Client: We are completely unable to plan a construction project successfully. We want you to use AI to do all of our project planning for us. Our requirements are that instead of needing to spend any money or time planning we just want to press a button and have a computer instantly put together all of our project plans for us. The program also needs to identify optimizations on it’s own and change all related plans enterprise-wide. All copies of the plans should be able to be kept up to date at all times so we’re never looking at an old plan again. We also want the ability to print.

Dev: …

“We want a program that will do our most laborious work for us with little to no effort on our part.”

@AmyShackles isn't that the point of automation tho
They're not wrong, just way, way under budget

@RememberMe ... except that AI isn't clairvoyant, nor omnipotent.

The idea to let a machine do project planning is in my opinion insane... Especially with AI.

@IntrusionCM Especially if a person can’t even get it right. Like what would it be measuring “right” or “correct” against? Especially since most of their project planning issues have to do with misinterpretation of requirements or constantly changing requirements not missing or incomplete ones.

@RememberMe Automation is great for repetitive tasks that have a finite number of variations and whose complexity is known. Automation shouldn’t really be used for things that involve strategy, have infinite variation and complexity, and need to take the will and action of humans into account. An unbounded problem is not a good candidate for automation.

Edited to add: … in my opinion.

Totally agree with Amy ..they just hear the term AI and think it’s the panacea rather than the fad which will come and go…buffoons of the highest order!!!

"We want to lay off all our project planners but it turns out their job is really important. Can a robot do it for free?"

@IntrusionCM I was talking about the "do the hard work" part, not project planning in general. And tbh you can already get systems to remove a lot of the cruft (supply chains, operations, etc) if you're willing to shell out. Since it's a construction project, all that would be a thing.

@AmyShackles "strategy" is a very wide term. A lot of strategy and decision problems are optimization problems under sufficient constraints. You'd be surprised how well optimization can do for supposedly unconstrained problems, doubly so for ML based AI where the structure of said optimization is not clear and needs to be learned from the data.

You also may not be able to do everything with a given AI system, but you definitely can get the same/similar/better job done with a small team of experts/decision makers + AI. I don't think the intent is to literally built a one-button system anyway, for every automated system you need some human intervention and oversight as it is anyway, that's a standard assumption.

Let's take a lambda step, put some faith in my boy @boombodies and imagine that he did import god.h and managed to create an automatic project planner that knows its shit.
Imagine what happens when the client is about to use it for the first time. They bark some unintelible client-speak into an alexa-thingy, some fairy makes P=NP and voila! the software spits out some beautiful charts and sheets planning a whole operation.

The next day the client comes back PISSED saying the magic software does not works, because it deprioritized "the most important thing! I've set it to top priority!" - and also for every single piece of the request.
Or that the software canceled "an very important feature!" that contradicts two others.
Or that the software "makes plans that are too slow to implement!" because it assumes that people will work no more than 24h per day.

My point is, give a client a fish and they will complain about it for a day. Give them a fishing net and they will starve to death.

@JsonBoa That is brilliant!

@JsonBoa Fucking amazing, should be its own post