That’s a really good counter argument. You may have moved me over to the other side.

Yea, I agree with this, but I still don’t see what advantage the state of the art providers receive by adhering to an open protocol. If anything doing so would (on the margin) push users towards open source models when they might have been willing to pay for a more advanced model just to access certain plugins.

That being said, I do think that a standardized approach to a plugin ecosystem will arise. I just think it’s silly to expect any of the foundation model providers to participate.

I don’t think so, I suspect my argument holds no matter who is running the most advanced LLM. The market leader will never have an incentive to open source their “app store”.

The only way this breaks down is if by some miracle, an open source model takes and maintains the lead.

Yea, it would be nice.

But what benefit does any LLM provider gain by implementing/adhering to an open protocol? OpenAI is trying to build a moat around their service, from their perspective plugins are key to establishing a competitive advantage.

I can’t see this happening in reality.

The question isn’t just about whether or not you could get a job in 3-4 years, it’s about whether or not the investment makes sense. Unless, you plan to be employed in a field for >7ish years, then the answer is almost certainly no.

Are you confident you can identify a field that will still require your labor in 10-11 years?

I’m pretty sure we’re already there.

Assuming that all technologies exist and there is nothing further for humans to reach for. I could see myself wanting to enter a fully immersive simulation where I could experience life when humans still had a purpose. I’d want that simulation to be so immersive that I had no memory of my outer reality.

Heck I might even choose the time period immediately prior to the technological singularity as the setting for my simulation since that is probably the most interesting time for a human to be alive.