Words to live by
scratchee
joined 2 years ago
Whilst I’ve avoided LLMs mostly so far, seems like that should actually work a bit. LLMs are imitating us, and if you warn a human to be extra careful they will try to be more careful (usually), so an llm should have internalised that behaviour. That doesn’t mean they’ll be much more accurate though. Maybe they’d be less likely to output humanlike mistakes on purpose? Wouldn’t help much with llm-like mistakes that they’re making all on their own though.
The big sci fi win in stargate is how highly they rated internal consistency and having a scientific basis where possible. Apparently that was mostly because the actress playing Carter absolutely refused to tell bullshit gobbledegook and forced the writers to do it properly.
It’s subtle, and not always perfectly followed, but if you take the episode where they gate to the black hole, they have significant screen time justifying why the time dilation is so strong when the gravitational effects are so weak. It shouldn’t work that way and they acknowledge that explicitly, but obviously they wanted a fun time dilation story so they call it out and explain it as an unexpected side effect of the gate wormhole. So sure, they sometimes make science do what they need it to do for the story, but they try hard to justify it.
Star Trek meanwhile barely follows its own rules most of the time, let alone actually acknowledging real physics
Plus they cleverly evolved ahead of time to be camouflaged against all our plastic pollution, so their few predators keep choking on plastic bags instesd
Hey now, we have an unfair disadvantage, all the nutters are lumped in with us and bringing the side down.
I’m disturbed by the generic anti-war reference, it will depend how they clarify their position going forward. I’m ok with distancing from the ongoing Israel mess ofc, but I don’t see a low-military solution to Russia, at least any time soon. and I’m not sure I can vote for anyone advocating a reduction in military capability whilst Russia is building towards being a threat to Europe and therefore us.
Don’t get me wrong, I want a peaceful solution. I just can’t see one that doesn’t involve having a big stick to disincentivise Russia. I’d rather throw more money on a military that does nothing than hope that nonmilitary encouragements will work (ofc happy to have those too, but they didn’t work on the Nazis, and we don’t want another Neville Chamberlain-style appeasement scenario).
Of course, detailed opinions on complex issues weren’t going to make it into a statement this short, so I’ll wait and see, but I’ll be disappointed if I find I can’t support them just because they’re not taking Russian aggression seriously.
Yeah, that’s fair
The company did many things wrong, it’s an almost idealised example of total failure to take software seriously.
Most importantly they decided they didn’t need to test the software on their new machines because they’d already shipped previous machines running the software, so they “knew it worked”. The previous machines had hardware interlocks that made it impossible for the software to cause a massive dosing errors, the new machine was entirely software controlled.
Also they had exactly 1 “very smart” engineer build the software, who obviously wrote it for a hardware-safe machine. To be fair, I’m sure he was very smart, but safety critical and solo projects are not a great combo.
Also they had no mechanisms to ensure failures would be communicated to their engineer~~s~~ for investigation (failures were reported to them and then dropped into a black hole and forgotten about).
Also they didn’t even have any capability to test their machines after failures started popping up, because they knew the code worked perfectly so they didn’t need to waste any time or money on qa capability, massively slowing down their ability to fix things once people started dying
Software patents are pretty close to universally bad. Software moves fast and twenty years is ridiculous, when video codecs have grown to be biggest format and then been overtaken by their successors which in turn are overtaken by their own successors before the first codecs lose their patent then you know something is going wrong. Hardware patents have their place as you say, but software moves very quickly and can innovate just fine without the need for patents.
In theory you could make them viable by shortening the life, to just 5 years or something, but at that point the cost of administering them probably outweighs any benefits (if there would actually be any).
Copyright is another matter, I think we probably need that in some form (though the stupid length of copyright at the moment is even stupider for software)
If near infrared (1000nm) can become uv with the wrong material, surely visible light from the sun can do the same and would become an even more dangerous wavelength? Or is this an effect that only happens to near-infrared? Ive not come across it before…