AI that scans a construction site can spot when things are falling behind(technologyreview.com)
technologyreview.com
AI that scans a construction site can spot when things are falling behind
https://www.technologyreview.com/2020/10/16/1010617/ai-image-recognition-construction-computer-vision-costs-delays/
14 comments
taylorism, but with artificial intelligence! $13 million series A
From my experience in construction, the works are usually divied up into packages or 'claim cards'. Then each supervisors are supposed to report how complete each package under their control is.
Problem is, this can sometimes be a little tricky and the supervisors often fudge the numbers to smooth out how their works look. Ie, they may have done 50% of the work, but will only claim 20% because they know that the last 50% worth of the claim card will actually take way longer. They don't want to look bad at the end of each week cause the estimators did a shit job. So they'll take their gains and average them out.
A system that was more automatic and had way more compliance built in would be cool to see. I doubt however, that an algorithm could solve what is predominately an issue of people and estimation.
Problem is, this can sometimes be a little tricky and the supervisors often fudge the numbers to smooth out how their works look. Ie, they may have done 50% of the work, but will only claim 20% because they know that the last 50% worth of the claim card will actually take way longer. They don't want to look bad at the end of each week cause the estimators did a shit job. So they'll take their gains and average them out.
A system that was more automatic and had way more compliance built in would be cool to see. I doubt however, that an algorithm could solve what is predominately an issue of people and estimation.
If 50% takes longer than another 50%, the metric by which estimation is performed is, at least regarding time, flawed.
If a subcontractor were to report, "hey boss, we've gotten 8/10 items done this week, but the remaining two are going to take two weeks because $REASONS. The estimates appear to be flawed, perhaps because $X and $Y.", I suspect the entire project's management would benefit.
If a subcontractor were to report, "hey boss, we've gotten 8/10 items done this week, but the remaining two are going to take two weeks because $REASONS. The estimates appear to be flawed, perhaps because $X and $Y.", I suspect the entire project's management would benefit.
I suspect the opposite. The managers will start to panic and then micromanage everything, have crisis meetings etc. This will slow it down even further.
The only real solution to this problem is better long term estimates! The thing is that all estimates are partly BS because we haven't figured out how to predict the future yet. So the solution above is that the supervisor got a BS estimate and they are pushing back with BS progress. There probably is no other way.
Good managers will take the opportunity to learn more. Either there is a way to aid those who made the estimates, a way to help the people doing the work to get the job done, a need to relax the schedule, or some combination of them.
The goal is to get the work done on time and with the resources at hand. Micromanagement is often a hindrance to the goal, so good managers avoid it.
The goal is to get the work done on time and with the resources at hand. Micromanagement is often a hindrance to the goal, so good managers avoid it.
Good construction managers often go from project to project with no tie to a particular company. Same with most of the craft.
They're interested in surviving this project. There seems to be fuck all continuity out there.
They're interested in surviving this project. There seems to be fuck all continuity out there.
You've just identified why a system that's more automatic is risky to naively trust.
Here's my prediction: the AI is going to see a lot of jobs that progress linearly to completion and then stop dead at 90%. That having been said, it'll still be useful if metrics can be sampled out to guesstimate how much padding is needed to account for the 90% dead-stop time.
Here's my prediction: the AI is going to see a lot of jobs that progress linearly to completion and then stop dead at 90%. That having been said, it'll still be useful if metrics can be sampled out to guesstimate how much padding is needed to account for the 90% dead-stop time.
wanted to say this. "fall behind" is the deadline some quantity surveyor and job-quote expert expressed. The actual on-the-job work rate is a function of weather, concrete mix, unforseen post-design issues, pour quality..
Yes, this is hugely informative in a $million cost space, but it is also basically "work harder: we said you could"
Yes, this is hugely informative in a $million cost space, but it is also basically "work harder: we said you could"
Try managing a construction project sometime.
You're a DBA.
All the work done on my previous timber home, the delay came from uncovering the unexpected. I had mentally allowed 10% overrun on cost and time. We blew both.
Estimation is hard. I only do peripheral work in s/w estimation but my life-long experience is that its just hard, and even on million dollar projects, cost blowouts are common. I'm amazed how the london tube expansion, or the new sewerage went, all things considered.
Brisbane's cross-river rail project is 2x price and late btw.
Estimation is hard. I only do peripheral work in s/w estimation but my life-long experience is that its just hard, and even on million dollar projects, cost blowouts are common. I'm amazed how the london tube expansion, or the new sewerage went, all things considered.
Brisbane's cross-river rail project is 2x price and late btw.
Usually in my experience it's like any other jobbed out project -- all about the contractor and subs being overcommitted and unable to react.
