering the amount of hate out there for AI.  Is this even an unethical use case?  How do you get through to people on the issue?  I know you can lead a horse to water and you can't make it drink but if they'd just take a sip their minds will be blown.

The hatred is for AI in general.

AI in general isn't widely embraced by people due to the whole "this is a replacement for people" mentality it markets itself as.

Go figure , you tell people they are worthless and what a surprise... they don't like that.

[Mattson]

1 hour ago, CasualExtremist said:

I assume you are familiar with the hallucinations LLMs are prone to have? You have to be very careful in how you place your trust throughout the interactions.

 

You never elaborated, so if you find it worth your time, it might also be useful to cover the main counter points brought up by others. If you find it hard to summarize the negativity then you may not understand it enough to fully claim you disagree with it.

You're right I never elaborated but I was on their side for the last five years and have been only patronizing AI for the last three weeks or so.  Has there been any significant changes to their main counterpoints in the last three weeks?

[Biohazard777]

3 hours ago, Mattson said:

You see I am a last.fm user(Mattson402) have been for 20 years. 

...

After all that it gave me a script and had me run it.  It failed so I copied the failure into the LLM and it told me the file path for python is wrong.  Finally it runs but it keeps crashing at 130 pages(for reference I needed 4000 pages and a page is equal to one PC screenshot).  ChatGPT suggested I was getting rate limited.

...

Anyway... Thats it, thank you for coming to my TedTalk. Chat with my listening history uploaded. There is a link to the Chat with all my listening data if you want

It’s easy to catch errors when they’re stopping you from making progress... but once things seem to work, how can you be sure they’re actually working correctly?
Take this example from your chat:


Lets just focus on the last week or so, 16th till 23rd:

Lets compare that to what last.fm says:
https://www.last.fm/user/Mattson402/library?from=2025-06-16&to=2025-06-23

And I can already see it doesn't match what chatGPT spat out...

But lets be through, hover over the bar graph to get the values on last.fm
6, 34, 102, 193, 174, 135, 48, 163, 
and now lets make it a line chart, and put them side by side,
left: chatGPT, right: last.fm

Similar, but not quite right.
Did you go over the context window limit, or maybe the data wasn't scraped properly, maybe it just drew the chart wrong, or maybe the python code was ass...

Who knows, if you are unable to verify what it is doing... which brings me to: what else might it have messed up? 
 

3 hours ago, Mattson said:

It is absolutely staggering the amount of hate out there for AI.  Is this even an unethical use case?  How do you get through to people on the issue?  I know you can lead a horse to water and you can't make it drink but if they'd just take a sip their minds will be blown.

Nah man, I don’t hate AI, it can actually be a pretty useful tool.
But like any tool, you need to know what you’re doing to use it properly.

[wasab]

8 hours ago, Mattson said:

I am so amazed by this so I tried sharing it with the music nerd communities because I figured they'd appreciate it but instead I amassed negative down votes and received ao much criticism you'd think I'm occupying the Rhineland

I used to be downvoted so much on reddit, you'd think I was burning down Moscow and the Chinese Summer Palace all at the same time. 

[dogwitch]

its more of 1 side is trying to do hate base tribalism to the point of harm or harassment.

the other side wants a civil debate on the matter.

 

[SpaceGhostC2C]

7 hours ago, Mattson said:

> I am not a programmer

> using ChatGPT to write code is a bad idea

Pick one. Because if you're not a programmer (or you are, but nor a very good one), it's definitely a good idea. Microsoft did not buy GitHub for no reason  

I work in an environment where there is a lot of coding mostly by people without formal training in programming, and the productivity gains are insane. Maybe not necessarily for writing everything from scratch, at least I don't think I've witnessed an example that allows me to tell either way. But as an assistant to build on your first attempt, or debug? Way more efficient than endlessly browsing stack overflow 

[Mattson]

6 hours ago, Biohazard777 said:

It’s easy to catch errors when they’re stopping you from making progress... but once things seem to work, how can you be sure they’re actually working correctly?
Take this example from your chat:


Lets just focus on the last week or so, 16th till 23rd:

Lets compare that to what last.fm says:
https://www.last.fm/user/Mattson402/library?from=2025-06-16&to=2025-06-23

And I can already see it doesn't match what chatGPT spat out...

