[markwu2001]

PCB Designer on the team:

Holy cow this is awesome! The integration of project context with an LLM enabled more than the sum of its parts when it comes to information access.

When designing circuits for fast iterations, you're constantly managing risk and time expenditure; this leads to copying a lot of "flight heritage" circuits so to speak.

With the LLM, I'm able to get highly condensed starting points for better circuits when it comes to simplicity, understandability, or performance. As of right now, all I have to do is fact check the solution which is much easier than synthesizing one myself. I'm citing the example where I got it to design me a viable high gain analog frontend that meets my gain, bandwidth, noise, and biasing requirements.

[boringuser2]

I have no idea what you're talking about, but my experience 100% mirrors yours in most cases as a programmer.

GPT-4 produces a lot of my code now.

I'm leap-frogging my team a bit in productivity because they still don't like it, but it's so close to being undeniable.

[anthlax]

I’ve been unable to replicate this. Could you please show me an example? I ask everyone who makes this claim and have yet to see a concrete example. I just can’t get it to do anything useful for me. I feel like I’m missing the boat!

[porkbeer]

It wont write your entire program, and you have to already know enough code to know when it gave you garbage, but.. I find I can have it tackle small chunks and in some cases even glue them together in a usable way. It can often remind me of stratigies I would not have thought to use, good or bad. It can also do some basic debugging, including seeing things my tired eyes often miss. That said, you kinda already need to be able to code or you wont know the wheat from the chaff. It feels a lot like managing a remote esl person you will never meet.

[MuffinFlavored]

I don't get this. I asked GPT-4 to help me build a machine learning program for historical stock prices. It spat back about 8 errors, functions that don't exist, nothing compiled, multiple logic flaws...

Which one of us is lying?...

[ugh123]

Probably neither of you? I've had better luck coding with ChatGPT when I ask it very concise questions, mainly at the function or small class level, but still asking for enough functionality to feel satisfied. It helps to give it a "shell" of some code, or a starting point of code to build from. I also make sure to ask it not to include 3rd party libraries (I'd consider those under separate discussions with it). It's truly helpful. And that's saying nothing about copilot integration in the IDE which is sooo good.

[newswasboring]

None of you are. You are using the tool improperly. The tool cannot spit out entire applications yet, especially if all you told it was the final outcome. I see GPT-4 as a junior programmer in my team who is unusually productive. It cannot do the design for me, but executing it is a piece of cake for it.

[MuffinFlavored]

> The tool cannot spit out entire applications yet,

Yesterday I had a ticket at my job to extend some functionality in a code base that was probably 200 files and 100+ lines of code in each file, and that's before any `import` references to other libraries.

How can you feed all of these tokens to GPT-4 in a cost effective way so that it knows about your application well enough to recommend/pull off code completion at a human-like level?

[newswasboring]

I'm working on exactly this. I think we can use embeddings and langchain agents to do this.

[SanderNL]

The largest models like GPT4 have the interesting property to really, really finish what you started. If you start with flaws of any kind, it will continue to produce them. The inverse is true as well.

This is an actual thing[1] and it’s something larger models are actually worse (better?) at. They score higher and higher on the loss function (did I predict correctly), but their utility (does it work) goes down.

Just thought it was noteworthy.

[1] https://arxiv.org/abs/2102.03896

[MuffinFlavored]

> If you start with flaws of any kind

I asked it to start and it provided me the flaws :D I can share the prompt if you'd like.

[SanderNL]

Please do, I’m curious.

[boringuser2]

Asking GPT-4 to make you an entire application as a complete amateur is absurd and shows you don't understand the context window.

[bombolo]

Maybe he just writes really really really simple code all day and faked his way into a developer position? And then his team mates have to fix/redo anything he does?

I've seen it happen several times.

[boringuser2]

Wrong.

[gfodor]

I’ve been interested in learning how to design PCBs, I wonder if this tech helps constrain the set of things one ought to focus on learning, with the assumption that AI can fill in other areas.

[markwu2001]

I love teaching PCB design! As with you, I'm also really curious about AI's contributions to learning. My opinion is that the automation of search and execution increases the impact one person can make so more time can be spent on connecting the dots at a higher level. Whether or not education catches up in time is another story haha. It's pretty open ended right now but it's undeniable that a huge educational gap there.

[gabereiser]

This is what's been holding me back from PCB design. Like, I know the math, but there's forces at play beyond just V1/R1=V2 kinda stuff. I'm really keen to see if an LLM can assist a hobbyist in making sound, quality, designs that aren't going to break his/her wallet when the sheet comes in and none of it works.

About 12 years ago I designed an RF addon board for an RC car (to toggle lights). Even something as simple as switching with NPN transistors looks good on the bench software but when the chips came it didn't work at all. Did I mistakenly ask for PNP? Ugh! I used fritzing at first (still great software) and then upgraded to Eagle. I want to do something beyond Raspberry Pi/Arduino work but not quite ready to design my own radio/motherboard. I know electricity but I don't know why I would need a capacitor/resistor/transformer other than the very basics.

What resources would you recommend since you love to teach PCB design?

[neodypsis]

This XSeries on edX [0] looks promising for some of the prerequisite knowledge. Then for PCB design specifically, there are specialty courses on Udemy, for example, "KiCad Like a Pro" [1].

[0] https://www.edx.org/xseries/mitx-circuits-and-electronics

[1] https://www.udemy.com/course/kicad-like-a-pro-3e/

[neodypsis]

Have you tried KiCad? There are some interesting courses on Udemy about using it for PCB design.

[built_with_flux]

Here is nice demo video

https://www.youtube.com/watch?v=5aCN92kOxTU

[nullsense]

How often is it able to get it right first try?

[markwu2001]

I'll interpret "right" as in giving correct information. I found that it depends on the type of question asked; if it's somewhat general information such as design concepts and design topologies, it impressively nails it 99% of the time for my limited testing period. The main area it gets things wrong relates to very specific information that has only one source: e.g. datasheet specs. I heard the team already has plans to mitigate hallucinations on this front.

Also, I find myself having a conversation as I would with a real person, so it's less of spitting out the right answer for any technical question I might have, but more of a design buddy to consult when you're blocked on your search through the massive electronics solution space. Of course, this is a very preliminary use case and I can't wait to see what other people use it for.