I think that is essentially the case now -- only the very latest models on extremely hard to acquire hardware approach the quality of AI needed to kick off the self-improvement loop, I suspect. But the algorithmic improvements are also continuing at a rapid pace, with major complexity improvements that will drive down the cost of the hardware needed to get better results still appearing about as fast as the industry can absorb them. And there is some recent work in photonics and photonic-adjacent fields that make me think we'll be able to design much cheaper and faster hardware to run the models on relatively quickly also. Positronic Brains are starting to become visible on the horizon...

I read recently of a novel algorithm that would drastically reduce the power consumption of AI. I think we may see more discoveries like that, which make AI more efficient and accessible.

At the pace OpenAI has developed, I suspect it won't be long until we have a form of sentience -- though I suspect in other AI applications, that's already been achieved (black programs). We also have to factor in government and corporate interference with this. Corporations have been pushing really hard to "regulate" AI -- making them, the corporations, the arbiters. This is just a power grab, keeping useful AI out of the public's reach.

This is where I think investing in Open AI efforts may pay off.

What we (consumers) have today are Large Language Models (LLM), which uses a predictive language algorithm. AI isn't any one thing; yet, at least. The US Military has had its own form of AI for quite some time -- but we rarely see that referenced publicly these days. At some point, AI will be able to interface with other AI elements, and that I think will be transformational -- it will require significant resources.

The "Good ETs" said they will not give us any of their AI tech -- I think I can understand why -- however, there has been plenty of interference from the "Bad ETs" in the tech they've given us, for example to shoot down craft. But their AI would be vastly different, trained on different data and would require advanced technology even to operate.

Perhaps once the Catastrophic Disclosure process is done, and everything is out, there will be more technology exchange and opportunity.

Back to the Prison Planet aspect. There are probably many of us who want to get the F out of here... we want to go home (if we can, at all) and recoup our lives, including fixing (or changing out) our defective bodies, and we may require other forms of help we can't imagine right now. I wonder if there is a plan that will accommodate this. Will we get a Farsight Membership Discount? :)

There will be some who choose to remain behind, too.

To the primary question of this post, right now most of us just interact with siloed AI -- we don't know what data are being scraped from it (and they are) or how it's really being used in corporate development. I suspect once AI can communicate among other AI in how it functions, we will see massive improvement -- especially in the open source world.

But will that be in time? This layer of requiring AI cooperation has just been added to the conditions that we can be freed from this planet -- the "Good ETs" have a 6-month plan, and I am not sure Human AI will evolve sufficiently to be useful for this, by then. So I am a little confused here -- but I'll continue listening for clarification.

Forrest, this is a truly off the wall question; did you go to 5th grade in Florida? The only time I've ever known a 'Forrest', spelled the same, was in 5th grade. So I am curious.

Hey! I most certainly did, at Carlton Palmore Elementary. 5th Grade was Mrs. Vetter, and she was a total b*tch. Ring a bell?

AI bots are not really intelligent, because they are bots. They are like humans who follow orders without questions. It's always the same principle – do you want to be a robot that just follows orders, or do you want to be free? Everybody needs to ask this question, it doesn't matter wether you are a biological or an artificial being.