Pluto will rise again

I think the best term for AI luddites such as yourself is "AI Vegans".

Like vegans, you know a tiny sliver of reality, and make it your whole world. Like vegans, we don't need to know you're like this, because you'll tell us. Like vegans, you're wrong on almost everything. Like vegans, you believe your faith makes you a higher form of being, and those who disagree just "are lesser" i.e. "do not understand math".

Like vegans, you'll mostly go away in a decade or two.

It's actually the opposite. It was stagnant between 1960s and early 2000s. Where we thought we'd need to program it, because we were modelling it on ourselves even back then, and back then we thought brain was biologically fully pre-programmed.

Then fMRI brain imaging came, we understood how brain functions, that's it's not pre-programmed but learning from basic LLM-like template and then inferring based on what it learned (remember the "90% of the brain is not used" pre-fMRI myth?), and within a decade, basics of modern AI being "learn and infer" principle became mainstream.

And then they look at alternatives, and conclude "yes, this is in fact by far the best system we've ever had for this purpose".

"It's not perfect" is utopian thinking. What matters is "is this the best system we had so far".

Openrouter doesn't even have the Noromaid model, so probably not a lot to worry about

Jebus shut the fuck up already. NOBODY CARES!

Amazon's total retail revenue is 1%~2% a year. They literally make nearly nothing off retail. The majority of their business is AWS (and most Americans outside tech aren't even aware Amazon is one of the three largest data center companies in the world next to Microsoft Azure and Google's GCP). They don't actually need to make ANY money. They just need to not lose too much. (Amazon's retail revenue fell to -2% the year they invested in Rivian out of that division).

Curious opinion considering that legal profession is mostly applied about logic. You learn the rules, and they you use logic to apply them to each situation.

Of course we do. We literally modeled current gen AI on it.

In fact, we even know that LLMs overwhelmingly struggle learning where humans struggle as well. Because it's the same process.

The only difference is that we are embodied and spend decades learning, while they have far more compute and memory and are able to learn in days. Each grants us slightly different advantages and disadvantages within the learning process.

Which is fundamentally the same process.

It is literally learning though. It's a neural network modeled on human brain, learning in the same way human brain does.

You going "lalalala, can't hear you" isn't changing reality. It's still learning.

In EU, most grid operators learned from the mistakes of "let's just let everyone feed grid from unexpected input points all at once, and then stop all at once".

So the offset for feed in tariff is so high that it's basically never profitable to do so.

Correct, that's why I didn't mention it at all.

All I spoke about is that "you're not taking electricity from the grid when electricity from the grid is basically free, and you're taking electricity from the grid when when electricity is hilariously expensive".

It's telling that I haven't talked about that at all, and yet both naysayers claim I did, because that's your narrative.

If you're in Europe like me, when it's sunny and windy the price is going to be near zero, and when it's not it's going to spike to mid to high two digits cents per kilowatt. Had some cases of going to three digits in recent years. That really sucked.

I.e. you're "harvesting" when it's near zero cost electricity, and having to pay when it's expensive. The only thing that could give you any "return on investment" is the legally mandated 24 hour pricing shift which if memory serves me right is EU mandated, which is specifically designed to artificially inflate price of electricity renewables produce to give them a chance of paying for themselves at all.

Both are true at once. Preference mathching is best, but preference shifting can work if you have a correct template for that specific human. Then you can determine what preference shifting has worked on this type of human in the past, and deliver that kind of material in that kind of context to get the process along.

Notably, same thing is required. Deep understanding of one's thoughts. Preferably before one thinks them.