awwaiid

Well every time I see a locked door I def _think_ about what it would take to bypass it. Especially those business office glass double-doors with motion detection and a hand-lock on the bottom.

I knew declarative languages would eventually win!

Claude tried to shrink the number with that stuff too, but I directed it to keep only proven existing common tech. I also included delivery ligistics, maintenance, electric grid operators, etc. It would take a huge investment to get to the "current" level of automation worldwide. I can copy the estimates somewhere if you like. What is your estimate?

I've been trying to program myself out of a job for the last 25 years. I think I'm doing it wrong.

Farming is already pretty scaled in general, right? Napkin math with Claude's help says we can feed everyone with about 250k workers. The internet says we are currently using like 2 billion. Plenty of room for optimization even without additional technology advancement.

If you've got time to kill -- that says to me that it is faster, and therefore better. Time is money. How much time are we talking? If this is a 10x speed up on your research cycle then that sounds like a subtle revolution.

I think the idea is that it cost a lot more than $20/mo/user to create the tech, and that investment in the tech-creation won't be recouped by your tech-consumption.

I think it depends on how far out you run the numbers for the area under the curve and all the various costs and profits. Seems clear to me that the individual tech-creation might end up being losses but that the net global will be gain very shortly as we continue to adapt and accelerate based on using the tech.

OK so first compile python to a NN. But next let's twist or overlay that onto a Transformer-based NN. Then we can have a Transformer Virtual Machine (TVM) execute arbitrary programs.

Use some of that transfer-learning (adding weights on top of each other) and an LLM can be "born" with an algorithm deeply encoded.

... or because I don't want to stab people.

My dad told me once "Son. You don't get in trouble if you don't get caught". He was in real life expressing annoyance at having to punish me for something dumb, but my preferred interpretation is that morality transcends law.

Yeah but if my brother can't pass it, that doesn't mean he is NOT human.

When you peer into the soul of the machine it delicately resolves to `while(1){...}`. All Hail The REPL.

I think it's more like humans are a chaotic choir of subsystems all doing their thing and tossing up their directives until some sort of "win" happens or the volume is loud enough in some direction that it then gets reverse engineered into a "thought". But yes.

A useful illusion of Free Will is that we can to some degree predict the future state of the world and influence the future state of the world. To make meaningful agent type activity you have to have a world full of causality, in other words --- and to have causality it is awfully nice to have some deterministic properties (or science might not work out so well). So! I for one am a hard-compatiblist. I think that not only is "useful free-will" aka Agency compatible with determinism, I think that determinism is necessary!

(concepts probably stolen or bad copies of Daniel Dennett models)

:shipit: haha

(a) I like your analysis (b) maybe the wings are just for show so that you flap them out at clubs every now and then right when the music hits the drop or something and everyone is like WHHAAA????!!!!

Isn't zero-precent unemployment the exact opposite of the star trek utopia we were promised?

I should go watch that debate. From "Elbow Room", the rough idea I've gotten out of it is that traditional Free Will is meaningless and so he invented a new definition which we might now call Agency.

I don't recall how far he took it, but for me determinism is very helpful (if not required) for predictability, and predictability required for meaningful Free Will. So I'll have to go watch the debate, but my bet is that Dennett was willfully (pun intended) refusing to accept the other-definition of Free Will in favor of his own. I'll report back after I listen to the debate :)

I think ... right ok, I guess harder-to-fail but really it is easier-to-fail, easier to remain in a failure state, as a kid right? Same thing eh?

Yes -- the whole thing should be how many hours it takes to buy X instead of how many dollars.

It's then interesting to see how that looks in different economic brackets. Maybe in one bracket a car costs 1000 hours, and in another the same car costs 350 hours (so might as well get a 700 hour car. Still cheaper).

True.

However we might think about this differently if we flipped it to "our standard is 20% completely untested".

Uncoverage communicates the value much better.

A thousand hands on a Ouija board.

Yeah, but those evolved on high-gravity planets are smarter (and maybe stronger) since they too must calculate thrown-object trajectories but have to do so faster. Our brains use just enough energy, but no more, to keep us alive. Being smarter than we are would be a waste .... but if we HAD to think faster, we would evolve to match.

So maybe they'd figure it out.

I think they call that chemistry and it is indeed an important part.

I mean... That's it. That's the example. All Is Chrome. (small letters: or safari)

Well the Perl people also invented the defined-or operator (//), so there is that.

Sounds like a feature rather than a bug!

Maybe this is a missing-comma type issue, and they mean that the last time that AI came up with an antibiotic was 60 years ago

That would be a funny settlement -- "OK, so $10/month, and we'll go back to 1950 to be fair, so that'll be .... $8760"

I think the gap between attributable knowledge and absorbed knowledge is pretty difficult to bridge. For news stuff, if I read the same general story from NYT and LA Times and WaPo then I'll start to get confused about which bit I got from which publication. In some ways, being able to verbatim quote long passages is a failure to generalize that should be fixed rather than reinforced.

Though the other way to do it is to clearly document the training data as a whole, even if you can't cite a specific entry in it for a particular bit of generated output. It should get useless quickly though as you'd eventually have one big citation -- "The Internet"

I love pry+byebug so much that I'm having a hard time wrapping my head around why I might want to replace them with this. I guess for a marginally more out-of-the-box and universally available tool, eh? I'll have to look into the extensibility aspects.