artificial intelligence

[SE-Bologna]

Hello, everyone.
probably already reading the title you will have come the orchite since the topic is now omnipresent.
In spite of this I would like to open a small discussion here, in my opinion the theme will be central in the coming years also for mechanical design.

For example, I believe that the boarding (the paper will not disappear shortly from the workshops) will soon be completely automatic and there will only be a check.

do you have news of first cad applications where is implemented seriously?
for now I have found only generative design applications (already present for years) spread/recycled as applications to, which are not.

what projects have siemens, dassault and ptc?
How do you think our work will change?

peace and good

 

[falonef]

Hi.

if you are looking for this on the web should jump out some article, video about it

 

[Steppermotor]

I don't know what future developments and the impact on our work will be.
from that little I could read about how the ai works the only application that I think would be really useful (as long as there is not already) is the search in a pdm similar parts for shape and size.
this to avoid the proliferation of equal parts on different groups in environments like mine where there are thousands and thousands of components and many years of modeling in the db.
Plus it could help you find quick solutions without banging maybe to solve a problem already faced.
Maybe I said all the crap. . .

 

[re_solidworks]

Hello, everyone.
probably already reading the title you will have come the orchite since the topic is now omnipresent.
In spite of this I would like to open a small discussion here, in my opinion the theme will be central in the coming years also for mechanical design.

For example, I believe that the boarding (the paper will not disappear shortly from the workshops) will soon be completely automatic and there will only be a check.

do you have news of first cad applications where is implemented seriously?
for now I have found only generative design applications (already present for years) spread/recycled as applications to, which are not.

what projects have siemens, dassault and ptc?
How do you think our work will change?

peace and good

I'm not very hard on the subject to, but since I understood it works by looking for info on the web and putting them together quickly. considering that on the web of free data does not find itself, I do not see how theai could work. could be implemented in companies with an important database, but I still see it hard. Moreover, the noble design is not "copying" existing solutions, but creating new more economical, safe and efficient. Can you do that? I think she needs to evolve a little earlier. perhaps in the short term will help those who produce series products becoming a kind of more evolved configurator.

 

[Steppermotor]

Well maybe then I explained badly.. .
my reasoning was based on the fact that already in administrative areas it seems to be used to do repetitive things and compare data already in possession of the corporate db.
in our reality we do not need to look for solutions on the internet but avoid duplicating parts and solutions already present in the company that with the passing of the years and the change of designers remain dormant, knowing that they exist maybe they could still be useful.
design from scratch then is another thing... that there is not to which I keep and sometimes not even title of study or master blasonati or courses "hyper teaching". .

 

[stevie]

design from scratch then is another thing... that there is not to which I keep and sometimes not even title of study or master blasonati or courses "hyper teaching". .

We'll talk in a few years.

 

[Steppermotor]

I hope I'm already retired. . .

 

[SE-Bologna]

@falonef , on the web every day they jump out new things but it is all mixed (for now) in a great cauldron. I posted here to get some specific info about mechanics.
for example ptc:

https://www.ptc.com/it/blogs/cad/will-chatgpt-ai-revolutionize-engineering-and-product-development

but it is a fairly generalist article and perhaps put there because something must say, they can no longer ignore the subject.
@re_solidworks @steppermotor the aai is based on deep learning algorithms that analyze large amounts of data.
I make a banal example: a software cad manufacturer could recover 100 thousand (the numbers must always be great) models and drawings of biella associated with real results of breaking tests, fatigue and reliability. at this point could be able, given some generic input specifications, to generate model and design of the best possible biella associated with these specifications.
complicating everything you could get the whole engine. It is obvious that then you use it as a work draft, going to improve, but 80-90 percent of the work in terms of time is already done.

Another example could be to have 100 thousand drawings of trees and automatically produce the tree table that I am modeling. at the limit I might have to change some tolerance but in general the table will be fine to 90%

Hi.

