I was in love with electronics but I realize that electronics is becoming something outdated with fewer and fewer practitioners as a hobby.

Programming is gaining more and more practitioners as a hobby.

But now I am surprised that programming seems outdated, I no longer need to know programming because AI applications have appeared that write the program by themselves according to the user's wishes.

Search on the internet:
Android application written entirely with Ai
or,
WEB page written entirely with AI.

How do we adapt to this change that seems to be a technological revolution?

 

By not believing in the AI hype of actual intelligence in today's systems.

The people that love hardware and love software will use these systems to expand and improve human resources, when and if they become useful to expand human capabilities using human intelligence.

Those without the love, well, sorry.

 

Too bad the US government didn’t have AI when they needed someone to build the website to manage Obamacare enrollment.

This is intended as a joke, just in case…

My point is that real software development is many orders of magnitude more complex than anything a current AI can do.

 

I'm a programmer by trade and I use AI all the time to help me write software. I can say with certainty that computer programmers and engineers will always be in demand because the AI itself will always need to be augmented with new training data. Not only that, current AI is only capable of solving specific problems which has to do with the way the AI is designed from the bottom up.

If you aren't a programmer or mathematically inclined (tensors, linear algebra, etc.), it might be hard to wrap your head around what I'm saying because the concepts are so abstract. AI ultimately represents the collective knowledge of humans which is always changing and improving. This knowledge is encoded into symbols which further encode more symbols.

Eventually we reach a brick wall where the AI runs out of symbols because it has no way to detect and interpret them. Oddly enough, these symbols arise from the human limbic system not so much higher brain function. Without the limbic system (think of an array of sensors), humans (and our AI extension) has no way to interpret and interact with the world in a way that is consistent with the scientific method.

Consider self driving cars, unless a human tells the AI that a red light means stop, it may never reach that conclusion through its own calculations. It may figure out after watching 1 million videos that cars almost always stop at red lights but what about the cars that don't? Without software engineers, there will always be that uncertainty which of course is a risk to life and property.

You mentioned an app or website and the fact is these tasks are relatively trivial compared to computer vision in self driving cars. There are fewer variables and getting those variables wrong carries less consequence.

As time goes on, government and corporations will be under increasing pressure to educate people on the limitations and capabilities of AI as well as how to program it. Currently we are in a situation where most people have absolutely no idea how AI actually works which is really no surprise just like how most people don't understand how an airplane works. Will we ever get rid of aerospace engineers? Probably not and why? Because the conditions of human flight are always changing in response to newer technology and risk factors.

This is my view on the topic and there is plenty of heated debate on the internet. My point is humanity has a ton of work to do before we get an AI that doesn't need a human to keep it in line.

I also suspect AI will never reach that point simply because the universe is too big and it's full of symbols that depend on other symbols which depend on other symbols and so on. The irony is most folks severally underestimate the hardware that is their own brain and body and how much information is processed per second and what this information represents relative to reality. Each of us is essentially a super computer running on hardware that has been evolving since life began.

By comparison, current AI models represent a small portion of what the brain and body are doing every single moment. Obviously AI is really good at certain things but so is a 100 year old mechanical calculator. The question in my mind isn't so much what AI can do, but what it can't do and that's a very long list.

 

I don’t worry about whether others are interested in my hobbies. Some danger of electronic supplies drying up I suppose in the extreme, but so far that’s fine. I have enough supplies to last me a very long time…

 

I was in love with electronics but I realize that electronics is becoming something outdated with fewer and fewer practitioners as a hobby.

Programming is gaining more and more practitioners as a hobby.

But now I am surprised that programming seems outdated, I no longer need to know programming because AI applications have appeared that write the program by themselves according to the user's wishes.

Search on the internet:
Android application written entirely with Ai
or,
WEB page written entirely with AI.

How do we adapt to this change that seems to be a technological revolution?

I'm a seasoned programmer (as we used to be called) and only recently toyed with all this AI stuff. Frankly it is very useful but has limits and can be wrong, so trusting what it says is not advisable but aside from that it is very helpful.


Look:



Now you could dig this out by searching through articles, samples, GitHub and so on but what a slow process that is (I've done it many times). Now ask a further question:



You can learn a lot with this kind of interaction, huge time saver but it won't replace you, it enhances you if used sensibly.

Bear in mind too, that you can ask exactly the same questions a few weeks later and might get a slightly different answer.

 

I'm a seasoned programmer (as we used to be called) and only recently toyed with all this AI stuff. Frankly it is very useful but has limits and can be wrong, so trusting what it says is not advisable but aside from that it is very helpful.


Look:

View attachment 346504

Now you could dig this out by searching through articles, samples, GitHub and so on but what a slow process that is (I've done it many times). Now ask a further question:

View attachment 346505

You can learn a lot with this kind of interaction, huge time saver but it won't replace you, it enhances you if used sensibly.

Bear in mind too, that you can ask exactly the same questions a few weeks later and might get a slightly different answer.

As a seasoned programmer it is useful because it's a time saver, not a learning experience. For learning, it's horrible IMO, as it fools you into thinking you are a seasoned programmer. When in fact, you are a paint by number coder.

 

Here's more AI for you, this time Visual Studio:



I typed the function and signature, then the two braces and this is how Visual Studio reacted, you can see its suggestion. The code exists elsewhere in the app (and destined in fact, to be replaced with this helper method) and the system has inferred that this is what I wanted to write and it's pretty good, this is more or less what I'm doing, I just press TAB and it will insert that suggested code.

Do others here see in their IDEs? I suspect Visual Studio Code also does this too but don't know about other tools.

