An example with an oversimplified diet to illustrate the point I think you’re trying to make: You have a diet that’s exclusively cake and you’ve determined that you need 2000 Calories of cake to maintain your weight. That 2000 Calories figure is an estimate and we don’t know exactly how much of it we’re actually absorbing. In reality, it’s actually more like 1800 Calories. Now all of a sudden, you switch your diet to eating exclusively cookies. You measure out exactly 2000 Calories of cookies and eat the same thing every day. But your Calorie estimate is wrong and you’re actually eating 2100 Calories of cookies per day. Now you gain weight on this supposed 2000 Calorie diet.

I argue that this doesn’t matter either. If you see that you’re gaining weight, then it means you’re eating too much. Reduce your Calorie target and you’ll be back on track. In a real world scenario, you’re going to have a much more varied diet than only cake or only cookies, and each item will come with their own measurement errors. But for most people, their diets are varied in a fairly consistent way, so these errors are also consistent on average. If you ever make changes in your diet (e.g. completely cut out McDonald’s), you’ll change both your estimated Calorie intake and target like in the example above. Adjust your numbers accordingly based on how your bodyweight moves and you’re good.

Of course, other ways of dieting are also effective. It depends mostly on what you can adhere to and your goals.

I know someone with relatives in China where they gave their kids nicknames that roughly translate to “first baby”, “second baby”, etc. They’re all middle aged adults now and they’re still addressed by the same nicknames. So you have kids listening to their grandparents talk about “second baby” and imagining a baby, but then you meet them and it’s an old man.

None of that actually matters when it comes to weight control. What matters is that the linear relationship is retained in your proxy measure of Calories. Meaning that if you eat two pieces of cake, you’ve doubled your Calorie intake compared to eating one piece.

Many pay a lot more without ever learning this lesson. I’d say it was money well spent.

Something like “Fediverse Obituaries”?

Privacy concern for sure, but given that you can already tie different photos back to the same phone from lens artifacts, I don’t think this is going to make things much worse than they already are.

someone could create a virtual camera that would sign images

Anyone who produces cameras can publish a list of valid keys associated with their camera. If you trust the manufacturer, then you also trust their keys. If there’s no trusted source for the keys, then you don’t trust the signature.

I think there’s enough people who care about this that you can just provide the data and wait for someone to do the rest.

Couldn’t you achieve the same thing using a ceramic bowl with a lid/plate?

An air fryer is a convection oven. But not every convection oven is an air fryer.

If you’re able to give precise instructions on what to do to resolve the issue, then couldn’t you also automate it and make it seamless from the user’s perspective?

with so few input

Yes, few inputs. Not none.

I’d call math/logic/reason fairly strong by themselves.

What does strong mean in this context? It’s a very useful tool. No one is denying that. It just doesn’t tell us anything about the universe without input from that same universe.

We’ve had wireless phone chargers for a long time.

That’s two items. The second one costs extra.

Bamboo rats go through three stages when eating bamboo. The initial phase is the quick motion of the bamboo entering the rat’s mouth. The intermediate phase is the movement of the teeth chewing the bamboo and the final stage is the state of the rats’ teeth right after eating the bamboo.

Um, what?

Mathematics is all about developing logical tools. Basically things like “if we start with this assumption, then you can make this conclusion”. After you’ve developed all of these tools, then you can look at the universe around you and apply those tools to your observations in order to come to new conclusions about that same universe. There necessarily needs to be that input that ties it back to reality. Mathematics on its own doesn’t tell us anything about reality.

I often wonder if it’s worth being better rested at the expense of losing so many wakeful hours.

I’ve been sleeping at 9pm and waking up at 8am fairly consistently for several months now. Still waiting on that elusive 6am wake up time.

I don’t know if it would be correct to say that the civilians have nothing to do with the war. As far as I’m aware, the IDF’s intended target is all Palestinians, civilians included.

It may be worth editing your post to specify that you’re talking about LLMs. There’s no indication that this is your intent without reading your responses in the comment section.

Ethical meaning : “private”, "anonymous, “not training with your data”, “no censured”, “open source”…

Private, anonymous, and uncensored. Those are easy. There’s plenty of pre-trained LLMs out there that you can download and use however you like.

Not training with your data, not possible as far as I’m aware. LLMs rely on the availability of a huge quantity and diversity of data. There isn’t enough of that available that also come with consent of the creator for this usage.

Open source isn’t well defined for machine learning models. Lots of models have their code and weights available, so if that qualifies for you, that’s also easy to find. Huggingface hosts most of them.

To accomplish that, you would need power. In this capitalist society, power comes from money.