 

[SE-Bologna]

in practice the help should start from the "low" part of the work, not from the creative one (type the generative design).
I add this quote read recently (I do not find the source)
"I want you to help me do the washing machine and clean house, not compose music or paint that are the things I would like to do in my spare time"

 

[PIETRO2002]

I think there are developments that we can't even imagine now, which will impact our professional and private lives. For example, in the medical field now the doctor makes a diagnosis with his knowledge, tomorrow he will be diagnosed by helping himself with the help, so having so many data and experiences available, logically the human relationship between doctor and patient will remain of fundamental importance. In our field, choosing and calculating a mechanical organ for a particular application will be much easier and intuitive, even if "for now", the designer's creativity will still have its importance, but I wrote "for now". . .

 

[TETRASTORE]

I add this quote read recently (I do not find the source)
"I want you to help me do the washing machine and clean house, not compose music or paint that are the things I would like to do in my spare time"

linedin

 

[TETRASTORE]

theai is evolving very quickly in all sectors and will revolutionize our working system.
Until now it has been often associated with the chatgpt note which, thanks to a notional system, offers more or less consistent answers by analyzing the large amount of data in its possession and, in the technical design field, its contribution can be summarized by a description of autodesk: " today, thanks to self-desk you can improve the exploration and creative resolution of problems, automate boring and repetitive activities and provide individual users and teams access to powerful analysis tools with which to make informed decisions".
but the true waiting frontier can be generating design proposal from ptc that states: "generating design is a 3d cad feature that uses artificial intelligence to create optimal projects autonomously from a set of system design requirements. Designer engineers can interactively specify their requirements and objectives, including favorite materials and production processes, and the generation engine will automatically realize a project ready for production as an initial point or final solution. As a result, engineers can interact with technology to create higher quality projects and innovative products more quickly".
In practice, taking the example of a reducer (product I know a little better), my expectation would be to provide most of the application data and I would eventually like the complete project already calculated, verified and designed in addition to the cycles and processing programs related to my means of production. Of course, during this journey I expect also alternative proposals that I will evaluate. utopia? no, all this would already be possible but, of course, those who work at these innovative solutions will tend to release them gradually to mutize them adequately, a little as they do with the mobiles for which, the solutions already available to you add a little over the years to every new version of the device.
ptc says that with the generational design also technicians with a reduced experience will manage complex projects but, in my opinion, for many years the experience and preparation of technicians will be indispensable to obtain excellent results working less.
It will be a different speech of those who are now confined to subordinate activities such as a particularist of the projects of others who, especially if he is young, it is necessary to review his professional future because in the course of a decade will be overcome by technology.
on this I do not see a negative impact but I see new opportunities to grasp, it is only about beginning to understand and adapt to a new vision of work and specializations, a little what happened when the era of computerization began a few decades ago.

 

[lucacort]

seems quite relevant to me:

gNucleus : Generate FreeCAD models by GenAI - FreeCAD Forum

 

[ultratech]

I think it will be a great revolution.

if he talks a lot about it since he entered chat_gpt in the limelight at the end 2022.I think it will change and variety the type of professional figure required.
I believe that they will implement modules on how to draw brackets and small objects.

interesting links but become dated very early in this field every month there are great variations every month:

Generative Artificial Intelligence Software Tools cad 3d - Convergence Advice​

Generating AI software tools are a new generation of cad 3d software that uses artificial intelligence (ai) to help engineers design better products faster.

convercon.com


Thank you.

 

[francescodue]

I do not know much about the situation, but I would like to express my opinion.
given a set of precise rules (e.g. on the table) you can train one to the even with smaller amounts of data. This would mean that medium/gross companies can build their own tailor-made ia.
a possible alternative would be that the regulatory bodies receive this training work.
or, for example, professional orders or equivalents.
with regard to the use on the design of components, I see it more as a "comparator" of solutions: given for example a biella (to reconnect to the above mentioned above) and inserting all the parameters required, theia should present to me the different solutions that favor the individual "aspects": standardization of measures rather than resistance or instead greater applicable force or ease of manufacture (I am not a mechanic but I hope I have made the concept understand).
in any case the designer will evaluate which solution to use and refine it until the final result.
in any case, it would be mandatory that before submitting the results the message was always shown "these are all wrong projects, they must be checked"; This is because, at the moment, it seems too easy to get wrong results even if they seem sensible.
in the various media we often talk about Marche errors of the ia, so obvious to be absurd; But the real danger lies in the small details that escape.

