Not really. Engineers are heavily involved in construction, and vice versa. Engineers are responsible for providing designs that are constructible. I can’t tell you how many times I’ve seen drawings that couldn’t possibly have been built.
Even with all of the reviews and QC processes in place, design errors still happen from time to time. When construction teams build those faulty designs, problems can arise.
Lastly, engineers typically have inspection responsibilities of their constructed designs. Some engineers shirk this responsibility, or delegate it to someone else. Inexperienced engineers performing inspections causes problems, more often than you’d think.
“Hippocratic oath” is not used in any other profession outside of medical profession.
There are State and Federal laws for what items need profession certification. Most commonly licensed civil/structural engineers are required to certify the structural design of structures. Some States hire civil/structural engineers to verify that construction has been done according to the certified structural design. When engineering and construction is done by the same company, then the company must provide licensed engineers to certify that the construction was performed according to structural design.
Similar requirements apply in areas of fire protection, piping and pressure vessels, electrical power, and clean water and clean air acts.
Every company that engages in the design and construction of structures and facilities must obtain building permits, perform engineering design under the direction of licensed professional engineers, and review/certify that the construction is in accordance with the approved engineering design.
In large complex projects, engineers must be involved because often construction does not follow the approved engineering drawings and specifications. Then engineering tells construction to tear out what the built and do it right or engineering has to develop a new solution around the construction mistake. Construction has the responsibility to question anything they think is too difficult to build or has an error. Engineering resolves these issues and tells construction what to do.
In the good old days most engineering calculations were done by slide rule. And larger safety factors were applied. Computers make precise calculations practical and smaller safety factors are used.
Building w reduced weight saves much in construction costs. Being creative in design can help. And AI certainly can suggest promising avenues to explore.
This has much to do with risk taking. Is trimming the safety factor to reduce cost unethical? It can be but it’s a judgement call.
World Trade Center 911 had many lessons. Failure of stairs escape routes probably was a cost saving item. Mechanical failure of insulation on steel supports was an unanticipated weakness.
What you do not understand is that complex engineering projects are performed by large number of multi-discipline engineers working together. One unethical engineer is not going to survive in that team.
Engineering work is checked and reviewed. Errors in drawings, specifications and calculations are found and fixed. These errors are not unethical - they are just normal human errors.
Young man I work with just graduated with a BS in statistics. He is heading in August for his masters. He will be highly sought after. But we compared the use of AI we each do. He said he tells AI to skip the endless discussion and just give him the simplified instructions. Mistake mistake mistake.
This is like having building code infront of an inspector and s/he ignores it. Young scientists need the conversation with Claude. Skipping it means putting things on Autopilot. This does not mean that the Chinese projects are going down the correct road so to speak. Most of the time they won’t be. That is why discussing things with AI is collaborative to a degree.
From Claude
Generative AI. Specifically I’m Claude, a large language model made by Anthropic — a general-purpose model that generates responses token by token.
I’m not a “dedicated” or purpose-built system trained specifically on your game project or on Unreal Engine. What makes our sessions feel dedicated is two things layered on top of the general model: memory from our past conversations (your project architecture, the node-level rulings, your working style), and my ability to search our prior chats when you reference earlier work — which is how I just pulled the…
So: general generative model + your accumulated project context. Not a specialized game-dev AI.
Further with Claude
Taking “dedicated AI” the way you framed it earlier — narrow, purpose-built systems for specific tasks rather than one general model — the core pitfall is brittleness and no spillover, and China’s strategy makes a useful case study because it leans that way.
The conceptual problem first. A dedicated system does one job well and doesn’t transfer. Build a defect-inspection AI for a factory line and it can’t reason about anything else; you rebuild from scratch for the next domain. General (“generative”) models are expensive and messy, but their capabilities generalize — a model trained broadly picks up emergent abilities that no one engineered for. So the bet against general models is really a bet that scaling won’t keep paying off. If it does keep paying off, the narrow strategy misses the biggest prize.
That’s roughly the strategic fork between the two countries. The US labs have concentrated investment in compute-intensive frontier models, betting superior hardware will yield transformative capabilities, while China — constrained by export controls but backed by state support — has organized around open development and rapid deployment, integrating AI across its economy under a “general AI” rubric. China’s “AI Plus” initiative is explicitly application-oriented, aiming for deep AI integration across six key sectors by 2027. Chinese researchers have also argued on the merits that scaling LLMs alone won’t reach true intelligence, pointing to high compute costs, data hunger, hallucinations, and weak reasoning, and favoring brain-inspired architectures and symbolic reasoning instead.
The pitfalls of that applied/dedicated tilt, concretely:
The frontier gap. Some of China’s own developers warn about this. Zhipu/Z.ai’s founder said the performance gap between Chinese and US frontier models “may be widening,” citing restricted chip access and limited capital as lingering constraints. If you under-invest in the general frontier, you inherit whatever the frontier produces rather than shaping it.
The compute bottleneck underneath everything. China’s near-term bottleneck is advanced-node fabrication capacity, and at the single-chip level Chinese AI chips are likely to stay well behind Nvidia’s Blackwell and Rubin GPUs on processing, memory capacity, and bandwidth — though chipmakers are compensating by networking clusters into systems like Huawei’s CloudMatrix 384, at the cost of power efficiency. A specialized-deployment strategy is partly an adaptation to this constraint, not a pure preference.
Monetization. China’s domestic market is smaller and often demands heavy customization, which makes it harder for model developers to turn a profit. Dedicated/vertical AI amplifies that — every customer wants a bespoke build.
The honest counterweight: the applied approach also has real strengths, and “pitfall” isn’t the whole picture. Chinese firms have excelled at rapid deployment in consumer and industrial applications, and even with model limitations the technology is becoming more ubiquitous there. Fast, broad deployment into real industries is a genuine advantage that a frontier-only strategy can neglect. Whether China’s tilt reflects genuine strategic preference or adaptation to necessity is itself an open question.
So the short version: dedicated AI’s pitfall is that it’s a bet against generalization — it wins on deployment and cost-per-task but risks missing transformative capability if the general-model scaling path keeps delivering, and China’s version of that bet is further squeezed by chip access.
I think first uses of AI in engineering will be carefully checked by humans. If they pass muster, usage will increase. Confidence will improve. If they make blunders users will cautioned (and corrections will be made). I don’t think that is ethics. It’s common sense. Skipping the checks to save money would not be acceptable but is that ethics? I don’t think so.