I used to love playing this PC game when I was a kid in the 1980s. Brings back so many aromatic memories!
I’ve been watching for a while now the evolution of job ads around the use of generative AI for creative production. It’s almost always for the purposes of marketing, which comes as no surprise, of course. Though you do very occasionally run across “content factory” manager jobs, or positions directly tied to social media management.
This one below is by a company called Polywood that seems to make outdoor furniture out of recycled plastic, looking for an “AI Imaging Artist,” and it squarely falls into the marketing category above. Though, I think it’s also noteworthy because it seems to be one of the few that tries to bridge that gap between marketing and product/conception – which I think is exactly the right direction for integrating this kind of work. I’ll see if I can’t pluck a few choice snippets from the LinkedIn ad. It kicks off with this:
Create images people won’t believe aren’t real.
Create what doesn’t exist. Inspire what comes next.
Personally, I like real wood (and spent the better part of this summer making picnic tables). But the job ad still keeps me reading, which is honestly kind of rare these days:
Working alongside our Creative team, you’ll create realistic outdoor environments, visualize products before they’re manufactured, develop localized lifestyle scenes, and bring creative concepts to life for product launches, advertising, email campaigns, social media, e-commerce, and more.
Every image you create will need to do more than look beautiful. It must accurately represent our products, reflect our brand, and help customers imagine life with POLYWOOD furniture.
I’m not sure just how advanced they mean with regards to “visualize products before they’re manufactured,” but this is personally something I think is highly worth leaning into. Find some way to rapidly create and float out publicly theoretical/conceptual designs created using some kind of gen AI pipeline, and when they’re able to generate ___ level of market proof and public interest, you create new products based around that.
Reminds me of another company, Arcade AI, that lets users design jewelry and other types of product categories with an AI assistant, and at time of purchase, the physical manufacture of the (presumably one-off?) design you created with the tool gets routed to a vetted human artisan or production shop. I think this is kind of a cool direction?
I guess this is ultimately an “I’m old enough to remember when…” post. In this case, I’m old enough to remember when journalism actually reported on things that happened, instead of merely being opinionated ranting. Don’t get me wrong, opinionated ranting has its place. And I happen to believe the apex of that particular mode of technological expression was/remains blogging. Hence, here I am, engaging in the same… but I’m at least not calling it “journalism.”
The other “I’m old enough to…” piece of this particular article linked above from The Verge is: I’m old enough to have worked on prior generations of this content-labeling problem, especially around fake news, satire & mis-/disinformation. For literal aeons, users of web platforms have asked the product masters of those platforms for effective labeling of all different kinds of content. And while many trial solutions have been attempted, very few of them have ever stuck, or caught on in a massive way.
Why? Because accurately detecting different types of content is a complex, costly, and highly ambiguous problem. As well as being potentially highly volatile politically. The author of that Verge piece states:
“Surely improving the user experience for your millions of users is a worthwhile investment to fend off competition?”
While the casual well-meaning outside observer of a web platform might be forgiven for assuming that’s how it works, I can say as someone who has seen a little bit of the workings of the sausage factory at least, that this is decidedly *not* how it works. My experience has been instead: users ask for a thing (often for years), and thing rarely gets built unless it meets the product vision and business case the product is aiming to solve for.
The author of this article also seems to implicitly think that everyone else thinks like them and nobody IRL wants to see gen AI content on web platforms. But my recent anecdotal research – in this case experimenting with and observing gen AI-assisted anonymous ‘faceless’ Facebook meme pages – strongly suggests that the market overwhelming *wants* so-called AI “slop.”
Consider the following example:
This is a meme I stole from another AI meme page, and had ChatGPT come up with a new variation of the “punchline” text. I like to think of this technique as ‘stolen valor’ – you just take things you see performing well elsewhere, and copy them outright or create slight variations them. But the thing is, dumb AI-heavy memes like this work. They get engagement, they get likes, they get comments. So much so that there is a cottage industry of people who have connected their FB pages to automation services backed by gen AI tools, and those accounts just churn out boatloads of narrowly focused and consistently branded content on a given theme: single moms, coffee lovers, US females 55-65+ years old who like saying “good morning” to each other online.
With gen AI, you can now serve (or create) any kind of niche content with the greatest of ease. It’s a bonanza. This YouTube video gives a decent overview of how simple it is to set up a system like this:
My impression after looking into this a good bit is that people who are philosophically or politically against gen AI in the media or on social media do not accurately represent the vast majority of people who are simply happy to interact with cute content that makes them feel mildly good without challenging them too much. And that happens to be an area generative AI excels in.
