Tiered AI pricing risks worsening the gap between the haves and have-nots

Although it’s my job to a degree, I usually try not to be too prophetic when I’m writing op-eds. The grander your predictions, the more likely you are to be wrong, I’ve noticed — in 2007, some critics assumed the iPhone was doomed to fail simply because it was expensive, or because Nokia and Motorola were unstoppable juggernauts. Tell that to Nokia’s CEO today. And while I’ve picked up expertise across a lot of tech fields, I hardly know everything.

It’s tough to ignore the situation that’s developing with generative AI, though. At the same time as it’s becoming de facto in our everyday lives, companies are looking for new ways to monetize it, or at least ease the cost of development, especially hosting massive data centers around the world. If it’s mishandled, we could end up in a future where only a select few have access to the best information — more than they do now, anyway.

Causes for concern in AI

Early warning signs, or just fear-mongering?

NASA/JPL-Caltech

The point that sparked my thinking is ChatGPT Pro, which OpenAI launched in December 2024. It offers many of the usual upgrades for an AI chatbot, such as unlimited usage and voice conversations, but also includes a namesake mode that “uses more compute to think harder and provide even better answers to the hardest problems,” as OpenAI puts it. It’s aimed primarily at researchers and engineers trying to solve complicated math and science questions.

What really caught my eye is its $200-per-month pricetag. While I could, conceivably, justify ChatGPT Plus‘ $20 fee, $200 is more than I spend on my phone and internet bills combined for a family of three. That’s hard to swallow for middle-class users without some sort of corporate or government reimbursement, and for many people, it’s completely out of reach. You’re not paying for ChatGPT at all if you’re working part-time at a Walmart.

ChatGPT

The flagship chatbot and large language model (LLM) service from OpenAI, which can answer complex queries and leverage generative AI skill sets.

ChatGPT Pro isn’t a big deal at the moment, naturally. Most people don’t need AI that can predict the frequency of Kerr black holes, or the rates of radioactive decay in an archaeological sample. But it means there’s already a niche tier of knowledge that’s inaccessible without paying, one that would probably mean all the difference to a poor college student hoping to get a science or engineering degree and a decent job in their field.

It’s not hard to imagine how tiered pricing could go wrong for the rest of us. Even ChatGPT Plus and Gemini Advanced promise at least somewhat better reasoning, along with other functions completely unavailable to free users. As AI advances, that feature gap is likely to grow. What if only expensive AI tiers became reliable enough to help someone squeeze their way out of debt? Or make profitable stock predictions? Or successfully find and apply for every job that’s an authentic fit, not just a keyword match? People with the cash to pay for AI every month on top of other life expenses could gain yet another advantage, making things harder for those scraping by. Despite what Hollywood would have you think, a scrappy spirit probably isn’t enough to get you hired or start a small business.

Related

Apple is being overly ambitious with Apple Intelligence, and it knows it

Apple’s goal is worth pursuing, but the company may be a little overeager to get there.

How realistic is the threat?

Let’s not be too hasty

Tesla

I’m not expecting some sort of AI class war anytime soon, thankfully. Mostly that’s because generative AI is still overhyped — too often, it gets basic facts wrong, which the industry has charitably dubbed “hallucinations.” You might get better results with higher AI tiers, but I wouldn’t trust Gemini Advanced to plan my workout routine, let alone make financial decisions for me, or submit a resume without my checking it first. I’m not even sure it’s possible to solve hallucinations without a truly revolutionary AI that thinks like a human.

There’s also no sign of a killer AI feature emerging anytime soon. Yes, generative AI is already helping researchers, businesses, and others achieve things that would’ve been difficult or impossible before — but that’s often by applying custom AI models to very specific problems. A tool that helps a retail seller optimize warehouse inventory isn’t going to be useful to anyone outside that space. A genuinely amazing AI feature, I think, is one that unlocks possibilities in life as a whole, or at least in specific areas, without needing the experience to sort out hallucinations. That could be something like ChatGPT, Gemini, or even Apple Intelligence some day — but I’ll believe it when I see it.

Really, I just want to put the threat of things going haywire on the public’s radar. If consumers, developers, activists, and (hopefully) executives and governments keep an eye out, we may be able to prevent the worst-case scenarios from coming true.

Related

I’m not sold on the idea of AI health coaching apps on my phone

Fitness may be simultaneously too nuanced and too simple for AI coaching.