Has AI been used yet for routing or GCA

[Jason Connor]

As artificial intelligence is taking off all over, I wondered if anyone has used (or admitted to using) AI for things like routing a course on a piece of land, or any other aspects of design.


Even if any of us non-architects have played around with it to see how it would layout courses.

[Charlie Goerges]

As artificial intelligence is taking off all over, I wondered if anyone has used (or admitted to using) AI for things like routing a course on a piece of land, or any other aspects of design.


Even if any of us non-architects have played around with it to see how it would layout courses.

I haven't heard of it, but we've discussed the logistics of it here before and I believe we came to the conclusion that something like that would eventually be possible, but bear in mind what AI actually is. Right now, on the text side, it just an extremely complicated predictive text. Additionally, it's only based on whatever inputs it has. I wonder how many topographic maps and correlated golf course routings have been fed into any of the graphic AIs? It's probably not a lot.


Someone speculated that one of the possible architectural uses would be for a user to input a topographic map of some existing or former course and an algorithm would search existing topographic maps to find as close a match as possible for the construction of a replica course. I could see that working, but it's not really AI.


I will say that I don't think we should waste any precious real-world sites on AI designs.

[Matt_Cohn]

Can somebody post a topographic map for a site (or a few) in this thread and I'll put them in ChatGPT with some guidance and see what happens?

[Tom_Doak]

Can somebody post a topographic map for a site (or a few) in this thread and I'll put them in ChatGPT with some guidance and see what happens?

Matt:  Sure.  But first send me all the data about your job so I can feed that into AI.


Seriously, though, why would any designer turn over the fun part of the design work to AI ?

[Joe Zucker]

Can somebody post a topographic map for a site (or a few) in this thread and I'll put them in ChatGPT with some guidance and see what happens?

When ChatGPT first came out, I asked it if it could read a topo map and if it could route a golf course with some guidelines. It answered yes, but I couldn't figure out how to get data into the AI engine.  It would be interesting if you gave it the topo maps and routings of specific architects and asked it to create something based off of this data. 

[Ben Hollerbach]

Seriously, though, why would any designer turn over the fun part of the design work to AI ?

What I could imagine is AI could be useful prior to a site visit to identify potential routing options for further exploration. While this is an activity the architect could do as well, the AI may present options not inherently noticed by the architect upon first exploration.

For example, one could "teach" the AI about MacDonald's Template holes by sharing with them topographic examples of those holes and then asking the AI to use that knowledge to locate potential template locations within a new site map.

While it's not doing the job for the architect, it may still provide a useful tool to the architect to test and explore options beyond their current practice.

[Andrew Harvie]

Coca-Cola used AI to make an advertisement about a month ago, and got flamed for it (rightfully so). Sports Illustrated had AI stories being written by fake journalists, and that didn't go over too well either. There are benefits to using AI in certain fields as an enhancer like spell check or organizational tools, but I think creative jobs are safe for the time being because we value the human touch, mistakes, and creativeness that comes with it. I would think a golf architect, and especially the art of routing, would fall into that.




We've seen the difference between architects who design in an office vs those who are in the field and the quality difference in the past 30 years, I don't see much benefit in removing the human touch entirely.

[Matt_Cohn]

Can somebody post a topographic map for a site (or a few) in this thread and I'll put them in ChatGPT with some guidance and see what happens?

Matt:  Sure.  But first send me all the data about your job so I can feed that into AI.

Tom, I'm not quite sure what you're getting at—I think the point of this exercise is that I'm NOT going to provide ChatGPT with your specific insights and expertise. I want to see what it can do on its own. (I know there are some issues with LLM's learning from copyrighted material, but that's a discussion for another place.)


It can be a hypothetical or already-constructed site, of course!


As a tutor and college admissions consultant, a lot of my job can be done by AI already, and in 5 years AI will probably do all of it better than I can. But I've learned a lot of ways to make AI work for me. For example, in just a minute or so I can get ChatGPT to generate a 20-question multiple choice quiz, with an answer key and explanations, on any topic and at an appropriate level for whatever my student might need help with. I might edit a couple of the questions, and then we can work through the whole thing together during our meeting to help the student study.


That wouldn't be so different than AI generating a dozen possible routings on a piece of land that you then review, pick a couple of favorites, and adjust slightly as you see fit.


If ChatGPT can't do that now because it's not sufficiently text-based, I'm sure AI generally will be able to do it a few years down the line. To answer your question about why a designer would turn over the fun part to AI: only if it produces better results, right?

[Matt_Cohn]

Can somebody post a topographic map for a site (or a few) in this thread and I'll put them in ChatGPT with some guidance and see what happens?

An initial attempt with a random topographic map I found online did not go well. 

