AI is going to struggle at building a consistent internal model of the domain into the software unless you’re able to give a structured explanation of the domain.
If you’re just giving it a set of inputs and expected outputs, it’s not going to generalise well and fail at out of sample input, unless the AI already understands the domain from its training set.
Being able to give a structured explanation of a domain (and being able to judge if the internal model of the software makes sense) is not the same as having experience in a domain.
Lots of ppl with domain experience can tell a right output from a false one, but can’t tell you why.
I agree, but only because YouTube is a wild west of trash, not because children somehow don't deserve to be entertained. I think that distinction should be made. Instead of focusing on barring children from the bad stuff, it might be worth trying to attract them to the good stuff. (I hear PBS Kids is a good app to leave your child alone with. No personal experience of my own though.)
You are contradicting yourself. If you're hoarding the data for yourself you're not going to develop something useful. Sharing the data means that it will be integrated into the big LLMs, which will be useful "for their own country".
>The facts show that just like the amount of labor is not fixed, neither is the size of the economy (fixed pie fallacy) and as more work is done, the economy grows
Your reply is a glib thought-terminating cliche strawman that doesn't address their point at all. Interesting!
Those theories are based mainly on the effect of Cuban immigration in Miami, however they lack a control so you can't really conclude anything.
Besides, yeah, if you hire people who will work for any salary, the amount of jobs will increase, but salaries will decrease, for locals as well. After some time, locals will flee sectors where the migrant workers are brought in, creating further self-inflicted "labor shortages"...requiring more migrants!
The main winners are capital owners, who, thanks to the migrant workers, can now acquire a larger part of the added value generated by workers.
They're available. Though you probably shouldn't invest to heavily in gen1 (production) sodium-ion batteries. It's looking like they'll be obsolete pretty quick.
I would agree if I could buy a AA battery that would power my toothbrush for a year. Or one that could be rechargeable with easily-available chargers reliably for a decade (and without having to drop 0.2V to achieve it...).
Any kind of consumer power technology can only ever be truly "good enough" if it never causes any inconvenience or significant cost.
Here are some rechargeable AA batteries that were specifically tested and worked in toothbrushes [1].
For almost all devices there is no good reason to care that the nominal voltage of NiMH rechargeable batteries 0.2V lower than the nominal voltage of alkaline non-rechargeable batteries. Alkaline batteries have a steeper initial discharge curve and pretty quickly drop below 1.3V.
If your device has trouble with 1.3V it is either going to almost instantly stop working if it is a high load device, stop working after using maybe 10% of the battery's capacity of it is medium load, and maybe 30% for a light load.
On the lasting a decade or more front, I'm still using 19 of the 24 1st generation Eneloops I bought sometime before March. 2 died and 3 are missing. Last time I went through and measured their capacities, about 3 years ago, they averaged 1886 mAh. They were sold as having an average 2000 mAh capacity with a minimum of 1900 mAh.
I've also got 15 4th generation Eneloops bought 2014-08. Those are also all still fine, with an average capacity of 1960 mAh.
You might wonder why I bought the 4th generation ones since the 1st were still fine. It is because they greatly improved the self-discharge. 1st generation was specced at retaining 80% charge after a year. 4th generation is specced at retaining 90/80/75/70 after 1/3/5/10 years. I've got some lower power applications where changing batteries is annoying, so I want to minimize self-discharge.
> For almost all devices there is no good reason to care that the nominal voltage of NiMH rechargeable batteries 0.2V lower than the nominal voltage of alkaline non-rechargeable batteries. Alkaline batteries have a steeper initial discharge curve and pretty quickly drop below 1.3V.
The few times I've measured the voltage of my non-rechargeable AA batteries (which, granted, was infrequently, and not recently), I haven't seen them drop below 1.3V until they've been in use a while.
And I've much more reliably observed that when I try to use rechargeables in my electric toothbrushes (Oral-B Pro Clean, the kind with separately moving round and long brush sections, which are, alas, no longer available anywhere I've been able to find), they start out very sluggish, and gradually descend to near-uselessness, while using non-rechargeables makes the toothbrush very energetic at the start, declining fairly steadily over a month or three, with it matching the level of the rechargeable at something like 2/3 of the way down.
I don't know what to say. I may not like it, you may not think it's actually intelligent, you may not think it's going to change the world - but how can you not see that this is revolutionary?
I remember before LLMs, someone on HN made a bot to program automatically by pulling the top rated answers from stackoverflow. To me agentic coding just feels like the next iteration of this.
And LLMs in general feel like an iteration on search.
The strengths and weakness of LLMs are already apparent, and in my opinion unlikely to change from here.
No. Nondeterministic output is not revolutionary. Technology forced down our throats by a few companies and executives who are licking their lips at the idea of laying off people, even if laying those people off means garbage products, is not revolutionary. Slop is not revolutionary.
Perhaps what people forget is that every great product builds on the past in a way to improve it. Buggy software and lame copywriting and kids not learning is not revolutionary. The people continuing to prioritize quality will be the revolutionary ones. Garbage is not revolutionary.
Thinking about it I have to revise my statement somewhat. I have seen The Great War, Technology Connections etc and my Youtube algo is after 15 years very tuned to me.
The issue is somewhat that this stuff needs to be pushed more into peoples feeds and not pregnant spiderman videos.
AI is going to struggle at building a consistent internal model of the domain into the software unless you’re able to give a structured explanation of the domain.
If you’re just giving it a set of inputs and expected outputs, it’s not going to generalise well and fail at out of sample input, unless the AI already understands the domain from its training set.
Being able to give a structured explanation of a domain (and being able to judge if the internal model of the software makes sense) is not the same as having experience in a domain.
Lots of ppl with domain experience can tell a right output from a false one, but can’t tell you why.