DeepMind releases Lyria 2 music generation model

I think there’s an even bigger issue than this.

Many people would not know how to live under such a system. By this, I mean that I strongly believe people would become severely depressed or insanely stir crazy.

I’ve been on an extended sabbatical after 20 years in tech. The first year was magical. “I could do this forever” I told myself, and actually considered it.

The second year was more complicated. I could feel myself drifting away from myself. The structure of work and the rewards of working on big projects were now fully missing, and I could feel this growing emptiness that needed to be filled.

For health reasons, I entered the 3rd year, and by that point I needed more major psychological intervention. I’d become severely depressed and while I knew that getting back to work might help, I was now in a position where going back to work sounded impossible.

I’m not claiming that my experience is universal. But I’ve started to find more accounts that are similar to mine. I’m also not saying it’s impossible to replace work as a form of necessary challenge and satisfaction. But the societal structures do not exist to fill the void.

For better or worse, we’ve been a species that relies on “work” in some form to live. I use quotes because clearly this has looked different ways over time. Hunter/gatherers certainly had a different set of tasks than the modern city dweller.

But ultimately I’m not convinced that we’re equipped to live satisfying lives without some form of striving for survival. In a post-work era, I think a lot of us will go some kind of crazy or experience depression.

I don’t think most people are aware of how awful things can feel after enough time away from work has stacked up.

It reminds me of that feeling when going on vacation somewhere nice. “I could just live here forever”. But the reality is that the thing that makes the vacation feel incredible is the contrast from normal life. Remove the contrast, and things become pretty flat.

Edited to change “most” to “many” in the 2nd paragraph because that better reflects my belief.

The idea that we no longer need labor is preposterous. Where are the machines to take care of children or the sick, or to build and maintain housing?

What kind of science fiction world do you think you live in?

Who is "we"? The entire human race?

Expecting all humans across different cultures and languages to come together and figure out basic income for 9 billion people is absurd. This kind of cooperation never happened and probably never will. People are completely unable to cooperate at the massive scale this requires, let alone solve far smaller challenges like mitigating outbreaks or making an effort to avert climate change.

"We" is not a thing.

It's called Moravec's paradox: https://en.wikipedia.org/wiki/Moravec%27s_paradox (he is cited by the Elephants dont play chess paper)

I think you're missing the point of that statement a bit.

Have you heard that AI music? It's absolute shite. AI is a much better boilerplate coder than a musician.

I guess automatic sorting, washing at the right temperature, drying and maybe folding is what someone who washes their own clothes would wish for.

Talking to the appliance is probably not that high on their list.

> the only laundry-related AI you can hope for, is a cloud connected smart speaker telling you it can't wash with unapproved third party detergent pods

This seems unnecessarily fatalist.

Laundry folding machines exist[1] and there were attempts to create a consumer friendly one, so far unsuccessful. Technology advancements could make that happen. At least that's what I'm hoping for.

[1]: https://en.wikipedia.org/wiki/Laundry-folding_machine

> I want Al to do my laundry

Do you want a washing machine with Alexa built-in? Be careful what you ask for.

https://www.samsung.com/uk/washers-and-dryers/bespoke-ai-lau...

Someone (probably from here) would develop a local inference model probably - or hopefully.

Perhaps much of what humans be doing in this regard is - validation, and the physical work left. But we're not there yet, of course. Vibe coding takes a lot of manual labor, and it's nowhere near for the actually complex tasks such as... multi-part CTEs munching gigs of spatial-temporal data.

Speaking of AI in music - well, perhaps many will welcome some tools when you have to:

- clear hissing - process levels in tedious clearing - auto-removal of aaah, oooh, eeerrmm and similar - podcast restoration, etc.

but of course, nobody wants darn model singing in the mornings, and composers definitely don't need anyone to make up melodies, drum rolls, or bass lines for them.

I see deepmind advance their offering, still I find it difficult to imagine any of my producer friends embracing such abomination, and particularly giving it is a remix tool before all else, and not a composition tool. People love details the same way a painter loves details.... dilettantes think all this irrelevant, they really can't be wrong more.

> Well you know, all those fun and creative parts in software engineering has been taken over by vibe coding and now

Maybe on your YouTube shorts playlist but not in real life. People doing real work are not vibe coding. The previous perpetual react learner turned ai vibe coder certainly is doing vibe coding, but not for money from a job.

> Well you know, all those fun and creative parts in software engineering has been taken over by vibe coding

What? In what way? Fun and creative parts are thinking about arch, approach, technologies. You shouldn't be letting AI do this. Typing out 40 lines of a React component or FastAPI handler does not involve creativity. Plus nobody is forcing you to use AI to write code, you can be as involved with that as you'd like to.

