I’ve been working in tech since before ChatGPT launched, and I was a journalist covering flying cars and other tech breakthroughs before that. I know a thing or two about tech hype cycles, how to look past them, and how to see around the corner to find what’s worth fearing – and what’s not. I’m not right one hundred percent of the time – no one is. But I am right often enough to feel comfortable being more annoyed than afraid with the current state of affairs.
I’m not worried about the AI bubble and AI doomerism more broadly for these reasons:
- Technology waves, like the market, rise and fall, leaving behind valuable infrastructure for people to use to create new products and services. From telephone lines to fiber optic cables to data centers – we’re always building new capacity. Believe it or not, the most successful company to leverage emerging capacity is likely still in its infancy despite the huge IPOs over the last few months.
- We’re going to transition from technological performance to service. We’re already seeing this with use cases like Muse. The average user doesn’t care about whether a model meets or beats a benchmark; they care about what the model can do to make their life better (however they define “better”). In enterprise, businesses want to control for risk, which makes models trained on unlicensed data somewhere between uncomfortable and a complete nonstarter. Hello, opportunity!
- The political winds will continue to shift, and regulation will shift with them. People care about not having their work stolen; they care about not having deepfakes made of them. They care about making sure data centers are responsibly built. They are sick of slop. The election results show that politicians ignore these concerns at their peril – that risk doesn’t disappear with one election. Delivering on promises to voters is difficult, and political winds shift when those promises are not met – and sometimes they shift even when they are.
- There are a lot of different types of AI. There’s AI that’s well-suited to government use, and AI for commercial use, and even AI for personal use. Some companies require AI that’s trained domestically. Others can work with models trained abroad. Companies are increasingly looking for models that meet their needs over models that meet every need to save both money and time. That means there’s still room for a new wave of different approaches to AI.
At Sublimate, we’re designing for this next wave. We’re designing for what comes after people get bored with watching large, multi-modal models crush benchmarks and climb the asymptote to “AGI”.
We’re preparing for when individuals and enterprises recognize that the power of AI isn’t in developing The One Ring – one large model to rule them all. It’s in building the capacity to give people the AI models they need when they need them – all while providing the customer with the confidence that the data the model is trained on has been authorized and licensed, and they have the rights to use the outputs in the ways they want to.
Underneath all of the hype, the drama, and the fluctuating markets is a simple truth: we all collectively have a say in what happens next with AI every day. The elections are one piece of evidence and now the markets will serve as another. We also have an opportunity to make our voices heard beyond the ballot box and Wall Street. For example, you can choose the models with which you share your data.
An AI apocalypse is not a foregone conclusion, and a market bubble inflating and bursting is not the end of an entire industry or the global economy as we know it. Last I checked, internet-based companies are up and running despite a wave of outages and, in the case of some services like Microsoft Office, their restoration is being met with reluctant applause.
I don’t mean to make light of the pace of change or the risks involved in the development of AI and the technologies and infrastructure that enable it. I also don’t mean to minimize the losses for real people in the market downturn. I am one of them. However, I think it’s important to provide a pragmatic message as an alternative to the hyper-optimism and nihilistic doomerism that has been roiling at the extremes.
I’m not afraid of the AI bubble because, at Sublimate, we have an alternative vision for how AI can work for artists, creatives, and everyday people who want to control their intellectual property and the right to generate AI outputs against it. We see how value can work through AI in unique ways long after current market volatility has subsided.
.png)
.png)
.png)