So, if my hypothesis is true: that most people actually *like* or at least engage with gen AI content, then why would platforms want to create a way to enable audiences to filter out what potentially is one of the highest-value content categories for everyday users? The Verge author almost gets there, but via different formulation:
“Allowing users to filter it out regardless would go against all the effort these platforms have undertaken to profit from AI: They want you to embrace the slop factory.”
I don’t think that’s quite true though. They don’t want you to embrace the slop factory. They just want you to embrace the factory itself. As long as you post your article links and meaningless hot takes on social media, and beg people to like and subscribe, they don’t care which department you want to work in: the Slop Side or the Authenticity Brigade. It doesn’t matter. They merely want users to be eternally subject to the totalizing effect of their product ecosystems. And as long as people keep feeding those machines, letting them run their lives, and begging for changes from those who administer them, nothing will happen. Everything will just keep chugging along. All the way from the lowest of low-effort homebrew AI meme FB operators on up through the Complainer-Industrial Journalism complex. We’re all feeding at the same trough. The systems that need to be changed go radically deeper than “AI.” If we stop the critique and the inquiry there, we’re basically missing the boat entirely. And as Adorno said, wrong life cannot be lived rightly. Not with all the settings and filters in the world.
Had a fun time talking about using AI to help produce books on the Hello Future podcast on IHeartRadio. Listen here.
Had fun recording this appearance a few days ago on the Kim Komando radio show & podcast about my AI books and albums.
The folks over at Campside Media did an excellent job distilling a complex story down for a general audience. I enjoyed participating!
I meant to reply to this Consequence of Sound AI music article from December, but household affairs interrupted everything for me.
Don’t have much to say here, except that it’s a shame an authentic, deep, and personal interview got reduced to… this muddle.
My question for journalists at this point is: if Andrew Frelon is so inconsequential in AI music, etc., then why does everyone feel the need to keep talking about him and his work?
I’ve had a number of things going on which have prevented me from updating here much, but just wanted to point out that somehow Billy Joel released an AI music video almost a year ago that I never heard about until recently:
I don’t much care for the AI effects here, as I find the facial expressions don’t really feel like they match up too cleanly with his, if you look at vintage video of him (like the BBC Old Grey Whistle Test videos), but not a bad try I guess. There’s just something lacking about the musculature below the surface for me… I do love the song though, and have been playing it on guitar.
Basically the title: music journalists should stop parroting Spotify’s PR claims about how they’re taking action around AI music. There are tons of examples of this, but Christianna Silva’s piece on Mashable (from one month ago) is one that caught my eye this morning. Everyone seems to have just uncritically reported on the words Spotify said, and not digging deeper at all into the actual actions Spotify has taken so far.
I realize because of Spotify’s monolithic position in the industry that even just them mouthing the words “we are doing something about AI” is somewhat mildly newsworthy, but in my experience as someone who uploaded a lot of AI music (820 songs) over the past several weeks is quite the opposite.
Silva wrote, at the end of September following Spotify’s announcement:
On Thursday, Spotify said it would start doing just that, saying in a press release that “aggressively protecting against the worst parts of Gen AI is essential to enabling its potential for artists and producers.” The platform is integrating a new spam filtering system, AI disclosures, and “improved enforcement of impersonation violations” like deepfakes.
As someone who spent years working enforcement for a platform, none of these statements give the impression of anything other than enforcing existing rules, and doing a sudden big sweep to give the public impression something is happening. It’s reputation management, imo, and little more.
Why do I say that? Because, as I said, I uploaded a huge amount of songs in a short time. In one case, I uploaded 300 AI songs in one night. There’s not even a way to label them as AI at time of upload in Distrokid, let alone surface that label in Spotify or allow users to take action on it.
All this reporting pretty much rests on taking Spotify’s word at face value, which can be problematic in journalism, as you end up whitewashing the message of others to appear more legitimate than it might otherwise seem.
Via Rolling Stone India:
As David Bowie once said, “Always remember that the reason that you initially started working is that there was something inside yourself that you felt that, if you could manifest in some way, you would understand more about yourself and how you coexist with the rest of society. I think it’s terribly dangerous for an artist to fulfill other people’s expectations — they generally produce their worst work when they do that.”