But lets be through, hover over the bar graph to get the values on last.fm
6, 34, 102, 193, 174, 135, 48, 163, 
and now lets make it a line chart, and put them side by side,
left: chatGPT, right: last.fm

Similar, but not quite right.
Did you go over the context window limit, or maybe the data wasn't scraped properly, maybe it just drew the chart wrong, or maybe the python code was ass...

Who knows, if you are unable to verify what it is doing... which brings me to: what else might it have messed up? 
 

Nah man, I don’t hate AI, it can actually be a pretty useful tool.
But like any tool, you need to know what you’re doing to use it properly.

Fair point though to be fair its close enough and its not like I'm trying to do something objective here like using it to calculate pie.  In the end of the day its about music which is an entirely subjective field.

 

Try not to lose sight of the forest through the trees.  Sure if I was trying to solve cancer this would be incredibly problematic but in the grand scheme of things this use case is of the least importance to humanity's well being so if things are slightly off the world isn't going to bthe worse for it.

 

Out of curiosity how far from truth is the model?  Is it more than 10% or are we picking nits here?

 

EDIT:  since posting this I was reminded that last fm stats exists and you can export your history to CSV without writing any code at all.  So in the future I'm going to be feeding that data into chatgpt.

 

Also the kernel reset on chatgpt and it lost the csv so I attached it but you can pull any profile on last fm stats

 

And finally if you don't have last fm but do have lots of playlists on Spotify you can use a webapp called Exportify that rips your Spotify playlists and that data is valuabe if you want chatgpt to make relevant recs, especially if you crafted the playlist yourself.

 

lastfm_scrobbles.rar

Edited June 24 by Mattson
to make this moot

[Sauron]

11 hours ago, Mattson said:

It is absolutely staggering the amount of hate out there for AI. Is this even an unethical use case?

I don't think there's anything unethical about getting music recommendations from a set of your own listening data per se. In terms of using LLMs broadly you should bear in mind that right now the power cost is being hidden from you because these companies operate at a loss, one could argue you're wasting power and polluting needlessly compared to using a tailor made algorithm (especially if you already have access to one in the form of lastfm pro). Make of that what you will.

[dogwitch]

1 hour ago, Sauron said:

I don't think there's anything unethical about getting music recommendations from a set of your own listening data per se. In terms of using LLMs broadly you should bear in mind that right now the power cost is being hidden from you because these companies operate at a loss, one could argue you're wasting power and polluting needlessly compared to using a tailor made algorithm (especially if you already have access to one in the form of lastfm pro). Make of that what you will.

they dont us as much power and polluting. there are other sector that do far more on that.

[Sauron]

2 hours ago, dogwitch said:

they dont us as much power and polluting. there are other sector that do far more on that.

I said "compared to using a tailor made algorithm", I didn't say it's the single most polluting thing you can do.

 

Also, while there may be a couple sectors which overall pollute more, the AI industry's contribution to air pollution is significant and not to be dismissed. https://spectrum.ieee.org/data-centers-pollution

[Millios]

14 hours ago, Timme said:

ChatGPT is my go-to anime advisor. 

Elaborate on that one
Wdym "go-to anime advisor"?

If you mean that you need an AI to sum up an anime for you I would suggest just reading the brief descriptions studios and mangas already give you

16 hours ago, Mattson said:

It is absolutely staggering the amount of hate out there for AI.  Is this even an unethical use case?  How do you get through to people on the issue?  I know you can lead a horse to water and you can't make it drink but if they'd just take a sip their minds will be blown.

Lemme put it this way cause this is why I personally hate some AI and some of its uses

 

If you need it to do your work (aka let it do your actual work for you) like a programmer asking chat GPT to write his code for him, steal the work of others to generate slop content, use it to trick others etc etc then yes the hate is justified and 10000% needed cause you are causing harm to others and to yourself

 

If you use it for tasks like yours aka "find me music I might like" where you never find it the works of others, or as a tool (for example in video editing rotoscoping can be done with AI assistance and it saves HOURS of work) or something as stupid as these italian brainrot memes nobody will gaf or they will understand

 

Here's the thing tho, the most widespread usecase is generating images using the actual work of others so you are hurting their livelyhood and wasting resources for a free crappy copy

 

The hate for many tasks is 100% needed, the hate for other tasks is from ppl who dont see how useful it can be

[Mattson]

11 hours ago, Sauron said:

I don't think there's anything unethical about getting music recommendations from a set of your own listening data per se. In terms of using LLMs broadly you should bear in mind that right now the power cost is being hidden from you because these companies operate at a loss, one could argue you're wasting power and polluting needlessly compared to using a tailor made algorithm (especially if you already have access to one in the form of lastfm pro). Make of that what you will.