I've had the good fortune to participate in projects where stuff really had to get done and cash was flowing. So a mixture of well-defined scopes, significant performance bonuses and non-performance penalties were implemented, and somehow the downstream contractors were magically able to respond to the unexpected.
I've had the good fortune to participate in projects where stuff really had to get done and cash was flowing. So a mixture of well-defined scopes, significant performance bonuses and non-performance penalties were implemented, and somehow the downstream contractors were magically able to respond to the unexpected.
Absolutely. Adding a "stopwatch" to make humans more efficient is having negative effects on it's employees.
The article is quick to say people won't lose their job, but that's not the only fear about automation. Employees hated taylorism because of the reduced autonomy, unrealistic expectations of consistency, pseudo-scientific increased workloads and sometimes mind-numbing boredom.
The article is quick to say people won't lose their job, but that's not the only fear about automation. Employees hated taylorism because of the reduced autonomy, unrealistic expectations of consistency, pseudo-scientific increased workloads and sometimes mind-numbing boredom.
Just wanted to call out a company already active in this space called reconstruct:
https://www.reconstructinc.com/
Their software is pretty cool, came out of a research group I was working in. Originally, we were using drones to scan the site every day and building a point cloud. I know much of the research team went to the company (I was only working on the scanning system).
Pretty sweet stuff and honestly one of the best uses for a drone I’ve seen.
https://www.reconstructinc.com/
Their software is pretty cool, came out of a research group I was working in. Originally, we were using drones to scan the site every day and building a point cloud. I know much of the research team went to the company (I was only working on the scanning system).
Pretty sweet stuff and honestly one of the best uses for a drone I’ve seen.
There must be a better way. I work on a project that does lidar scanning now to track construction progress and its a beast. Keeping up with the latest data is a huge problem (and this is not a tech problem, its a people problem). Issues are raised that are not really issues and so forth due to ever changing data in a project.
Image scanning is a stepping stone to true project tracking which I am hoping to work on next! Wish I could say more.
Image scanning is a stepping stone to true project tracking which I am hoping to work on next! Wish I could say more.
Gathering more than one progress data set separately is the only way to get a single data set. If you try a single measurement or combine them to early you will get [many] layers of people making stuff up and the further up you go the more nonsensical the data gets.
If someone has only half (or less) of the data to forge his lies out of the creative book keeping window is almost gone. Employees with 10-20 years of experience generate lies of undetectable mind blowing quality. I've seen people negotiate their way out of 100% of their job arguing they didn't have time for it.
If someone has only half (or less) of the data to forge his lies out of the creative book keeping window is almost gone. Employees with 10-20 years of experience generate lies of undetectable mind blowing quality. I've seen people negotiate their way out of 100% of their job arguing they didn't have time for it.
This is so true and absolutely correct. The issue though is not gathering our data (progress) its comparing it against they're engineering drawing etc.. that are constantly changing on a file share that we never know about.
You want it geo-referenced? Good luck.. so you end up putting a bunch of people on the team working for them to get decent data which bloats your budget and then you look bad to client.
You want it geo-referenced? Good luck.. so you end up putting a bunch of people on the team working for them to get decent data which bloats your budget and then you look bad to client.
Have to find cheap measuring points.
Does it tell me when shoddy work is being done to stay in the time frame? Does it measure the concrete mixture or check that the rebar spacing and type are correct for that type of building?
Yeah. My impression about construction is that we don't need more schedule micromanagement. We need something to put the metaphorical (and/or literal) fear of God back into the hearts of every subcontractor and entire supply chain. I can't imagine that many corners being cut during construction of bridges and cathedrals in medieval times.
In my experience the opposite is true in construction today, at least where I live.
Everyone lives in fear of something with more wrath than a god. Liability.
Because of this they are not willing to do anything even slightly outside of code, and any decision that needs making is a phone call to the engineer rather than making the call on the job.
This slows everyone down as nobody is keen to actually be the one with responsibility. The engineeer who feels the same way also make sure that everything is overspecced to all hell to make extra cirtain that they won’t be liable for anything failing.
Everyone lives in fear of something with more wrath than a god. Liability.
Because of this they are not willing to do anything even slightly outside of code, and any decision that needs making is a phone call to the engineer rather than making the call on the job.
This slows everyone down as nobody is keen to actually be the one with responsibility. The engineeer who feels the same way also make sure that everything is overspecced to all hell to make extra cirtain that they won’t be liable for anything failing.
Well, but it is part of the "requirements imposed by society". For example, a few years ago in my city a supermarket roof caved in during reconstruction, killing 54 people; one cause of the event was an after-project underspecced change to a single key component, and the responsibility for that fact was quite important (I believe the civil engineer who signed off that change was sentenced to 6 years prison for manslaughter) - and arguably we do want everyone involved to consider that responsibility and not just "make the call" that seems sensible without going into detail, since not going into detail kills people.
