Hahaha this hits home too hard, back in early 2000s people would moan all the time whenever they spotted a hint of autotune, in 2026 its the industry standard.
I think its really speaks on the incredible ability of people to be able to be stuck in the past rather than new technology being "bad".
This is an amazing comment. I'm old. I was born in the 70s, grew up in the 80s and 90s and miss those times so much. But that is because I was young, immortal, the world was mine to discover.
In 20 years people will be missing the 2020s too. It is just human nature to complain.
That's just not accurate. I haven't studied SWE Bench Pro in detail, so I can't tell you exactly what the flaw is, but SOTA models routinely make bad architectural choices I have to intervene to fix.
TL;DR its very effective as it directly tests model on REAL codebases: "The benchmark is constructed from GPL-style copyleft repositories and private proprietary codebases". The use case is very real.
It doesn't sound to me like this benchmark is attempting to measure architecture design. As far as I see in the paper, they do not evaluate the architectural quality of a task completion, only whether the model is capable of completing it at all.
Could you elaborate? I hear some people say a big model should be driving a smaller model, I hear some people say a small model should be driving a bigger models.
When I have an expensive task that is clearly defined, I will get opus to write an LLM workflow for it, and then I will execute it with a smaller model. (Starting with the smallest one, and then upgrading if the task fails.)
But this is a single well defined task, designed by me and Opus in concert. If I need ongoing agentic work, Opus would be too expensive. I'm not sure if Haiku is big enough to be the driver yet. And Sonnet is probably too big! Haha.
(Grok looks promising, optics aside... Grok 4 Fast was almost there but not quite. Great for interactive / realtime (steered) work though.)
But I'm thinking you need a smallish model which can delegate both up and down. I'm not exactly sure what that looks like though. Cause the model needs to be big enough to know that it's struggling... Instead of pattern matching to something stupid and getting stuck in a loop trying to solve it the wrong way.
All of the major model's memory are handled by smaller more specific models.
I do not know about the future, but I believe, like the human brain (the amylgada + cerebral cortex), AGI will have smaller but more specific submodels running in parallel to craft an compelling heuristic.
The thing about Spotify is that is NOT driven by record labels, it is an platform for the individual meaning an individual can upload their music in an laissez-faire situation.
If they disallow AI artists tomorrow, they are going against what they created the company for.
>then they should close their ears rather than trying to maim others
criticism isn't maiming anyone. I don't know about you but I was taught that debating culture is part of a living society. A lot of people think that the human centipede dopamine machine that is "AI art" is a disaster for us and instead of acting like the three monkeys as mature adults we can critique this
the idea that we should "close our ears" is of course itself the very appeal of AI content, it never challenges or complains, that's the appeal of AI music, AI boyfriends and so on.
Criticism is healthy and != maiming, but it is apparent that mainstream culture nowadays labels "AI Art" consumers as dumb and tasteless. I think this is maiming.
>"the idea that we should "close our ears" is of course itself the very appeal of AI content, it never challenges or complains, that's the appeal of AI music, AI boyfriends and so on." -> People also do shrooms that arguably have a similar effect, who am I to judge?
It is dumb and tasteless and nobody's being maimed for pointing that out, it's DFW's Infinite Jest made real. The psychedelics and shrooms analogy is very appropriate, because it gave us a generation of escapist idiots who fried their brains and retreated into their own fantasies. Instead of going out, producing, debating and collectively making art.
Who are you not to judge? It should be judged, it's narcissistic and solipsistic. if the AI psychosis is comparable to the 60s and 70s at least there's hope it'll exhaust itself as quickly.
Do not agree with what you say at all.
some people are usually afraid of going out because people like you label them narcissistic and solipsistic so they stick to LLMs.
They are soulless because often they are AI covers of existing songs trying to mimic the original artist so that they get substituted in as replacements for the original with limited realization on the part of the average listener.
Which of course that means the AI covers get the listens and the associated revenue instead of the original artist.
