I’ve been peering into the LLM world this week, and honestly, I think I’ve just accepted that AI is inevitable at this point.

The amount of collaboration, research, development, and just overall movement happening around AI is kind of insane. Things are being created way too fast to pretend this is some passing trend. It feels like every week there’s another model, another tool, another company integrating AI into something.

Recently, I learned about the whole backlash around Google pushing LLM onto people’s devices, and I get why people were upset. Nobody wants an LLM forcibly installed on their phone, especially if it’s using their GPU, battery, hardware, or whatever else for something they never asked for.

But I also think there are two sides to it.

To me, there’s a huge difference between having an LLM on your device and having some cloud-based AI service constantly connected to your device.

Like, if there’s an LLM running locally on my phone, I don’t automatically see that as a problem. The issue is whether it’s sending my information somewhere, tracking what I’m doing, or quietly reporting my activity back to some company. That’s where it becomes creepy.

An LLM by itself doesn’t really “report” anything unless it’s given tools, permissions, or network access to do that. So I don’t think the problem is simply “AI is on my device.” The problem is when AI is forced on you, hidden from you, or transmitting things without you clearly agreeing to it.

Honestly, I’m not completely against having an LLM on my phone. I’m actually okay with it, as long as it’s local, transparent, optional, and not siphoning information from my device.

And there are some real benefits to that.

For example, imagine being somewhere with no signal. Maybe you’re in an accident, or you’re dealing with an injury, or you’re stuck somewhere and need help figuring something out. Having an offline LLM on your phone could genuinely make a difference. It’s not going to be perfect, obviously, but it could still be better than having absolutely no tool or knowledge available in that moment.

That’s the part I think people might be overlooking.

A local LLM could be useful for offline situations, emergencies, travel, accessibility, translation, troubleshooting, or just having a basic assistant that doesn’t need to connect to some server every time you ask it something.

So I guess my point is: I don’t think “LLM on device” is automatically bad.

What’s bad is forced AI. Hidden AI. AI that phones home. AI that pretends to be local but is actually sending everything to a server. AI that gives users no real choice.

There’s a big difference between a private tool running on your hardware and an inference service quietly connected to a company’s cloud.

And I think that distinction is going to matter more and more as AI gets shoved into everything.