 

[Fulvio Romano]

in practice the help should start from the "low" part of the work, not from the creative one (type the generative design).
I add this quote read recently (I do not find the source)
"I want you to help me do the washing machine and clean house, not compose music or paint that are the things I would like to do in my spare time"

Well, we'd miss it. robotics (in the broad sense) is servile work (from the Czechoslovak "robota") that helps man, not that replaces man.

I think there are developments that we can't even imagine now, which will impact our professional and private lives. For example, in the medical field now the doctor makes a diagnosis with his knowledge, tomorrow he will be diagnosed by helping himself with the help, so having a lot of data and experiences available.[...]

Already today the diagnostics made by the aisle is a few percentage points higher than that made by human doctors. but we talk about diagnostics for images, that is to compare a situation with other analogues. five milliseconds of aisle contain more than ten years experience than a doctor, until we play in the field of the aisle I win you. medicine is not only diagnostics. it would be like to say that an ultrasound machine is better than a midwife because the midwife without an ultrasound is wrong to understand the sex of the unborn child.

I remember that already fifteen years ago an interesting study of google showed that google himself could predict an epidemic a week before the local hospitals. banally because people look for symptoms on google about a week before going to the doctor. but this is not medicine, it is data analysis.

Until now it has been often associated with the chatgpt note which, thanks to a notional system, offers more or less consistent answers by analyzing the large amount of data in its possession and, in technical design [...]

this comment is very sensible. in daily design we do a whole series of repetitive and value-free operations. We have to do it, there is no justification that I keep. being able to give it to a software would be beautiful, and it will soon be. but how to diagnose for images is a micro part of the medicine, so this part is a micro part of the design.

I spend two words on chatgpt which simply does statistical analysis on the probability that a sequence of words corresponds to expectations. If you ask him a question about a well-treated topic on the internet, the answers are incredibly fast and centered. but try to ask if man has ever been on the moon, or if someone tries to control our minds by spreading aerosols in the atmosphere, or even news about the sphericalness of the earth. two things happen, first of all repeating the same question several times you get diametrically opposed answers second at that moment what is more preponderant and recent on the internet. the second is that the internet is so full of wild theories that chatgpt can easily deduce things that violate the principles of physics and common sense.
But it doesn't end here. try asking questions about niche sectors where information is objectively scarce. chatgpt begins to "invent" non-existent facts in order to converge their statistical algorithms. I happened yesterday to ask him to write me a small post on a very particular subject, and it brought me out that the first object was designed and made by a well-known engineer. I said, "I was convinced that I was prepared on this subject, and yet I never heard of this news." I tried to investigate before I published. There is no trace of that event anywhere on the net. I tried to ask the same question and that "historical fact" is banally gone.

there is another important thing to know about the generative ones. they generate synthesis from vast amounts of data. the result is typically used to write texts that then go on the internet. the second generation therefore will be found to synthesize data from synthesis already carried out from some other to. This creates a very dangerous drift because the amount of data increases, but the information in this data becomes increasingly lost, making the above described phenomena more and more frequent, increasingly insidious and difficult to discover.
There are techniques, called civetta patterns, which are inserted for example in visual generatives. this is not done today in chatterbots, i.e. text generators.
Eye. .

 

[infallibile_GF]

