Someone asked me yesterday, what sweeping changes I anticipated in my field that would have an impact on consumers in the next ten years. They said so with no traces of sarcasm nor mockery. Setting aside how preposterous the question is, I thought I’d sort of answer it with some correspondingly utterly useless, baseless predictions. I’m not an expert in any field–but I have a website so I practically am one. (to be clear, that is sarcasm denouncing denigrating statements of many self-proclaimed internet “experts.”)

The Death of the Application

Browser based everything will be the norm. WASM/Web GPU will be the default target for “applications”, allowing for easy cross-platform development. Just write something for Chrome and you’re done. I say Chrome because I’m not hopeful for a Firefox future. I am, and will probably continue to use, Firefox (or rather, LibreWolf), but I will not expect that most people will stick to this for long if WASM proper takes off and there start to be major differences between how content is served via Chromium contra Gecko.

Oh yeah, and more SOCs. Apple’s M chips have shown how powerful SOCs can be–I think this could be the default for Windows laptops too.

Always online everything will be the norm.

Edge Models

Models will be fine tuned on edge devices, a resurgence of voice assistants of yore–but somewhat smart now. Or more than that, models are almost entirely run on edge devices as raw power fails to bring adequate improvements and we turn our interests to efficiency.

The Death of the Transformer and the LLM

I think this is the safest bet. In ten years, something else will be the bleeding edge in AI. With this, I’m somewhat bearish on Nvidia. They’ll have the capital and talent to adapt to the market, but if they don’t they may be toast when the current GPU architecture suddenly isn’t the fastest way to train models. Some bleeding edge ASIC might cut them out. I know it’s crazy to be bearish on Nvidia, but again I’m making baseless predictions and am looking a decade out, so might as well go wild.

Resurgence of Statistics

This is almost certainly because I’m a statistician, but I think we will see a return to “statistics” as opposed to “data science”–i.e. somewhat of a transition back to concerns about estimation error instead of solely a ruthless demand for accuracy.

Adoption and Rejection of Day-to-Day Model Usage

I think that model usage (be it an LLM or whatever may supersede them) will be split by application instead of person. Personally, I’ve been avoiding LLMs for most things as they write worse than I do, and the effort taken to rectify their mistakes is often more than just writing from scratch. Of late though, I’ve been trying to use LLMs more. I’ve settled on DeepSeek as my go-to, and with R1 I’m using LLMs far more than I would’ve even a few months ago. I think as time passes, this will become the norm–with the question only being which tasks you use a model for, as opposed to whether you use one or not. I don’t think this is a bold prediction, and is certainly how people are acting now. I think this will be default in a decade though, probably with norms established in industries or positions or such. Experts would reject models in their areas, and accept them in others so as to boost productivity on a whole.

Increased Focus On Accuracy

Models need to be more accurate to be useful. Ternary truths are key. I want to write about this soon, but I’m thinking of nullable bools as a means to express uncertainty for decisions–things aren’t always yes or no, sometimes you don’t know, and that should be made explicit.

Dead Internet

Botting will be a major problem–more than that, it will be the defacto standard. I’m going full Dead Internet Theory.

Education

I have no idea how it will look, but it will have to change. Perhaps there will be an increased focus on critical thinking, synthesis, & intelligence as opposed to rote memorization? But I’m sure such a sentiment was made alongside the advent and widespread adoption of the calculator & the internet. We shall see.

Alea iacta est.