Isn’t it funny that many of the best AI chatbots and content generators are now based on subscription models?
Actually, this isn’t funny at all. Most of these systems were trained on content created by others, and most will continue to use internet searches for new content to provide their generative answers, all while ultimately charging us for those answers.
The good news is that major AI companies like OpenAI are gradually signing deals with media companies, which will address some of the concerns that exist between AI companies and media companies regarding content scraping and training. However, the AI landscape is still the Wild West, and for visual artists, the picture is far murkier.
Whenever a new system like Ideaogram comes out, it’s clear that the company behind it has no problem with reusing intellectual property. And as far as I know, they don’t pay a dime to the owners of Superman, Mickey Mouse, and countless other beloved characters.
It is inevitable that some of their publications will look, sound or read like their training materials.
And don’t even get me started on YouTube creators. There are reports that Apple, Nvidia, and Anthopic have trained their AIs on thousands of their videos. Having been a YouTube creator myself for a while, I know how much they depend on video views (and the ad units played on them) for their livelihood.
I’m not suggesting that generative AI systems reproduce artwork and information as their own, unaltered. These systems use their training to shape their work, to know what word should come next and what brushstroke, line, or visual style might be most appealing and make sense in the context of the prompt. Inevitably, however, some of their production looks, sounds, or reads like their training material, and that final product would be nothing without everything the AI models were trained on.
The costs of AI
It’s expensive and time-consuming to develop high-quality, generative AIs and train the large language models behind them. According to a new report from the Washington Post, it took Amazon a year to train its massive Alexa AI update. The costs of that effort are likely being passed on to us.
Unlike most chatbots and image and video generators that have come to market in recent years, Alexa AI may launch with a subscription. It’s not clear if existing Prime members will get it, or at least some form of Alexa AI, for free. I pay nearly $140 a year for Amazon Prime and still find it to be a good value. However, I would expect to get Alexa AI as part of that subscription package. If not, I won’t pay another $10 a month for access to this smarter assistant and at least I’ll still have the Alexa I’ve known and used for more than a decade.
The honeymoon period in AI was short and uneventful, but that is changing quickly and the future of AI access is clearly in the pay-to-play model.
Even companies that don’t currently charge for AI access are likely to do so in the future. Samsung doesn’t charge for Galaxy AI, but when we asked the company’s executives about the possibility of charging in the future, they only said it would remain free until the end of 2025.
There are also reports that Apple may start charging for some parts of Apple Intelligence. The company will only roll out small elements of its AI platform next month and into 2025, but considering how many billions Apple makes from services each year, it stands to reason that Apple would want to incorporate AI into its service offerings and reap at least a small monthly harvest of AI service fees.
A lesson well learned
In a way, I admire the AI industry for understanding something that the established print media only understood in the early years of the Internet when it was almost too late: you can’t just give away something that has been the economic basis of your company for decades.
When the internet came along, it was considered a completely open and ad-free place. At some point, ads started appearing alongside the content of almost every media brand in the world. No one charged a dime because ad impressions more than made up for lost subscription revenue. Until they didn’t. Display ads became white noise and media companies watched their profits plummet. At some point, brands started putting up paywalls and finding other revenue streams like e-commerce.
For a short time, AI was free and accessible to all, but that is changing rapidly and the future of AI access is clearly paid.
Despite its shortcomings, the beauty of the early internet was its wide availability. If you had a computer or got access to one and were online, you were on the internet. Yes, there was a digital divide between rich and poor at first, but because of the low barriers to entry, it was inevitable that eventually the majority of people would have access.
The path to access to AI will clearly be different. With the world’s largest technology companies all working hard to develop their own AI, and many considering tiered access to the highest quality and most useful AI (longer videos, full content creation, more powerful productivity tools), the landscape of future AI could be permanently shaped by haves and have-nots.
However, unlike the internet, AI is still based on information and content created by others, most of whom remain unpaid. This imbalance, I fear, will continue until all AI and model builders find a way to pay everyone for their help in building this AI revolution.