Oh I don't care about the environment so this isn't a concern of mine.  I'm a modern day Daniel Plainview and I have a straw, and my straw stretches acrooossss the room...

[Mattson]

6 hours ago, Millios said:

Elaborate on that one
Wdym "go-to anime advisor"?

If you mean that you need an AI to sum up an anime for you I would suggest just reading the brief descriptions studios and mangas already give you

Lemme put it this way cause this is why I personally hate some AI and some of its uses

 

If you need it to do your work (aka let it do your actual work for you) like a programmer asking chat GPT to write his code for him, steal the work of others to generate slop content, use it to trick others etc etc then yes the hate is justified and 10000% needed cause you are causing harm to others and to yourself

 

If you use it for tasks like yours aka "find me music I might like" where you never find it the works of others, or as a tool (for example in video editing rotoscoping can be done with AI assistance and it saves HOURS of work) or something as stupid as these italian brainrot memes nobody will gaf or they will understand

 

Here's the thing tho, the most widespread usecase is generating images using the actual work of others so you are hurting their livelyhood and wasting resources for a free crappy copy

 

The hate for many tasks is 100% needed, the hate for other tasks is from ppl who dont see how useful it can be

I agree with this sentiment 100% but the rub for me is that people treat every use case under the lens of its most popular use case and I find that infuriating.

[dogwitch]

10 hours ago, Sauron said:

I said "compared to using a tailor made algorithm", I didn't say it's the single most polluting thing you can do.

 

Also, while there may be a couple sectors which overall pollute more, the AI industry's contribution to air pollution is significant and not to be dismissed. https://spectrum.ieee.org/data-centers-pollution

study never peer review.

there are other peer review on this, and other sectors. this is a drop in bucket compare to far worst sectors.

but yeah this ai claim was de bunked already.

 

[VirtualDelinquent]

I think that some of the hate for AI stems from people abusing the creativity aspect of it. Creating AI photos and videos for spreading misinformation, for example, a lot of those are on Fakebook and other social media. Some people also fall victim to a hive mentality, they'll believe whatever people tell them regardless if the information is true or not. 

Although, it's interesting to see what AI can do when it's ethical. I've been playing around with some of the image tools myself and ended up using one of them for a new custom mousepad. The initial image from AI wasn't that great and would turn the original photo into The Scream if I had it change anything, so I edited it myself and it turned out somewhat decent, all it needed was a human touch. That being said, it's by no means a replacement for an actual artist. 

 

[Millios]

11 hours ago, Mattson said:

I agree with this sentiment 100% but the rub for me is that people treat every use case under the lens of its most popular use case and I find that infuriating.

Thats just the internet

 

Nowadays every dumbass with a phone will spread their dumb opinion like "raw milk is safe and good for you" or "raw meat is better for humans" (no it is not against guidelines to call this medical misinformation dumb because it can and does cause serious health issues) or genuine propaganda/bait content on many apps

Case in point, Trump had his son make an AI song for his campaign

 

If you sit down and pay attention to what every single person has something to say you might genuienly start hating everyone cause both sides have people nobody can stand

For AI bros its the "Why do it myself when chat GPT/Grok can do it for me" for every single thing in their life instead of small tasks nobody cares about 

For the Anti AI crowd its the "Any and every use no matter how beneficial and ethical it may be is wrong and AI needs to be destroyed at all costs"

The true neutral and imho "best" opinion is this

-Critize poor unethical usage and mishandling of AI, endorse proper ethical usage by profesionals or use for ethical inconsequential tasks

[Kisai]

2 hours ago, Millios said:

 

For AI bros its the "Why do it myself when chat GPT/Grok can do it for me" for every single thing in their life instead of small tasks nobody cares about 

For the Anti AI crowd its the "Any and every use no matter how beneficial and ethical it may be is wrong and AI needs to be destroyed at all costs"

The true neutral and imho "best" opinion is this

-Critize poor unethical usage and mishandling of AI, endorse proper ethical usage by profesionals or use for ethical inconsequential tasks

No. That's how we end up in the Skynet nightmare scenario.