Artificial intelligence. a term that today goes very fashionable, has a not know what science fiction, almost magical. And yet what is imaginary artificial intelligence? It would be nothing more than that thing that until yesterday they had algorithm, but it didn't do well, it wasn't cool. algortimo. the other one yesterday instead we called it software and the other still yesterday was called very generic computer.
what we call artificial intelligence today is nothing less than the natural evolution of computer programs and systems. It's not an invention of the last year. It will now be 30 years and even more than, without we knew, we use software based on aisles.
I don't know I could make a stupid exempt in the automotive field: to the early 2000s, i.e. 20 years ago, the first electronic control automatic changes arrived on the cars with a function of self-learning. that they were adjusted according to the driver's style and were able to interpret and also to predict the pilot behaviour. i.e. they knew whether the driver was driving in a sporty way, or it was quiet and adapted the subsequent changes. Today a change like this would call it an automatic change with the rtificial intelligence. 20 years ago, she simply changed her mind.
tornanado to mechanical design and applications of ai to mechanical design. I believe that times are not mature at all, because in this sector precision is everything, indeed absolute precision is fundamental. wrong a tolerance, a share, can cause damage to thousands of euros, or in some cases, even damage to things or people. Therefore, a system that works a bit at the carlona as the current ai is not admissible. Yes, of course one remains impressed by the immiges generated by the aisles, which seem true photographs and then looking better turns out that he created people with three legs, or with two right feet, or strfalcioni of this genus. all gross errors, assulutamen inadmissible in the engineering field.
then a fundamental requirement for all systems ai is to have access to exterminate databases, from which to extrapolate the required results. in the case of our mobile phones, search engines, etc.. systems have easy play in having billions of data of all of us available to process the results, but in the field of mechanical design that you do? allows these software to have access to business servers, all companies globally, so without any control? so they can pull out a nice project, maybe copied who knows from which company, maybe even covered in abrevetto and meanwhile the projects, calculations, data of my company will make happy a Chinese company that thanks to the will have copied my ideas? I see it unenforceable for the moment and also in the short and medium term.

 

[Vittorio]

ai It scares me a bit, but it is an inevitable process like all the processes that have brought us the compact technologies of our daily life.
even with the first cad there were those who fought to remain clinging to the tecnigraph.

so by curiosity I asked gpt-4 what he thought of himself.
answers:
positive aspects:efficiency and speed: the ia can manage large volumes of data and complex tasks with a speed and efficiency that exceed the human ones.precision: reduces human errors in critical fields such as medicine and engineering.Automation: frees people from repetitive tasks, allowing them to focus on more meaningful activities.decision support:provides advanced analysis to help in complex decisions.customization: improves user experience by customizing services and products.innovation:stimulates progress in areas such as health, energy and the environment.Accessibility:makes technology more accessible to people with different skills.
negative aspects:unemployment: automation can lead to loss of jobs, especially in less specialized sectors.bias:if not properly managed, ia algorithms can perpetuate existing prejudices.privacy: the collection and analysis of personal data may threaten individual privacy.safety: vulnerable ia systems can be exploited for malicious purposes.dependence:excessive reliability over the age can reduce human decision-making capabilities.Ethics:the use of theia in sensitive contexts raises important ethical issues.

 

[falonef]

... so they can pull out a nice project, maybe copied who knows from which company, maybe even covered in abrevetto and meanwhile the projects, calculations, data of my company will make happy a Chinese company that thanks to the will have copied my ideas? I see it unenforceable for the moment and also in the short and medium term.

this already happens today even without the ...

 

[PIETRO2002]

ai It scares me a bit, but it is an inevitable process like all the processes that have brought us the compact technologies of our daily life.
even with the first cad there were those who fought to remain clinging to the tecnigraph.

so by curiosity I asked gpt-4 what he thought of himself.
answers:
positive aspects:efficiency and speed:the ia can manage large volumes of data and complex tasks with a speed and efficiency that exceed the human ones.precision:reduces human errors in critical fields such as medicine and engineering.Automation:frees people from repetitive tasks, allowing them to focus on more meaningful activities.decision support:provides advanced analysis to help in complex decisions.customization:improves user experience by customizing services and products.innovation:stimulates progress in areas such as health, energy and the environment.Accessibility:makes technology more accessible to people with different skills.
negative aspects:unemployment:automation can lead to loss of jobs, especially in less specialized sectors.bias:if not properly managed, ia algorithms can perpetuate existing prejudices.privacy:the collection and analysis of personal data may threaten individual privacy.safety:vulnerable ia systems can be exploited for malicious purposes.dependence:excessive reliability over the age can reduce human decision-making capabilities.Ethics:the use of theia in sensitive contexts raises important ethical issues.

It's a bit like nuclear energy, you can use it to produce energy and development, or you can use it to destroy the planet. I hope that this decision will be taken by men and not by a car, although sincerely, I have not so much confidence in world leaders lately!