AI for engineers
Y Combinator just put out a call for more AI tools for engineers. Can and will AI replace engineers? In the near term that’s pretty unlikely for a litany of reasons – engineering work is not a low hanging fruit like other more formulaic work, for many engineering tasks poor or no datasets exist to train ai, there are too many conflicting constraints to be balanced. Can AI be an efficacy tool for engineers? Absolutely!
Need to summarize or find something in a 300 page standard?
ChatGPT, Claude, Sharly and many other tools can do that.
Need to find something across multiple standards?
ChatGPT and others excel at this and can give citations to make sure the answer is the correct one. Here’s a great example for human factors – https://chatgpt.com/g/g-nIgwIPFbx-ergogpt I have most of these standards, but some are in paper format while others are in PDF. This tool searches them all in moments making a hours long task only minutes.
Need to write a complex Excel equation?
Akkio is free with no login
Generative design
Various CAD packages have had generative-like FEA features for years. These tools when applied to structural FEA would remove material where it is least efficient in carrying the given loads yielding a minimal weight part. Usually the output part would not be manufacturable so some material would be added for DFM to cast and CNC such a part. With metal 3DP these parts can be realized as optimized!
Autodesk has taken this further than ever with more tools and configurations to make the process faster and more effective.
Startups like Neural Concept are working on even more robust solutions. Imagine making a multi-week design process take days? One can hope.
Document authoring
With the proper training data AIs like Google’s Gemini and others can generate similar documents. For simple form letters proof reading may not be required, however with more complex documents a knowledgeable human is still required to be in the loop.
Industrial Design and Product Design Assistance
Tools like Leo and others can help ideate design and, in the case of Leo, write product documentation. The outputs of these process take multiple iterations to improve and give somewhat superficial results, but at the same time getting to this level without AI assistance does take time and AI makes this faster instead of replacing the designer per se.
Limitations
Like any tool, AI has its limits. Understanding those limits is important. Asking it to do too much in one go will probably lead to a poor answer. Break the problem down and/or don’t ask for too much. Also answers need to be double checked. We’ve all heard the stories of “AI” making up answers or references. From all the tools I’ve listed above one can conclude that there is no killer app or single tool for all problems. This splintering of the AI engineering market will continue with multiple tools required to complete holistic projects.
Security Concerns
AI engines like IBM’s Watson are built around enterprise use. Other engines like Claude are privacy focused though OpenAI with ChatGPT-4 would argue that they are ready for enterprise use too.