 

Here's more AI for you, this time Visual Studio:

View attachment 347583

I typed the function and signature, then the two braces and this is how Visual Studio reacted, you can see its suggestion. The code exists elsewhere in the app (and destined in fact, to be replaced with this helper method) and the system has inferred that this is what I wanted to write and it's pretty good, this is more or less what I'm doing, I just press TAB and it will insert that suggested code.

Do others here see in their IDEs? I suspect Visual Studio Code also does this too but don't know about other tools.

Like I said before, useful for an experienced programmer that already knows how to that that code but deadly to the task of learning for a beginner. Don't ever trust the code it regenerated (that means the 'AI' doesn't understand the code, only that it's statistically likely to be the next words in the code sequence) from the universe of previous human intelligence created examples.

They call the stuff it makes up, hallucinations, but the real word is Bu!!sh!t.

 

Like I said before, useful for an experienced programmer that already knows how to that that code but deadly to the task of learning for a beginner. Don't ever trust the code it regenerated (that means the 'AI' doesn't understand the code, only that it's statistically likely to be the next words in the code sequence) from the universe of previous human intelligence created examples.

They call the stuff it makes up, hallucinations, but the real word is Bu!!sh!t.

View attachment 347589

There are definite limits, in that example it continued to suggest SPI5, SPI6 and so on and was unaware that these simply are not present on the hardware I was coding for, it likely would have continued all day of I had let it.

But it is very helpful when writing repeated statements that follow a pattern, it can often correctly surmise what you are doing and that saves time and reduces error especially fatigue related error.

But I do not worry that I'll be out of a job soon, the real world is messy and edge cases abound, these tools don't do well in that domain.

 

There are definite limits, in that example it continued to suggest SPI5, SPI6 and so on and was unaware that these simply are not present on the hardware I was coding for, it likely would have continued all day of I had let it.

But it is very helpful when writing repeated statements that follow a pattern, it can often correctly surmise what you are doing and that saves time and reduces error especially fatigue related error.

But I do not worry that I'll be out of a job soon, the real world is messy and edge cases abound, these tools don't do well in that domain.

I've tried it once or twice and decided I liked typing much better than trying fix the bugs it created.

 

Hi M,
I'm having a break from a 6 year project, as programming is really hard for me, so I needed it.

I was given a pair of small radio transmitter-receivers. I wanted to program them, as a mountain rescue aid.
I have never programmed with Arduino or used AI, before.
I tried ChatGpt and over a few days of exchanges 'we' managed to do it.
C.

 

I'm a seasoned programmer (as we used to be called) and only recently toyed with all this AI stuff. Frankly it is very useful but has limits and can be wrong, so trusting what it says is not advisable but aside from that it is very helpful.


Look:

View attachment 346504

Now you could dig this out by searching through articles, samples, GitHub and so on but what a slow process that is (I've done it many times). Now ask a further question:

View attachment 346505

You can learn a lot with this kind of interaction, huge time saver but it won't replace you, it enhances you if used sensibly.

Bear in mind too, that you can ask exactly the same questions a few weeks later and might get a slightly different answer.

this is interesting. How did you give it the specific context of your personalized hardware. Ignore if you were using AI for standard SDK/DVK type hardware whose libraries and Application notes are available on the internet.

 

this is interesting. How did you give it the specific context of your personalized hardware. Ignore if you were using AI for standard SDK/DVK type hardware whose libraries and Application notes are available on the internet.

Hi K,
I think that all information that AI uses is on the internet, but it can filter out sections, depending on your replies.

As for when it keeps repeating errors for ever, (note my code is maybe not difficult for it) instead of repetedly replying 'your wrong', try starting again with a better question including what it kept repeating.
C

 

Hi K,
I think that all information that AI uses is on the internet, but it can filter out sections, depending on your replies.

As for when it keeps repeating errors for ever, (note my code is maybe not difficult for it) instead of repetedly replying 'your wrong', try starting again with a better question including what it kept repeating.
C

yeah. but the internet based knowledge will give wrong results for a custom PCB right? Since @Futurist says he generated code from AI, the hardware must've been commoditized or else AI wouldn't know.

 

yeah. but the internet based knowledge will give wrong results for a custom PCB right? Since @Futurist says he generated code from AI, the hardware must've been commoditized or else AI wouldn't know.

Hi K,
Yes, but as soon as you correct it to your specs, it is now on the internet.
C

 

this is interesting. How did you give it the specific context of your personalized hardware. Ignore if you were using AI for standard SDK/DVK type hardware whose libraries and Application notes are available on the internet.

That image was like the third or fourth question I asked on that subject. Write very clearly, minimize ambiguity and you'll get better responses. You can say things like "so if I did that twice, would it work?" but better to ask "so if I iterate the while loop twice, will that eliminate the propagation of the unwanted flag in the R3 register?".

Its, extremely helpful, routinely I ask it language questions rather than looking up the manual, or I'll post code fragments and a compiler error message to get guidance, much as you'd ask a coworker sitting next you "Hey Tony, what's wrong with that expression?".

 

Hi K,
Yes, but as soon as you correct it to your specs, it is now on the internet.
C

Only one unique copy with millions of others that are much more likely to be selected for the next statistical pick during the next update rescan. It will take a lot more than your correction to change it.

If it helps then use it but don't ever expect intelligence or understanding. Intelligence or understanding is still firmly in the human domain for now and IMO for the foreseeable near future.

 

yeah. but the internet based knowledge will give wrong results for a custom PCB right? Since @Futurist says he generated code from AI, the hardware must've been commoditized or else AI wouldn't know.

That's why specificity is important.

"How does an MCU do SPI using C?"

vs

"How do I write the SPI peripheral setup code in C for an STM32F446RE MCU?"

In my experience the more details you include in the question the better the quality of the answer.

 

I don't know if AI is learning you or You learning it But it helps some what but debugging is pain I't told me today that XC8 is junk the datasheet don't match its naming for simple
stuff