So instead of listening to "System of a Down", you get a cover from the AI artist "System of the Down" and now that you listened to that cover you start getting more covers in your recommended from them and eventually you are getting covers from them instead of from the real band since you started listening to them instead.
And even if it's not that extreme, the listener is getting served these knock off covers with no actual person behind them. If the listeners don't realise that's what's happening it will reflect poorly on the original creator and hurt their listenership (which wouldn't be impacted if shitty AI covers weren't being subbed in).
It even gets to the point that now you have artists who have upcoming albums and AI cover artist bots scrape the song list and upload auto generated "covers" of the unreleased original song to try and capture listens that would go to the original artist while people go to pull their music up prior to, on, and after release day of their new album.
Overwhelmingly AI songs on Spotify are autogenerated slop from bots trying to leech off of actual artists by creating a shitty knock off to skim some cash out of those artists' paycheck. (This is distinct from actual cover artists who at least contribute their own unique human touch to the covers).
If you want to make music and you happen to use AI in the process then whatever but Spotify has a major AI cover/clone problem.
"soulless" is a term YOU label people with for enjoying a certain type of music. This is your problem, not the people's problem for enjoying this sort of music.
But I agree with your second point, an AI should NOT BE able to mimic a brand/personality such that it brings harm to them.
Your third point, actually it is NOT distinct from actual cover artists, because if I tomorrow pick a system of the down hum and then remix it and it gets popular, their automated DRM system will probably C&D me and their label sue me to oblivion using the same AI.
> then remix it and it gets popular, their automated DRM system will probably C&D me and their label sue me to oblivion using the same AI.
Yes it will because remixing is not a cover. Remixing is a derivative work of the recording and therefore is subject to the terms of the mechanical license on the recording itself. The copyright on a recording is legally distinct from the copyright on the composition (the melody and lyrics).
You are protected under a provision of the Copyright Act of 1909 to modify and perform your own renditions of the original composition (provided it has been recorded previously). Provided you supply some level of original creative input and don't use the recording in your work, the original rights holder are required to provide you at no cost what is called a "mechanical license" for your cover granting you the right to distribute and sell the recording of your copyright.
In the past, granting of mechanical licenses was between parties (i.e. IP holder reaches out when they discover you made a cover and they grant you a license or you go to court and the court grants you the license. But as of 2021 a non profit body was formed under the guidance of the federal government to handle the blanket granting of mechanical licenses to any and all human covers (provided they did not use the original recording in their work).
If you sample the original work at all then you are now creating a derivative work of the recording and you must negotiate for a mechanical license (which is not required to be free). Likewise if you remix it.
AI nutters always use this argument that bots are merely doing what people are doing: "a person learning from 100 books is just the same as an LLM learning from 100 million books".
If you own an apple tree, and (as is common in some countries) a child leans over your fence to grab an apple, you don't lose sleepless nights over it. But if a corporation comes over and leans over with its 10 meter long robotic arm and takes all your apples then it's a disaster.
Scale matters. People don't care for small "losses" because we want other people to prosper. But AI steals from everyone and brings prosperity only to Amodei and Altman.
because no one wants to be forced to listen to the slop? some reason spotify is allowing them to dump a lot of ai "music" and then they get played without you knowing
i pay to listen to music now i have to not use discovery mode because spotify wants to earn more money by pushing ai slop to its listeners to not have to pay real artists
> because spotify wants to earn more money by pushing ai slop to its listeners
If people listen to AI songs, how does Spotify get a bigger cut, as you seem to be suggesting [1]? Or are you suggesting that Spotify is generating/has rights to this music?
AI slop is quite antithetical to Spotify's business model, they would actually lose money if say tomorrow, AI music dominates the charts, when AI music gets that good, people can just generate their own music and leave Spotify back in the mud.
It's probably the "AI artists" themselves generating false engagement and manipulating the algo for discovery.
I think its really speaks on the incredible ability of people to be able to be stuck in the past rather than new technology being "bad".
reply