I think nightcore versions of your pieces would be of interest...

[dead]

The anouncement of the "death of the band" is premature.I dont believ this for a second, nothing can replace live music played by humans, as the experience and the very real positive benifits of an audience comming together can not be digitised, in spite of the many attempts to package high tech simulated environments. There are hundreds of thousands of full time touring musical groups world wide, and millions of music festivals every year. Just because the bar is high, very high, to become a working musician, and has been forever, does not give credence to the notion that some nameless randoms get to make the call because there tube vid got 3 views.

In my opinion the referenced music fragment of that video is fairly "unmemorable", it is "just" a sequence of harmonized textures. And that is precisely what AIs (LLMs) are good at.

(I listened for ≈7 min from the reference point.)

(I am talking about the music, not the live performance itself.)

That what most people don’t care about

> theoretically it can react to crowds better than any human could

That's one area I'd expect AI to do poorly. Performance is a two-way dialog between performers and the crowd, with facial expressions and body movements from both the stage and the audience in communication. I'd expect any AI that's not attached to a humanoid robot to be less exciting to a crowd.

However, I am very excited about AI in some of the other contexts you mentioned, like as a music-writing or editing partner.

It foresaw streaming, but it was a direct response to Napster and file sharing.

Is there any big difference between using that and instead doing playback lip syncing and fake playing the guitar, like already happens sometimes at concerts?

You can convert your guitar signal to midi and quantize it.

Not sure that would have helped Jimmy "sloppy" Page getting famous though.

We've had that for years in the DAW and autotune and snapping to grid.

The result's pretty boring and interchangeable, and that's largely what AI music is trained on. Accuracy is not novel here. Ever since the 80s it's been increasingly possible to augment musical skill or lack of, with technology.

I don't think we're very close to correcting for sloppy intentionality. Only to correcting 'mistakes', or alternately adding them in the belief that doing stuff wrong is where the magic is.

My theory is that as the quality of these generative tools increase, we'll see the public opinion of them slowly shift. Regardless of philosophy (although discussing it is always fun), it just seems inevitable since there are so many more consumers than producers. And as you say, consumers are the ones that will primarily benefit from this new technology. As a consumer we care primarily (some could argue solely) about our own emotional reaction to the music —or more generally put, art-piece.

In practical terms I also believe that this will give rise to a lot of new consumer behavior, and, as you so aptly puts it "creative consumers" will become normal.

The ability to on-demand create more content to fill out some very narrow niche is a great example ("Today I want 24 hours of non stop Mongolian throat singing neo-industrial Christmas music"). Or maybe to create covers of songs in the voices of your favorite long dead artist. Anything from minor tweak of existing works ("I wish this love song was dedicated specifically to ME", to completely new works (Just look at how much the parody-music genre has grown since Suno and the like first appeared). The possibilities are near endless.

I agree, the story behind the work is part of how we humans view a creation and cannot be dismissed.

I think we're a long way away from 100% algorithmically created content. This far all I've seen is content that is created based on human inputs and ideas. I'm not aware of the Instagram incident you mentioned, but it too seems like the brain child of a human if I'm not mistaken.

There have been trending AI generated videos floating around lately for example. Which I found surprising at first. But they still had a human script writer (prompt writer?), director and a human editor. Someone who had a vision of what they wanted to create and share. My prediction is that this human-directed tool-like usage will be the standard for a long time, so I'm not particularly worried about humans getting removed from the process.

The thing is that in this century the creative interaction seems to be moving up the value chain- it wasn't long ago that people would say that being a DJ wasn't creative. Simply selecting songs wasn't creative. Now lots of people consider DJs to be creative artists that are communicating something with their human will of selecting and mixing tracks.

Unless the whole thing moves to a random AI generated slop stream app, whoever turns the knobs of the AI that creates the music will become the new "artist". Right now it doesn't seem like the AI creator "does" anything, but maybe future people will think that.

Personally I don't like to gatekeep art.

For example: If someone walks out into the wilderness and encounters a particularly fascinating rock formation or plant, something that was created completely by accident and without a artist or designer, but they find that the sight instills in them strong emotions or deeper thought, I believe they should be allowed to call that art.

Maybe this is just petty linguistics and semantics though, in which case we're drifting away from the topic at hand, and I'm sorry.

I could see myself listening to AI music in Spotify, not as the main act but at least as a Plan B.

