He says AI is really bad in counting bolts in a technical drawing. Whole point is, that if you do a 200 bugs AutoCAD weekend course you will learn that AutoCAD does that - without any AI. It just keeps count on how often the engineer placed a bolt into the drawing.
Maybe we all leave the "stealing" to the Chinese and their past who are now overtaking Europa having copied to understand and learn and evolve really, which is the only use case of an AI counting bolts shortly after answering: What does that do?
RAGed Ollama models are extremely useful for engineers. An Engineering Office should look for an Open Source AI developer that RAGs an Ollama model dedicated to mathematical tasks with all relevant books and texts related to the main field the Office focuses on from Civil Engineering Bridge Design and Wind Tunnel Behavior to 3D Toy Printing about 3D materials and their attributes.
To be honest, if there is a Bubble of PhDs that collect money for selling a pipeline to an online LLM, that is blunt fraud.
The local RAGed AI than is an Assistant that goes through all related mathematical data in form of a script creating a prompt for interactions having a dedicated folder with e.g. .txt files that are exports of the spreadsheet like calculations to reason about those.
That will be extremely specific questions depending on the very project and thereby environment. The Engineering Team could ask to change all statical values by changing the Aluminum used or mounting technique.
It will struggle if while RAGing the mathematical model information was missing like on Biomorphing in Joint Construction or Japanese Timber Joint creations, but excel in walking through impact heavy changes and making suggestions to improve characteristics based on specific questions.
I wonder if the Bubble will be an Investment Fraud Bubble by some people jumping on a boat they did not know about, except big money.
The Industry, like OpenAI will monitor the Open Source development along with seeking dedicated solutions that must be DIN normable, like a bolt, being standardized. An Outlook Psychological Analyzer would be one to filter for Angry or Supportive email tones giving advice on how to reply easening tensions. That is no problem for the Open Source Companies, because they naturally must focus on tailoring, it will just make investing into AI solutions a simpler decision.
Not if ... all newspapers are full of a new generation of White Collar Criminals filling up Prison Gangs.
Microsoft will have an entirely different approach onto creating that PSY-AI than Open Source developers. They will build a Data Center, which run on Linux already anyway rather than on Windows NT or consider Unix depending on which computer coding language will be used to drive the very large Ollama model being a copy of DeepSeek or ChatGPD and its clones facing millions of requests per day.
There is a great chance Microsoft will go for C or C++ and Unix creating deeply OS embedded ultra stable software combined with load balancing software as they do for their Cloud offers like Drive or Gmail.
The training is then in opposite to Open Source the simple part. They buy the Harvard Library, or parts of it, and use books that have no Read, but a Borrow or Buy button avoiding all future Copy Right issues by size-of-cash while some guys like me eventually will buy a Hand-Scanner with to .txt converter Foo Fighting Metallica 2Pacs we all are for real and its the Amsterdam Public Library for a reason than.
The rest is rather simple being a new button or prompt in Outlook telling you about Mr Jones going soon ballistic, but being also able to point out nuances in his wording and making educated suggestions which power phrases to avoid for example and what to do, like no excuses make a clear announcement such as "It is broken. We can't fix it. We exchange it. Send it back to u"s or "We call you tomorrow at 09:00", asking you to send an email to FedEx for a pick up or make a calendar note.
You don't want them to answer emails. Trust me. No matter Ray.
#cyberpunkcoltoure