Are you saying that's a bad thing? The code exists for mostly good reasons. I'm more inclined to do things myself, but if I have to hire someone, I damn well want them to be "not willing to do anything even slightly outside of code".
I'm sure plenty of corners were cut back then too... Just the buildings where corners were cut probably don't exist today. Survivorship bias...
I think some of it is just luck in that there has always been someone fixing the roof and a fire keeping damp and frost away.
> I can't imagine that many corners being cut during construction of bridges and cathedrals in medieval times.
You can't imagine it because those bridges and cathedrals collapsed (and were probably rebuilt).
You can't imagine it because those bridges and cathedrals collapsed (and were probably rebuilt).
We do have a bit of selection bias since structures still standing were likely well built while those poorly built are simply gone.
You might be surprised, and horrified, to find out a not insignificant number of still standing buildings were built with no foundations. On a wing and a prayer, if you will.
https://thefogwatch.com/the-miracle-of-salisbury-cathedral/
https://thefogwatch.com/the-miracle-of-salisbury-cathedral/
Came across a house a year or so ago. The back half of the house was supported by an old tree stump. The person living there said the kitchen had an odd slope in the middle of it :)
I worked on a house where the previously built extension to the rear had been built over a soil pipe that let out into the ground, over the ~20 years since it had been built, it had tipped further and further and started to pull away from the house as the ground was eroded until we had to come in and knock it.
nice! I had one where I failed an inspection. Inspector comes out and signs off on the new job. "hey isnt this YOUR signature from the first time? Why did you sign off on it then if it is so wrong now?"
I sense Goodhart's law coming in to play eventually once people learn the metrics that contribute to the the input of the "AI".
Aren't these metrics increasingly more difficult to determine?
I think people will have no idea what the AI evaluates but they will make myths about it and adhere to those (which effectively triggers Goodhart's law). IMO a big chunk of SEO and content marketing is already based on myths around PageRank/YouTube/…
I think people will have no idea what the AI evaluates but they will make myths about it and adhere to those (which effectively triggers Goodhart's law). IMO a big chunk of SEO and content marketing is already based on myths around PageRank/YouTube/…
Manna ?
> There was the facade of “free elections,” but only candidates supported by the rich could ever get on the ballot. The government was completely controlled by the rich, as were the robotic security forces, the military and the intelligence organizations. American democracy had morphed into a third world dictatorship ruled by the wealthy elite.
Fact or fiction?
Fact or fiction?
That's what I thought of as well. I first read this short story a very long time ago and it's still something I think of often. For those who aren't acquainted, I think it's worth a read: https://marshallbrain.com/manna1
The ai could be
function isConstructionRunningBehind(project){ return true; }
function isConstructionRunningBehind(project){ return true; }
Can't wait for the boss firing me to be an AI!
I'm pretty sure this is already happening to Amazon and eBay sellers and gig economy workers that depend on Uber, Postmates, etc for their livelihood.
Sure, they may not be technically hired, but it comes with the same life devastating effect!
Sure, they may not be technically hired, but it comes with the same life devastating effect!
People seem to forget that the ending of Azimov's "I, Robot" is painted as a good scenario. ;)
This seems to be an example of management paying a lot of money for some expensive third party to come in, and tell them something that all the line employees already know.
I vastly prefer things falling behind to things falling
I can see the benefits of monitoring progress like this for a good project manager (who isn't micro-controlling or forcing a schedule, but who can keep the stake-holders at bay with good progress reports and early warnings about a schedule adjustment).
On the other side, an easier to implement and more worrisome idea (for me): AI that monitors your git repo to report you when you aren't getting things done quickly...
On the other side, an easier to implement and more worrisome idea (for me): AI that monitors your git repo to report you when you aren't getting things done quickly...
Probably triggers on workers drinking coffee.
I feel like McDonald's would benefit more from this.
I find it hard to believe AI knowing what's happening on a complex construction site and offer constructive ideas. Just a human looking at a daily photo should do.
McDonald's is 40,000 stores, all similar, 360 days of the year, multiple shifts.
It's just time until the DOM Pizza Checker gets lifted 2 meters and gets an update.
I find it hard to believe AI knowing what's happening on a complex construction site and offer constructive ideas. Just a human looking at a daily photo should do.
McDonald's is 40,000 stores, all similar, 360 days of the year, multiple shifts.
It's just time until the DOM Pizza Checker gets lifted 2 meters and gets an update.
Might be an interesting article if I could get through the popups.
It would be like an automated inspector that is up to date on the latest building codes at all time and is available 24/7