For example, I like a specific music genre, Italodance, which was popular in the 90s and then disappeared. The problem is that I have listened to all of it, as far as I know. No more is being made. If an AI model could make more for me, with decent quality, I'd probably listen to it.

Even music found on Spotify is artificially generated, carefully disguised as human curated

https://harpers.org/archive/2025/01/the-ghosts-in-the-machin...

Discussion: https://news.ycombinator.com/item?id=42461530

> We like the actors, the director, the artist, their taste, and who they are. It's why we have celebrities, because we get invested in the people behind the art.

But we can't know any of these celebrities as people. We only engage with their images created by marketing. Their stories are as curated and fabricated as the artworks they produce. Transferring these simulacra to AI personalities is merely another marketing problem to be solved.

Pop music is so simple, yet so difficult to make a hit. For some artists the music can be mediocre or even fairly bad, and still be a massive hit because pop music is essentially theater and their persona and mystique carries the day.

Some bands were terrible touring artists and rarely put on concerts yet made great careers as studio acts. Steely Dan would be one that produced many hits yet rarely toured, mostly later in their career.

The fundamentals of pop are totally understood. Yet what makes a hit is so fickle and difficult, the bar is extremely high

> The thing is for quite a lot people, music is just something they put in background w

You are absolutely right. These people only listen to music passively and so it doesn't make a big difference who/what made these tracks. Same for lots of commercial music (cheap TV show soundtracks, commercials, jingles, playlists for restaurants or shops).

But for anyone who actively listens to music and appreciates the style and evolution of certain artists, AI music is not acceptable. The very premise just feels wrong, if not outright insulting.

Identity of the creator and the piece has always been a big part of music appreciation. Imagine how dystopian would be if you ask someone their favorite and the answer is "the music that HitAI generated when promoted 'sad song for happy people' when I was 17, but their model has since changed and my subscription didn't include offline ownership, it's lost forever."

People have said that about robots and computers in the workplace, and indeed, since the 60s, more and more jobs have been automated. And there are less musicians now. Film scores and albums are produced with samples instead of bands or orchestras, reducing demand for session players, leading to less income for musicians, leading to less musicians.

And while automating dangerous jobs is a good thing, generating AI music isn't. It's not as unethical as generating deepfakes, but it's useless, and bad for society.

Good art requires a huge reservoir of artists. When AI music drives musicians out of work, it shrinks the reservoir, thus leading to lesser art.

The answers to your questions are in the comment you replied to. Part of their love is music is sharing it with others. They also like to fantasize about becoming a full-time musician. Both of those things are less likely if there is 100x the current volume of music from unknowns.

It's not about validation, it's about expression and communicationw with other humans. That's one of the key beauties of art and it's being flooded away with artificial, empty content

Find some good independent online radio stations that you like. There are successful ones out there, nts.live and rinse.fm in the UK for example. Obviously they exist in other countries too.

I say independent as most radio is stacked with adverts, but the above two seem successful without needing them.

I find the human curation far more satisfying than an algorithm, and most DJs want to support human artists not bland AI nonsense as they have a stake in the music industry.

SEARCH TIP: As the default search prioritizes newer content, it's quite tedious to find older takes on youtube.. there's a workaround though: web search for site:youtube.com and set a custom date range (ddg: click on "Any time", g: "Tools" on the right side) to sidetrack the big tech attention economy brainrot algorithm.. : D

Just imagine private model (own, little spech-to-text model) listening to the people playing DnD feeding to the bigger brother that in turn controls the music generation and feeds it back to the speakers as the story progresses.

:)

"Twitch plays (generates) music" goes brrr.

iMUSE on steroids. It will come for sure.

I think producing new and interesting music with AI tools will require models that allow for far more granularity, tweaking multiple tracks / layers, and this has to be exposed through a more professional and capable UI. It has to go hand-in-hand if it's going to be treated as anything other than a novelty or a toy.

Vibe coding has improved significantly in tandem with UI innovations that provide a more intuitive interface to the workflow. Although in the vibe coding space there's still a lot of room for innovation and exploration, especially when doing detailed task development.

> > cathartic when you're sitting at home in the dark because the power company continues to fail

> Ironical remark about the power drawn by IA assisted creation left to the reader.

Thanks for pointing that out, was scratching my head on that

The power grid in my country has been failing and unstable long before GenAI became a thing. We also don't have any AI data centers here that would be taxing said grid. But sure, enjoy making your glib drive by snarky comment.

TBH at this point HN should have a flag for "No Demo"

$10 they have no product, just a wait list and a note to embrace and extinguish some random music ai startup at some point in the nebulous future.