Sparkfun ALC,"less expensive than hiring a contractor or employee to design pcb"

[schmitt trigger]

This thread has drifted somewhat, about how AI and or virtual assistants will start replacing human activities.

Well, this is the weirdest so far; virtual reality influencers are trumping over flesh and bone ones:

https://www.bloomberg.com/news/features/2020-10-29/lil-miquela-lol-s-seraphine-virtual-influencers-make-more-real-money-than-ever?utm_source=pocket-newtab

[TimNJ]

This attempt at EE automation is obviously half-baked. I think we can all see that. But, I think it’s clear that, with some more “thoughtful“ design constraints built into the tool, it could be reasonably good for a specific variety of widgets. And if that’s possible, maybe those designs should be automated. Again, providing the tool is advanced enough to have have a clue about EMC, etc.

For other types of designs, let’s call them “technically harder”, I feel you’d spend way more time providing proper design constraints in order to get a decent output.

[T3sl4co1l]

Yeah, probably for the number of permutations in their library, it would still be more effort to program than to just design each and every one of them.

Nevermind if you include simple hacks to interface modules that aren't, by themselves, strictly compatible.  You could add glue logic, level shifters, comparators for simple A-to-D, or advanced level shifting; filtering for PWMDACs, filtering and ESD protection for the outside world, etc.  Not to mention support for high currents, thermal management, high voltages, module clearances/keepouts, mechanical constraints, etc.  The number of support options goes up exponentially, along with nearly 100% of the library being usable together then.

And again, anything slightly optimized, hire a real engineer.  Token / good-faith efforts could be automated, but don't expect it to pass, expect to have it reviewed at least.  And of course there's always ways to abuse modules to fail EMC and such, let alone when software is included in the equation*.  Even a small team of engineers will miss things reviewing projects like that.

*Famous example from some years ago: FCC passing rules (or wanting to? I forget how it went) regulating SOHO routers, because they're largely software defined, if not in the intimate details (usually implemented on-chip; are they usually mask programmed, or flashed?), then at least in the selection of power levels and channels, allowing users to violate rules/guidelines like using the minimum power needed.  And if the RF core is also hackable, then literally anything else that can screw up the spectrum (faulty modulation settings, overdrive causing IMD and harmonics, or even destroying the radio I suppose, which is even a merchantability thing, as is simply bricking your router).

Tim

[Cerebus]

Yeah, probably for the number of permutations in their library, it would still be more effort to program than to just design each and every one of them.

I do hope you're joking -  \$^nP_k = \frac{n!}{(n-k)!}\$ gets big rather quickly.

[T3sl4co1l]

I meant of designs that are desired in a more refined state.

Tim