So for defending high value fixed installations? Sure. But the problem is that there are more things that can be hit than you have systems to deploy to defend them?
So for defending high value fixed installations? Sure. But the problem is that there are more things that can be hit than you have systems to deploy to defend them?
Yeah I think there are viable options here but the complicating factor is the width of the line. The advantage of interceptors is that they can defend a fairly large area with a single battery. CIWS or similar have drastically lower range.
On the scale of "prepare to cry at the pump" to "buy an EV today" where are we at?
There are ways to limit this (showing only a small portion of the chatlog for example, like Monty Python's the Funniest Joke in the World), but those also have downsides (will miss because it didn't see enough of the conversation).
This is generally done but is far from foolproof, because the chatlog/context that caused the first LLM to go nuts may also cause the second LLM to go nuts. The filter LLM has heavily overlapping failure modes as the main LLM.
This is *so much liability*. I can't believe any lawyer looked at this feature before launch.
Oh yeah that'd be appreciated!
I'm kind of curious as to what one can do with this data. Are you open to publishing the non-Bsky flagged event feed? I have some weekend vibecoding ideas...
What's the heuristic for the flagging signal? I'm wondering what's the difference between "insider trade" and "extremely misguided conviction from one dude"... or maybe there's really no way to tell the two apart.
Oh for sure. Astroturf Chongqing content is infinitely preferable to astroturf Dubai content. That whole place is a slave state nightmare for new money assholes to openly declare their animus to any kind of morality.
*So much* lifestyle content about the Gulf States generally. Oddly enough now I see a *lot* of lifestyle content about Chongqing specifically. All feels very astroturfy.
He’s now tied his credibility to the US military’s credibility and boy do I advise against taking a long position on that one.
Exactly. There will be more atrocities and indefensible incompetence because look who is doing this! All of it is now on Carney because what… he wanted to score points in trade negotiations that nobody in the US will remember in 24 hours?
Yeah this is a total fumble. There’s a way to express condemnation of the regime without throwing in behind the US.
And this will get *even more unpopular* as time goes on because of who’s involved.
If you have a magically more productive workforce you rationally will use the extra capacity to go conquer new markets. A reaction in the opposite direction is an admission that they don’t think they can.
Yeah the main signal I read off this news is “our main products are tapped out and we’ve proven unable to tap into new products/markets in a meaningful way and are giving up”
It's really striking when visiting Asia how much more capitalist/competitive things often are... and how it actually does improve everyday life.
At the risk of no-true-Scotsmanning, the US is spectacularly bad at capitalism in many sectors.
In isolation I’d agree but the degree to which their competitors (OpenAI also) are in bed with the administration definitely gives me pause…
It is not coincidental that the biggest proponents of the modern notion of Singularity are *also* the biggest funders of immortality research!
A machine that out-compounds us is very consequential over long timespans! It probably won't compound enough within your {expected_lifespan} to save you from your meatsack!
But the Singularity people aren't talking about that because undergirding all of this faffing about is the foundational reason any of them think about this stuff: as a means to escape death.
Also, to add because this is a big bugbear of mine: we may very well invent a machine that is better at compounding than our society! It could become a gargantua over centuries or millennia! That's cool shit!
So sure, a hypothetical system where {rate_of_compounding} is high, sustained (or even increasing!) is very scary... but at that point we're far from the original definition and *quite* deep in the modern Silicon Valley one.
Also even in the a) assumption there's a baked in assumption that {rate_of_compounding} is high and doesn't decrease over time.
In almost all natural systems it's a sigmoid. Compounding rate slows over time! At *best* it is constant.
The only version of this idea that becomes a 🚨 five-alarm fire is the social one that the worst people in Silicon Valley worship, because it fundamentally refuses to model any significant growth constraint.
I think there's an interesting thought exercise about what happens if we invent a machine that can do this at a radically higher slope than our messy human fleshy machine, but also one can easily argue that the digital version of the machine faces the same scaling problems.
We shot people to the moon! We fly around like birds! We *have* a machine for reproducing more capable copies of itself. If we include groups and societies of people in the original definition we've already met it!
But we do this all day long. Each of us make stuff daily that would be inscrutably science-fiction to people from even 40 years ago.
A *person* has not become superhuman but large assemblages of humans are a machine for self-improvement. It's the basis of our civilization.
a bunch of cloistered worms decided to make my fucking friends the national punching bag and I will never, ever forgive them
Now that I think of it, it really is like sports gambling. A somewhat contained problem when you had to physically drag your ass to a bookie/hobby store.
And a much, much bigger problem when you do it without getting out of bed.