One wonders if someday we might be able to resurrect the neural network from dead cells by somehow reviving the connections between neurons. I imagine that the connections stay, but become dormant when the neuron dies.

There is nothing to resurrect. They get digested by the microglia.

Ah, I didn't know that existed. TIL

Perhaps, but I think that by the time we're that far advanced, strokes will be entirely preventable.

Strokes will never be preventable. You can mitigate them but a stroke isn't really a disease. It's a symptom.

An ischemic stroke (i.e. stroke due to a clot) caused by vascular or cardiac issues can be mitigated. A cryptogenic stroke however is idiopathic and therefore has no understood cause. These types of strokes make up 30-40% of all strokes. Unless we figure out their cause, there's no way to really prevent them.

But then there's also hemorrhagic strokes which are an entirely separate category that has causes and mitigations more or less diametrically opposed to those for ischemic strokes.

And of course those are just your broad painted categories and they are generally looked at as the start of a medical emergency but strokes happen all the time as a consequence of other medical emergencies.

Even if you could perfectly prevent strokes in generally healthy populations, those same people may still end up suffering from a stroke during a surgery or during/after a major accident or injury. No amount of preventative medication can prevent someone suffering a stroke caused by a brain bleed after a car accident. Likewise for someone with a crush injury, internal bleeding, or broken bones that end up throwing a clot which makes it into the brain.

So any advancement in halting and reversing damage from a stroke will be a massive boon for emergency medicine until the end of time. Unless of course we somehow find a way to cure/render humans immune to blunt force trauma or lacerations.

Sure you can. Just not with any technology on the horizon. But there is conceivable technology (e.g. medical nanotechnology) that could prevent strokes or stop them as they are happening.

Like what?

Like detecting constriction or loss of integrity of blood vessels, and doing the corresponding intervention.

The saddest thing here is not that it requires some future nanotechnology, but is achievable at the present scientific level, yet too expensive to develop, and wouldn't see FDA permission in a decade or two anyway.

I've generally assumed that AI would make developers get lower compensation because of the lowered quantity of developers required for the same output, but this raises the possibility of it actually increasing if more developers end up doing their own things instead of entering the broader labor market :)

It could increase compensation by growing the economy. (E.g., perhaps counterintuitively, skilled immigration has this effect.)

the problem is that very few to none SWEs “doing their own thing” will ever make a penny out if it. whatever they do, if it actually makes a little traction, will be cloned and copied in a week by someone else. this whole idea that “we’ll see a 1-person billion dollar startup” is as silly as it gets

Imo the main benefit of AI is allowing faster experimentation and more parallelism, as well as faster iteration.

Shouldn't there be a feedback system here preventing the scheduling of loads when cooling is degraded?

With hyperscalers for sure.

But this is the physical world, shit happens.

The algorithm didn't know that fuse was lose and fine at 50% duty cycle but was high resistance and going to blow at 100%.

I wonder if this is in response to the chrome incognito lawsuit.

cries I am sad at all of the cool designs that would work with optane if it had been widely adopted. We desperately need more points on the latency/cost per GB curve.

This doesn't math out to me just based on what I know of energy consumption numbers.

"is duplicated by" is amusing.

While I largely agree, it does raise the prospect of testing this iteratively. E.g., give a model some fake environment, prompt it random things until it does something "bad" in your fake environment, and then fix whatever it claims led to its taking that action.

If you can do this and reliably reduce the rate at which it does bad things, then you could reasonably claim that it is aware of meaningful introspection.

Yes, sort of. Generally you can measure the pass rate on a benchmark given a fixed compute budget. A sufficiently smart model can hit a high pass rate with fewer tokens/compute. Check out the cost efficiency on https://artificialanalysis.ai/ (say this posted here the other day, pretty neat charts!)