I'm never sure if I should be upset or happy when I've been debugging a problem for long enough that I finally decide to switch something out in order to improve visibility and that immediately solves the problem for entirely unexpected reasons. Particularly all the times when I couldn't readily discern why.
I mean yes, but that's far more difficult than one would expect in a layered system. You may be able to keep a concept aligned, but can you keep the meta concept aligned? Continuous learning means there is continous opportunity for a more powerful system of misalignment to form itself and take control of your alignment. Even worse is hidden layers of this misalignment that can hide itself by not using an interpretable language.
> Seems like AI could help massively there, by removing a huge bottleneck around technical elaboration and application seeking.
Ah yes, lets take the existing system that's already barely holding up and flood it with slop. What could go wrong?
> Since we lack the resources to fund every PhD with a crazy theory on what the next new subfield ought to be, how do we decide?
Do away with the perpetual uncertainty of the "will we or won't we continue to fund you" grant treadmill for practicing academics. Hold a yearly competition of academic prowess and intelligence open to any adult US citizen under retirement age. Award the winners a grant good for 40 years. If we allocated slots equal to 0.005% of the population each year (ie ~17k) that would represent 0.2% (ie 1 per 500) of the population in total at any given time.
Obviously I say that (mostly) in jest but clearly there are workable solutions if we approach things from a new angle instead of determinedly clinging to the status quo.
1. The AI chooses, of its own volition, to kill people. It is likely that by the time AI has this level of control and intelligence, it is too late to stop it.
2. A malicious person uses AI to cause a terrorist event or some other kind of catastrophe. This is more likely. The AI in this scenario is more like a tool. Attributing AI to the cause might be difficult, but it's likely that any new bioweapons which emerge in the next 1-2 years are likely developed by AI.
I think we should all hope that 2 happens before 1, but it doesn't feel good to hope for a catastrophe.
If this is what misalignment turns out to be I ... might be on board with it? At any rate it's nowhere near as concerning as what I had been expecting.
> Aren't there laws regulating bioethics that forbid such questionable practices?
Not generally, no. Most of it is by convention coupled with the fact that no reputable journal will publish your work these days if you don't go through an IRB for certain types of work. There's also the funding angle, with most work being paid for by one of a few very large governmental bodies and each of those having stringent rules about what they do and don't permit.
Basically pursuing certain lines of research would make you unemployable thus (almost) nobody does it. Unless it looks likely to turn a profit in which case you can start a company and try your luck with VCs.
That's a nice way of saying that it's a dependency clusterfuck.
I've never understood why we can't just expose the GPU ISA directly the way the CPU does. It's all getting compiled down at the end of the day so someone has to write a compiler for it either way. We'd be substantially better off IMO if it was all built directly into LLVM and then let middleware sort out the details.
That would require vendors to either stick with a single backwards compatible ISA like intel did for x86 or document how their graphics cards work.
CPUs manage this by changing the internal micro-architecture, but historically GPUs only needed to support a graphics API and used that abstraction layer to freely change the hardware.
I was so disappointed when I tried reading tintin in other languages and found the dear captain was straight up using slurs in those. I wonder whether the english language ones have been edited over the years to remove that sort of thing
reply