Mechanical Engineering

I think we can use it in situation when we need to find reason for our major manager to achieve increasing selary or additional vacation ...

But if serious in some cases this is quite dangerous to use Chat GPT in complex task's, you can use it in case for example when you need to write a simple macro on VBA (visual basic). I saw video when a person use it with excel programming, or maybe you can use it like a little bit faster type of searching in internet , its only on my opinion. I think during some time it will grought to much better system but as for today in mechanical tasks it's too young, and you need to be careful with the information which you get from there ...

Its good with Windows, office, sharepoint, Excel, vba, and solving problems with Inventor. Always tell your version of software first.

it is also good with Inventor and iLogic.

It writes a pretty great outline or preliminary engineering spec.

it can help you find a job, or write a resume, cover letter, etc,

and make lists.

if you are in software authoring, game scripting, etc, this will be life changing.

it can help with supply chain issues, to find vendors, select and review materials, etc.

Hi @Lano Kuyama,

My experience with ChatGPT is recent but very promising.

I have not attempted to use it for technical graphical or mechanical aspects of design, and I don't know if it has such capabilities. But I have used it for some aspects of "conceptual engineering" that are enriched with "verbal skills".

For example, the way in which a designer "interprets a problem" is not unique, precisely for this reason we say that "it is his interpretation" and it is full of subjectivities, which are often hidden behind hard data and statements that seem very technical and indisputable. But I've been studying creativity systematically for a long time (to try to teach it to my students) and I've learned that the contribution "verbal skills" make to mechanical design is truly immense.

In this sense, my "conversations" with ChatGTP have pleasantly surprised me. I have rehearsed with him the same "lateral thinking" exercises that I give my students in design classes... and I noticed that he has a large number of hits and almost the same mistakes as any human. But with a huge advantage: IT PRESENTS MENTAL FLEXIBILITY!

In other words, with "very little prompting on my part" he has managed to reconsider his proposals and greatly broaden his vision of the problems!

This is something that, despite the "youth of this AI", has pleasantly surprised me and makes me think it will be a wonderful design tool, if used properly.

I have also used it to make some applications in Python from scratch (which is a language I had never used) and it worked really well. What I have noticed is that you have to ask "short" questions, which respond to very specific tasks, and test and correct them with their help.

If you ask it for the code of a complex task, it will try to provide it, but it always breaks or makes errors that are later difficult to find (at least for me, not knowing the programming language).

Well, this is what I can tell you about it.

Kind regards

Its good for documentation and general project management but i have yet to see a use for more technical problem solving. Maybe as it gains more knowledge

An enthusiast has added GPT-4 to Blender, with which you can create 3D scenes and models. Here is Github link.

I think that over time GPT will be able to build standard designs from standard elements in CAD.

This technology will make life easier for engineers.

I messed around with it, Asking it questions that I "used to know" sophomore in college level questions. This is what I found, the principle was good, it looked up the right equations, but when it worked it out, it used all the wrong units/conversions. it would work in inches, but deliver in the answer in ft/sec. The equation would be by density, but the answer in cubic meters.... I would correct it, it would apologize, redo the work correctly and give the right answer. IF I corrected it with wrong information (say the density of water is .456 lbs/in^3), it would take what I said as true and add it in without correcting me back. interesting but I wouldn't trust it wholeheartedly yet.

You just give them ideas to replace you! JK :)

1. You can scan an object and use ChatGPT to generate its drawing
2. You can make it look at a product and propose the components
3. You can make it look at Space and suggest whether there is alien life out there depending on the facts not theories.
4. Make ChatGPT analyze the organization chart and decide whether all these layers and people are really needed.
5. Give it facts as-is and get plans for the future!

Haha, definitely generates a lot of excuses to give your boss when you decide to miss work.

You are right though, it's pretty useful to engineers/designers in different ways but is yet to see how it can be used to solve complex tasks.

Chris, I am in deep-tech inventions this does for me at all is not meant for engineering and deep tech programming, it is for primary school kids at lower level too weak.

never mind my writing missed something I meant this does not work for me at all.

I would have liked to see some of your "deep-tech inventions" work in your library (even if it is some images that do not compromise the confidentiality of the developments) to better understand your point of view and your assessment of the level of AI as "primary" school kids at lower level too weak".

Capabilities of 5D developments and multitask detection systems, some engineers out there are working precisely to bring the real design software in to the market with full automated animation of a 5D but still kept indoor.

The owners of Chat GPT just gave it access to the WWW for information. All hope of its usefulness is now lost.

I like it if chatgpt can assist me through some design processes like validation calculation, help with tedious research, alternative design, and the possibility of failure (like how bad this design can get). I try to ask a bunch of questions about the free body diagram, linear expansion, pressure, and buoyancy forces it pretty messes up. it got the right principle tho. but when it came to analysis and calculation it was wrong completely, if I corrected it with the wrong answer it was just agreed with it. I try to teach it to step by step till it can get the right answer, but if I ask another question with a different variable, it just goes completely wrong again.

Let me make a suggestion for the use of GPT, since it is still "very young", is in the process of learning, and its initial orientation was towards natural language issues and not technical problems or engineering analysis.

In that regard, I think it's vital to give you a proper "context" about which you can make inferences. Context is a (limited) amount of information that you gather at a given time, during a certain conversation. It is different from the "global knowledge" that you have access to from the www. It has a parallel with "working memory" and "long-term memory" in humans.

When a human solves a problem, his success depends on "contextualizing" it by bringing from long-term (huge) memory those concepts that are applicable to the problem, and loading them into working memory (which is very limited).

During an engineering conversation with GPT, it's good to start by "asking/going over" the basic applicable laws and operational issues that can cause a deduction to fail (such as handling your units). Once this context is built (limited to a few thousand words), it is more efficient to present the specific case that one wishes to solve and, eventually, help to find its errors by mentioning compliance with those laws and operational issues.

Helping this AI to contextualize has given me good results compared to the "robotic use" that many users give it, as if it were "an inexhaustible and unambiguous source" of knowledge.

Humans, in the face of an exam like the ones we ask the AI, also work better if the one who examines us "leads us little by little" towards the specific problem and collaborates with the search in our long-term memory and the load of everything that is relevant, in working memory.