A detector can be used to fine tune the generation until it passes, so you're going to end up with a lot of false positives (never mind how fake a lot of real photos look due to filters, which makes the task more difficult).

They have pretty well stated they can't scrape English data, as it has too much Western bias for their liking. They may be able to filter it, but as we've seen with ChatGPT, it's not straightforward, and things will fall through the cracks. That makes life difficult for censors.

In domains where they have access to large volumes of data that doesn't need heavy curating (outside of text), they should be able to do fine.

.. and after dumping on the US it still won't tell them 🤣🤣

What kind of data do you mean? I don't believe they have a high quantity of quality domestic text training data, and they have stated they don't want to use worldwide data. It's not clear how they plan to resolve this.

Being aligned to humans may help, but a human aligned AGI is hardly 'safe'. We can't imagine what it means to be aligned, given we can't reach mutual consensus between ourselves. If we can't define the problem, how can we hope to engineer a solution for it? Solutions driven by early AGI may be our best hope for favorable outcomes for later more advanced AGI.

If you gave a toddler the power to 'align' all adults to its desires, plus the authority to overrule any decision, would you expect a favorable outcome?

Which is what makes it weird, the core investment thesis was a miss, coming in below expectations for DC revenue.

That we disagree illustrates the problem, it's not unusual for there to be fundamentally different ways of seeing the world. It is a fact that the message an author is attempting to deliver, may be missed entirely by some people - and that's not necessarily a failing of the author, or the reader. A chatbot should in principle be able to pick up on this nuance pretty well, given sufficient data. It would need training data feedback from the reader though, which in many cases won't exist initially.

By and large humans aren't great at understanding other humans. Understanding a collective of humans (even superficially) is probably one area an AI trained with enough data will truly excel at. Making it a dangerous tool for spreading propaganda, which could be countered by AI readers/filters.

It's simply too much information for any one human to take account for (to model millions of readers), over time I would expect a new category of book to emerge which has minor variations that are tailored to the reader.

I never said it 'knows' or displays true intelligence, only that it performs at a level far above earlier chatbots that didn't come close to this capability.

Given me one example of an earlier chatbot that could code in multiple languages.

They are after wage growth, and wage growth has been trending down. Unemployment serves as a proxy for wage growth, but if wages cease growing there's no problem with full employment.

Your uploaded self will get terrible anxiety when moving between networks. I wonder if there will be uploads that refuse to move from the substrate they were uploaded to..

Salary data is available online, and it's nowhere near $1m. Maybe a small subset of critical staff, but even then not necessarily.

Organisms aren't closed systems, entropy as some inescapable feature doesn't apply

A common theme in responses on storytelling, have the ending 'and everyone got along harmoniously' or variants thereof. Ask it to tell a few stories and you should see it soon enough. There's almost certainly some engineering going on pushing it in that direction, which isn't really ideal for comedy.

Can you cite any legal challenges made by Uighurs in China? This is one of the core issues, when someone is mistreated there appears to be no recourse, or even the facade of recourse. Lawyers have faced punishment for even challenging the state.

It means there are countries where the majority of people are lean, and we don't need to much wonder what will happen.

That's almost three years ago now..

There is a big mother fucking difference between 'almost over' and 'will never end'. I never said it's almost over.

I didn't say Ukraine would win did I? I said it's a costly war and a resolution is likely to be reached sooner rather than later. Ukraine also doesn't have unlimited support, but it's looking like they may manage better than Russia.

Their budget surplus was massive in March, nobody could credibly say they couldn't fund the war at that time. That surplus has shrunk significantly since then, fundamentally different set of circumstances.

You cited Korean war as an example - how much does the Korean war cost their respective governments each year? Stuff all compared to what Russia is spending at the moment.

Pretty much, as they keep receiving external support to fix power infrastructure, as well as ammunition - while Russia continues deplete cruise missiles and other critical military hardware that they cannot easily replace while sanctions are in place, and loses revenue with Europe moving away from Russian gas altogether.

Which would be relevant if the OP had said '2 years ago' instead of 'than a year ago'

Don't need to take my word for it, data on revenue is straight from their ministry. https://www.wsj.com/articles/russias-shrinking-budget-surplus-limits-putins-war-options-11668174524

"Data released by the Russian Ministry of Finance on Friday showed that in the year to October, the government budget surplus stood at 128 billion rubles, or around $2.1 billion, down from a surplus of 2.3 trillion rubles in the same period last year."

Diesel is up 50%, crude almost flat on the yearly. That's a long way from 2-3 times higher.