Absent from the forum for a few months, I realized that we are starting to see research done on Chatgpt and in the solidworks doc the presence of AI is mentioned without further precision.
I also saw a potential concern about the future of our professions with the absurd fantasy that AI will put us out of work.
Obviously, this is not the case, but it is necessary to give arguments that would support this position.
If you are interested in analysis, let me know! I promise you not to drown in abstruse terms such as: * Token
Also interested, I use this tool more and more in the design, optimization and calculations phases of all kinds, there is a real time saving and surprising results. for the drawings and plans it's not yet perfected but it will quickly overtake us... be continued
We believed the same thing about programmers (web and others), taxis, researchers, teachers, musicians and singers, graphic designers... And yet it did happen.
@Sylk I hear what you're saying but you have to reason by degrees of complexity.
If I take your example of the taxi, it's quite simple to solve because it doesn't create anything. For example, if there is a wall, you stop, or if someone disengages, you slow down or shift. AI is a little more Sioux than agorytmic but it relies on a knowledge base and the AI is efficient to select the right action in 1millisecond without scanning the 800 terabytes of its base.
For the teachers it's a bit the same because if you look closely, they give back what they know with the added talent so that the student understands something.
But one of the great characteristics of AIs is that they are stupid if you haven't stuffed them with billions of cases. They are great restitutors with obviously after rearranging the information according to the prompt you did.
In the case of CAD it's different because you need to have an object knowledge base like you would have at grabcad. But if you ask him, "Imagine a machine that wraps chocolates and makes me all the parts that make it up, well rusk, it's a tench. Unless there is a company that has published all the plans for the latest machine which represents 6 years of R&D.
I haven't said enough already, I'm going to continue in a more structured way
@Scofield Can you give us some examples (tool in the design, optimization and calculations phase of all kinds, there is a real time saving and surprising results. for the drawings) ZZ
Hello, calculations of the weight of tear plate for example, calculations of the strength of gratings e.g.:
He knows the Solidworks manual quite well It also helps me to deepen the simulation and RDM, understand how the solver works... For supplier searches, sourcing... No need to go through web pages anymore, the bot does it for me...
You can even customize it, I put it in Jarvis mode and he greets me and calls me " Monsieur "
I didn't take the step of making him speak with a microphone like a real assistant but it's also possible...
I'm pretty intrigued by your use too. The cases of @Scofield seem quite interesting to practice... But I admit that I am quite helpless in front of the tool. The few times I tested it, I had the impression of doing a google search by keywords => which is quite far from what you seem to practice!
I agree on this point, it's much less stupid than google. Here is an example of a better efficiency of the A, which by the way is only an improved search system in this case. AI superior to google is explained by Google's initial funding system based on ranking and other algos.
Here is the question with a prompt that is a little stupid but which gives the expected result.
My little cat all clean! You know the object that looks like a power outlet that is always found to connect a workstation to the electrical outlet. Can you give me the reference on the net so that I can buy it
I took this example because two years ago I googled the reference for this type of plug and it took me quite a long time
GPT replies this in a few seconds To properly connect a workstation (PC, monitor, etc.) to a French wall outlet, the component you are looking for is often a mains power cord with a plug that complies with the French/European standard (CEE 7/7 or similar) on one side and an IEC plug (often C13) that fits into the equipment on the other. (IEC 60320 covers this type of IEC connector. Wikipedia)
Here are some examples of cables you can buy online:
@Tous To continue, I propose to reconstruct in the form of short texts, a bit like the summer soap opera, the result of my research and reflection on AI limited to the impact of AI on designers of complex machines. But also an update on the state of the art on what AI would be adapted to our professions.
I will post it to you as a PDF for those who wish to collect to have a global vision under the mouse.
Hello You have to discuss with the AI (basically conversational AI), giving it keywords such as google searches does not make it effective. Not particularly obvious at first glance but the result is not bad (but hey, you have to use it wisely and be careful if the AI is not validated by the company's IT not to put confidential data).
Not bad. What was your initial prompt to obtain this superb sizing result? Free version of chatgpt or subscription?
In search of new providers, AI is indeed not bad: it avoids having to go through 20 Google pages and click on 200 links.
I tried to extract data from pdf → excel via MS Copilot (a table that is on 4 pages of the pdf to be put in a single excel file: juxtaposition of the data of the 4 pages). Clean pdf with text (no image to pass in character recognition). The kind of shitty thing where you spend days on a repetitive job with no added value. The result was mixed: he did what I asked him to do by refining/refining the request 4 or 5 times. On the other hand, when I asked him to do the same on a new dataset, he completely went into a tailspin: infinite time and rotten results.
Personally, my use of AI (Perplexity and/or Le Chat de Mistral) is limited for the moment to adding didactic comments on the macros of my design or on macros uploaded here and there.
Considerable time saved when writing codes.
A significant advantage of not making typos or spelling mistakes (so much the better, it's not my specialty).
Allows a better understanding of macros retrieved from the WEB, or even the pure and simple elimination of some rather malicious codes... (And when you see what you can do with a simple macro, it's easier to take precautions.)
There was a time, not so long ago, when I tried to create (macro) tools from scratch with AI, the results remain extremely mixed.
Artificial intelligences still tend to hallucinate, even if it means creating functions that do not exist in the programming language used... and when an AI hallucinates we are confronted with a child of 3 to 6 years old who does not want to go to bed or brush his teeth, on the other hand he wants to watch the Snow Queen for the 12th time of the day... But sometimes there are some good surprises. Was it due to a better written prompt, a favorable alignment of the planets? It's hard to say (offer the same prompt to the same AI several weeks apart and it's likely that you'll get totally different results, sometimes better but not always).
Regarding the use of AI in our professions, we must not hide from the fact that we should not hide from the fact that we are not going to hide from the fact that we are doing so. We all already use Artificial Intelligence, only the form has become democratized.
A Google search? = AI (since 1998)
Calculation of the strength of materials in Solidworks = AI
Ben! Uh! CFD / Flow Simulation) Today no AI. It is a classical numerical solver based on the Navier–Stokes equations (finite volume method).
Calculation of the strength of materials (FEM/FEA) to date ditto, no AI. The calculation is performed via the finite element method (FEM), which is a deterministic solver. What may appear in the future will be for:
automatically detect critical areas,
offer smarter networks,
Accelerate repetitive resolutions. Currently, the core of the module is based solely on the EMF. so proportion of AI = 0%** today.
For google it's a little different! For " kernel " functions such as search: AI is now one of the pillars (not just an "option"). It is part of the core of the product for complex queries. For "assistant" functions (suggesting, completing, summarizing): AI is now very present, and is becoming more and more normal. But there are still areas at Google (some backends, or historical infrastructure) that are not purely AI but still rely on robust traditional systems.
To everything'
So, in the Google ecosystem, AI has a high proportion in new services/innovations, but not yet at 100% (some components remain classic in usual mode)