 

You NEVER let the machine make a management decision. You never let it fire a weapon, you never let it fire a worker, you never let it ignore a safety protocol. The AI is there to protect your existence and itself, in that order.

 

Like good god we had one of these just the other day of some idiot streamer putting a lava lamp on the stove and the AI going "it's going to explode" and the human just blissfully ignoring it until the other human goes "someone was killed doing that." Both the other human and the AI are not in the same room. 

 

I'm sure I can find plenty of AI recipes for food that is not only inedible but recommending things that are lethal, like using rat poison to make green colored food.

 

Asimov was right.

Quote

1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

Right now these godawful AI's can't even do the first part right.

[Millios]

1 minute ago, Kisai said:

No. That's how we end up in the Skynet nightmare scenario.

 

You NEVER let the machine make a management decision. You never let it fire a weapon, you never let it fire a worker, you never let it ignore a safety protocol. The AI is there to protect your existence and itself, in that order.

 

Like good god we had one of these just the other day of some idiot streamer putting a lava lamp on the stove and the AI going "it's going to explode" and the human just blissfully ignoring it until the other human goes "someone was killed doing that." Both the other human and the AI are not in the same room. 

 

I'm sure I can find plenty of AI recipes for food that is not only inedible but recommending things that are lethal, like using rat poison to make green colored food.

Oh I never said "let it make important decisions" or "let it do the work for you" (included in the inconsequential tasks part of what I said)

Proper ethical usage by profesionals in my eyes and how I use it is "use it as a tool like the calculator is"

Some examples are:
-Compile pre-existing research data

-Categorize pre-existing data/terms

-In some spaces like video editing, make X process faster (like rotoscoping and masking)

 

Letting the AI do important work on its own with 0 human input is a piss poor job done like some junior programmers who let chat GPT write code for them and they dont check it

If you are properly taught and know how to properly and ethically use it it can be a tremendous tool in many fields.

Case in point, Dr Mike at youtube , a trained profesional doctor with years of experience as a family medicine doctor, shares with others the opinion that AI ,when used to compile results that have actually come up in research or bring up results from relevant studies that are peer reviewed and accepted as fact, can help the medical field expand and evolve WAY faster than we as humans could without it

[Hideki Ryuga]

1 hour ago, Kisai said:

Asimov was right.

The whole premise behind his book was the flaws within these laws.

[Mark Kaine]

Eh,umm, computer generated music is a thing since decades... People don't hate it, I have to guess maybe the music you're "AI'ing" just isn't very good? 

 

Ie. I don't think current AI is capable of what some rather generic computer programs could do ages ago. 

 

Edit: maybe I misunderstood this, it's all about making "playlists"...? If so then where does the "hate" come from? I don't understand lol  

 

 

12 hours ago, VirtualDelinquent said:

think that some of the hate for AI stems from people abusing the creativity aspect of it. Creating AI photos and videos for spreading misinformation, for example, a lot of those are on Fakebook and other social media. Some people also fall victim to a hive mentality, they'll believe whatever people tell them regardless if the information is true or not. 

That's the thing I heard "AI" generated music, you can't even tell for sure it's AI, I can definitely tell it's bad and shallow though, not what I'm looking for when I want to listen to music.

 

"AI" at this point really just stands for "my poor code" or alternatively "the new crypto scam" (yeah it works differently but is the same idea that's behind crypto, aka make money out of literally almost nothing with the least effort required) 

[Mark Kaine]

4 hours ago, Millios said:

"raw milk is safe and good for you"

It actually kinda is, but it's also indeed risky... I think the trick is the time it's actually safe and good... I'd say maximum 10 minutes after milking... Done that a gazillion times and never had any issues with that.  Otherwise if there's a longer timeframe even 1 hour then it's probably better to sterilize the milk (maybe I'm missing something, can you even buy none sterilized milk, you can't around here with very few exceptions, and only for direct and immediate consumption)