How AI thinks about electronics.

Thread Starter

74F00

Joined Oct 26, 2026
58
The AI labeled it LED circuit, I would at least expect further explanation.
Oh is this why? What I meant by "I would at least expect further explanation" was if a human was doing it. All to my point that AI is acting unlike any human.
 

panic mode

Joined Oct 10, 2011
5,458
i don't know... it is only a toddler...
even if we look at the earliest days of generative AI (2020. Chat GPT), that would make it 5-6 years old. but for some reason you are comparing it to an adult human whos cognitive levels were sharpened over 100000's of years and who has specific technological experience. and you are complaining about details that are only caught by peripheral vision since not the main focus of the shopper. People have narrow focus and short attention span. but suppose you do zoom in and print it as a massive banner and place it at the entrance of some electronics store (whatever RadioSchack would be today). how many shoppers of that store, (not people shopping at Victoria's Secret or Home Depot) would bother to even look at it? that's right... nobody cares. and that makes it good enough that it does not justify further effort (computation/processing). in fact if you gather bunch of those people (RadioSchack shoppers) and ask them to sketch the actual working breadboard circuit (one they can look at), what do you think results would look like? like garbage - thats is how. they would not get even close to what AI did.
1793291577699.png

and today people are pretty dumb... they do not even know how to write in cursive, not to mention read each other's handwriting (and sometimes - their own). so comparing those "artistic skills" with a toddler AI, it...does not look good - for the humans.

just check this out, the most educated/competent people having problem with battery and lightbulb:
https://forum.allaboutcircuits.com/threads/mit-grads-struggle-with-lighting-a-light-bulb.71323/
 
Last edited:

Thread Starter

74F00

Joined Oct 26, 2026
58
i don't know... it is only a toddler...
But why are we treating it like a human? The human toddler is less capable because it has little experience (this includes muscle movements where they are not very stable). This AI has much more experience then any human has ever had!

but for some reason you are comparing it to an adult human whos cognitive levels were sharpened over 100000's of years and who has specific technological experience.
Why shouldn't I? Why are you measuring experience in years? I know that years of experience is alright to measure with humans, but with AI it is completely irrelevant. An AI model can ingest, analyze, and learn from millions of documents, images, or code lines in hours or days—a task that would take a human a lifetime or longer. And about the fact that we have sharpened our electrical understanding over 100000s of years is also not a problem because pretty much all of what we have figured out about electronics has been fed into AI. Also 100000s of years is to much, we really have only been doing this for 320 years (unless if you count a few ancient static electricity experiments in ancient history that now really is only 1 paragraph to fully explain to AI). But if for some reason you think time is a factor "let it sit" well then explain yourself, I don't see any reason why time itself not experience will help. Every day AI models are learning what to us would be a lifetime of experience. So yes, it is totally reasonable to compare AI to an adult human with specific technological experience because AI has analyzed much more electronics then anyone can even if they use their entire lifetime trying.

and today people are pretty dumb... they do not even know how to write in cursive, not to mention read each other's handwriting (and sometimes - their own). so comparing those "artistic skills" with a toddler AI, it...does not look good - for the humans.
Humans today are vastly more educated than at any point in history, with global literacy soaring from about 12% in 1820 to roughly 87% to 90% today. Perhaps today there is a less percentage of people that know cursive but why is that relevant? Cursive is simply a form of writing. Back then people didn't even have a spelling standard. Just look at their dictionaries (which were always changing you cannot find two that are the similar):
johnson_dictionary.jpg
This text is an excerpt from the ‘Plan of a Dictionary of the English Language’ (Sp Coll Mu51-c.25, pp 4-5)
As for Ebay, thanks for the advice, though I have seen enough good deals as to not dismiss it entirely.
 

panic mode

Joined Oct 10, 2011
5,458
you are looking at AI's ability to draw. humans draw by holding pen (or whatever tool) in their hand. the most basic form of that is to write text. if you want to compare AI/human based on artistic skills, isn't that a fair comparison?
also you are wrong about time. time is a factor... always. it takes time to read and process data. and Chat GPT from 2020 is not the same one from 2022 or from 2026. models are constantly tweaked and improved, hardware on which it runs is upgraded, so why do you assume that version from 2026 does not need to start learning from scratch? just because "it is online, on internet" it does not mean that it never gets reset. in face any moderate change to OS on computer or phone that you are using right now requires reboot. and when you get new computer it still takes time to load software on it. maybe a lot of what was learned before can be reused faster than learning from scratch but simply reading large data set is still time consuming. and in terms of maturity, AI is still in its infancy. your (rational) mind, seem to have problem accepting that.
 

