2023 Andrew Scheps • Thoughts on AI and Machine Learning

We were thrilled to recently have legendary Grammy Award-winning Mixing Engineer Andrew Scheps join us for an exclusive interview. Interested in studying music production in London, head to icmp.ac.uk/production
Andrew Scheps has engineered and mixed some of the biggest rock bands in the world: Green Day, Red Hot Chili Peppers, Weezer, Audioslave, Black Sabbath, Metallica, Linkin Park, Hozier, Kaleo and U2.
----------
Get social with us!
----------
Instagram: / icmplondon
TikTok: / icmplondon
Facebook: / icmplondon
Twitter: / icmplondon

[@MustakimAlMahdi]

"And I'll be proven wrong" this guy gets it

[@alextotheroh8071]

Some interesting takes on machine learning. It really goes to the heart of what creativity even is. I'd argue that all HUMANS are doing when they're trying to be creative is combining past experiences into something that sounds novel. I would not be surprised at all if AI is soon able to be "creative" in ways that utterly shock us, even if all they're doing is rearranging training data.

[@CrushingAxes]

Nice Andrew is awesome! Such a legend!

[@theICMP]

We couldn't agree more!

[@jjm9741]

uses Sony $99 headphones

[@SteveMcRayMusic]

Great insight! Thanks Andrew...!

[@RadioMartyT1B]

When Andrew Scheps speaks....I listen.

[@RandallFishwick]

A lot of great insights here. Especially worth a listen for those starting out in music production and recording.

[@XC-Fly]

"Listen widely" is my favourite advice. Especially, when i am asked, what kind of music i listen to. Music is about stirring emotions. Every genre trigger some specific emotions. No genre have the ability to touch all emotions to the max level. The amount of combined emotions reflects the endless number of genres and sub genres. When you digging only one kind of music, you`ll definitly miss fantastic music. If you like Metallica, you`ll probably like Vivaldi. Leave your "tribe "as early as possible To know the history of a specific genre, helps to understand that kind of music. Didn`t know nothing about blues as a teenager, but loved HardRock....and didn`t like Led Zeppelin. The understanding helps to feel it. If you don`t feel it, you don`t understand it. But you don`t have to like it. The charts2023 are an impertinence, like my english...

[@gregoryivesdolbyatmosmusic]

Great

[@Aeoxmusic]

A wise man once said "Listen to Andrew Scheps"

[@antiHUMANDesigns]

AI and such things are good sanity checks: I often throw some AI-based smart QE on the master or on a track to see what it suggests, just to see if my ears have become tired and I'm doing something really weird. Then I remove those and keep working manually, instead.

[@audiosignalpro]

I put the transcript of this into ChatGPT and it created a small black hole

[@drrodopszin]

Very tangential, but I got interested really in why are there so little amount of women in the mixing industry. So I started listening how people talk about why they ended up with audio. So far I never heard that anything gender related, but things like "I'm a nerd, and I got sucked into this, because I have found it fascinating."

[@alexandre7634]

Almost every great engineers learned from a great mentor, I think people weren't open minded about women working in the industry, I hope it's changing now

[@kostoglotow]

@@alexandre7634 oh please, enough with the "patriarchy" conspiracies. it's well established by psychological research, that on average men are much more interested in THINGS, and women in PEOPLE. cross-culturally even. hence more male engineers, and more female nurses/counselors etc.

[@drrodopszin]

@@alexandre7634 it's a bit anecdotal to say "every great engineer" without the stats, but alas, I was also not doing any of it.

[@alexandre7634]

@@drrodopszin Tom Elmhirst had Trevor Horn, Jaycen Joshua had Dave Pensado, CLA had Bob Clearmountain, and a lot of other engineers started in studios as interns like Manny Marroquín at Larrabee, they all had knowledge and feedback from amazing people

[@Matheus-ly6eu]

@@kostoglotow It's not "conspiracy", man... It's very logical. People need to see people like them achieving things to compute that they can, actually, achieve those things themselves. There most definitely is an unfortunate heritage from less inclusive times, in all arenas. Sometimes it can seem like we live in a "post-prejudices" world, especially when we don't feel it in our own skin on the daily, but it's not that simple...

[@iammusicp]

the cup of coffee is making me so nervous.

[@theICMP]

only Andrew's trusted with coffee in our studios ; )

[@alanduncan1980]

When drum machines were invented, they said it would replace all drummers. We know how that worked out. I don't know if technology will ever replace real humans playing real instruments.

[@theICMP]

So true!

[@citizenworld8094]

A.I is great. But my brain is greater. Andrew's brain is ever greater and he won't be proven wrong.

[@stevekudlacek]

AI = Boring Records… You should hear that, creative people!

[@ramspencer5492]

Machine learning is really cool for some things.... Right now people in charge of AI "songwriting" services have been massively cheating to try to make I have good performances... And human like timing It literally just stealing existing performances without consent... And morphine more than one voice together... Again without consent.... I'm trying to use AI to change the melodies and the performances enough that it will get by existing copyright infringement laws. These lights were never made for the era we live in. And the laws need to account for that. This direct stealing and morphine without your performances without permission needs to end!

[@antiHUMANDesigns]

Eventually, inevitably, creative work will be replaced by AI. That's just reality. Ai can do what we can do, very soon. Scheps is underestimating what AI can do. Our brains do the exact same thing: It combines the things we've heard into new things. AI can do the same thing, and is in fact already doing that. Neither an AI nor a human brain can create something from nothing. Both are neural networks that learn by training.

[@radiobar1634]

That is exactly what people don't want to believe. But you are right: when humans write new music, they "just" re-combine data/a vocabulary they've learned their whole life by listening to sounds/music

[@Bati_]

I’m a psychologist and cognitive neuroscientist in training and I also agree with this, however, one thing most AI models still lack is reasoning and general knowledge about the environment, so called “common sense”. However, in the future, I’m optimistic that AI models will be more close to Artificial General Intelligence (AGI) as well.

[@antiHUMANDesigns]

@@Bati_ Yes, the big difference between the AIs we have, and humans, is that humans experience a variety of things from their own perspective over decades. An AI is trained on a specific set of data for perhaps a month. But they have made an AI that combines the functionality of something like 600 different AIs, allowing this AI to perform some 600 different functions. That is essentially a "general AI".

[@Bati_]

@@antiHUMANDesigns I'm a bit skeptic about what you described here, I don't think that this can be interpreted as "general AI". Let me elaborate: my research interest is revolving around the topic of recurrent processing in the human brain. The human brain is a massively recurrent network. We know this from the neuroanatomical and physiological studies. There is an abundant amount of recurrent projections compared to feedforward projections. However, the state-of-the-art neural network models are nearly all feedforward architectures, not recurrent. Therefore, even though they perform extremely well, we couldn't deduce easily that what they are doing, even when combined their powers, will be analogous to humans. This will require a further line of research as a potential avenue. We are not quite there...

[@antiHUMANDesigns]

@@Bati_ The brain is more of a "spiking neural network", right? Those are infamously hard to train, which is why they're rare in AI. I'm not a neuroscientist, but I'm a C++ programmer. Combining 600 AI's, so that the same AI can do what chatGPT does, as well as beat the world champion in chess, play video games, and 597 other things -- it's close to being a general AI.