Thread Starter

74F00

Joined Oct 26, 2026
58
you are looking at AI's ability to draw. humans draw by holding pen (or whatever tool) in their hand. the most basic form of that is to write text. if you want to compare AI/human based on artistic skills, isn't that a fair comparison?
That is true. We draw by moving our hand in movements we have already done thousands of times. AI draws in a harder way. Its individually changing the pixels. When you took that photo of your writing a camera, it took all of your physical work and made them into a digital image of pixels. Now imagine making that image with pixelart.com, even the best are not that realistic with color and curves. Its really hard, trust me, I have an account on there and have trouble even making basic pictures. I don't think its very fair.

also you are wrong about time. time is a factor. it takes time to read and process data. and Chat GPT from 2020 is not the same one from 2022 or from 2026. models are constantly tweaked and improved and why do you think that version from 2026 does not need to start learning from scratch? just because "it is online, no internet" it does not mean that it never gets reset. in face any moderate change to OS on computer or phone that you are using right now requires reboot. maybe a lot of learned before can be reused faster than learning from scratch but simply reading large data set is time consuming.
I don't understand. The whole point relies on this sentence: "maybe a lot of learned before can be reused faster than learning from scratch but simply reading large data set is time consuming." which I don't understand.
 

Lo_volt

Joined Apr 3, 2014
377
We know that AI/LLM's are Machine Learning, just on a grander scale than any other Machine Learning effort. We also know that Machine Learning has been around since the early days of electronics. My favorite example is a coin acceptor. The designers of the coin acceptor ran multitudes of coins, fake and real, through their device and recorded the output. They segregated the good coin signal responses from the fake coin signal responses and used that as the operating criteria for their device. This was back in the 1980's.

The current spate of LLM chatbots is one that is attempting to answer a VERY broad array of questions. It takes an hellacious amount of stored data to answer every question we are asking. To quote Michael Crichton, "The designers are so concerned with whether they could, they forgot to ask if they should." Google's search engine was part way there already and at least pointed directly to source material without claiming authorship. My point, though, is why? Why do we need to digest this data in this manner? Was it not good enough to point to the source material? Why might I need to have an AI agent summarize the pdf when I can just read the original document?
 

Thread Starter

74F00

Joined Oct 26, 2026
58
The designers of the coin acceptor ran multitudes of coins, fake and real, through their device and recorded the output. They segregated the good coin signal responses from the fake coin signal responses and used that as the operating criteria for their device. This was back in the 1980's.
These AIs are a bit different. They use pattern recognition. So for example if of all of the data they saw the most common letter is "a" then they would only say "aaaaaaaaaaaaa" and nothing else (not to mention feedback into themselfs to further want more "a"s) but what these do is if they see they "appl" was just written they would know from probability that "e" was next. That is also how they answer questions. This is a much newer idea.
 

panic mode

Joined Oct 10, 2011
5,458
Humans today are vastly more educated than at any point in history,
exactly, they are fed or have access to more information than ever before, they go to school longer than ever before, yet they know and understand less. that is a terrible reversal...

consider exams from the past. ask people around you to answer this without access to internet or library.

1793302931879.png
 
Last edited:

Thread Starter

74F00

Joined Oct 26, 2026
58
AI makes mistakes because it does not understand enough. humans do the same to deceive intentionally.
Oh man that is some claim. I just realized that was said. I think we need much more evidence for the first sentence, and for the second sentence I completely disagree.
 

Thread Starter

74F00

Joined Oct 26, 2026
58
Perhaps they won't be able to answer some of those.
ask people around you to answer this without access to internet or library.
But lets be realistic so many people have access to the internet and library. And more importantly, we can't overload students, we have a lot more to teach nowadays. I bet that there is a list longer then that one with questions modern students know and older students don't.
 

panic mode

Joined Oct 10, 2011
5,458
Oh man that is some claim. I just realized that was said. I think we need much more evidence for the first sentence, and for the second sentence I completely disagree.
it is a definition of snake oil salesmen.
a good example is when you are buying a car or house or anything used. or when you order HDD that is sold as 1Tb (=1024*1024*1024*1024) but you get less than that (1000*1000*1000*1000). same goes with car fuels efficiency and bazillion other things.
or you find online deal you like and shipping is way more than product you are paying for. or you increment number of items and shipping is incremented linearly with quantity selected. or the shrinkflation packaging of food etc.
none of those are accidents, they are deliberate acts of deception.
 
Top