[Josh Bills]

I find this discussion interesting as in my profession(s), attorney by day, painter by night, I have been advised that for the attorney part, "AI won't replace an attorney, but an attorney that uses AI will replace those who don't use AI."  I honestly struggle with seeing how it can acquire the 25 years of practicing law knowledge I have and be able to come up with the right solution for a client.  I am trying to be open to AI assisting my work, such as reviewing contracts or medical records, but for lack of better word, I don't "trust" that it will do the job as good or as thorough as I do the job. Maybe someday it will have a higher trust factor.   


As a painter, I make decisions all the time based upon my experience painting golf courses and I can't imagine AI doing it better.  I can understand the arguments, the thought it can come up with creative solutions, but I like to think creativity in law or art should be left to a living mind, not a computer. 


To bring it back to this topic, doesn't  there have to be some human component to know that whatever the AI generates has any possibility of being useful? Even the quizzes you talk about Matt require your review as I'm sure it is not fool proof. 


I think the idea is intriguing, but I get any creative persons hesitation to rely on AI. 

[Matt Schoolfield]

I can't think of a worse application of AI (I hate that name when it's really just applied machine learning).

Machine learning is very good at doing something that a really dumb person can do if they had an infinite amount of time to practice. It might be good at building template courses, but the point of good golf architecture, is in the novel application of offensive and defensive strategies, which, I mean maybe we'll get there, but we are not there yet.

I just cannot stress how dumb LLM's are right now. If you wanted a dumb golf course, I'm sure you could build a model that makes them. However, It would probably end up being worse than, what is probably an the algorithmic program, running in PGA Tour 2k23 when you have courses get automatically generated.

The intellectual and creative fields are the last place that AI will be applicable, if ever. Never in my life have I lived through such a hype bubble in a technical field that was so interesting, yet so many people just decided was indistinguishable from magic. You can study this stuff basically for free, and I think more people should!

Can somebody post a topographic map for a site (or a few) in this thread and I'll put them in ChatGPT with some guidance and see what happens?

An initial attempt with a random topographic map I found online did not go well. 

https://www.usgs.gov/programs/national-geospatial-program/topographic-maps

[Matt_Cohn]

To bring it back to this topic, doesn't  there have to be some human component to know that whatever the AI generates has any possibility of being useful? Even the quizzes you talk about Matt require your review as I'm sure it is not fool proof. 

Correct. That said, the rate of improvement is pretty remarkable. Stanford School of Humanities and Sciences says that GPT-4 (early 2024) passes a highly rigorous Turing test. That's unbelievable. [Note: obviously this does not mean it's as smart or capable as a human; I'm just stating how amazing that is.] Does anyone really feel certain that AI technology won't be better than you at your legal/financial/whatever job by, say, 2030? And if it can't route golf courses by 2030 (since I assume that's not a high development priority), then I'm sure it won't be far behind.


I just think any statement that "AI will never be able to do ___ as well as a person because the human element is irreplaceable" is really missing the reality of what it can already do and how rapidly development is proceeding.


Of course AI will be able to expertly route golf courses in the not-too-distant future.

[Tim_Weiman]

To bring it back to this topic, doesn't  there have to be some human component to know that whatever the AI generates has any possibility of being useful? Even the quizzes you talk about Matt require your review as I'm sure it is not fool proof. 

Correct. That said, the rate of improvement is pretty remarkable. Stanford School of Humanities and Sciences says that GPT-4 (early 2024) passes a highly rigorous Turing test. That's unbelievable. [Note: obviously this does not mean it's as smart or capable as a human; I'm just stating how amazing that is.] Does anyone really feel certain that AI technology won't be better than you at your legal/financial/whatever job by, say, 2030? And if it can't route golf courses by 2030 (since I assume that's not a high development priority), then I'm sure it won't be far behind.


I just think any statement that "AI will never be able to do ___ as well as a person because the human element is irreplaceable" is really missing the reality of what it can already do and how rapidly development is proceeding.


Of course AI will be able to expertly route golf courses in the not-too-distant future.

Matt,


I can’t help but ask: what does it mean to “expertly” route a golf course?


That aside, I think it would be interesting to take the topo of famous existing courses and see what the AI “expert” comes up with.


Tim

[Tom_Doak]

I just think any statement that "AI will never be able to do ___ as well as a person because the human element is irreplaceable" is really missing the reality of what it can already do and how rapidly development is proceeding.

Of course AI will be able to expertly route golf courses in the not-too-distant future.

Matt:


Of course, that's not what I said, and I know the above was not the reply to my post.


I just don't understand the point of AI in this instance.  As I said, the work of routing the golf holes is the FUN part of the job -- the creative part.  Why would I want to pass that off to an AI, any more than I want to pass it off to one of my associates?  The whole point of the exercise is, what do I find in the land?


And if you got a well-trained AI and it spat out ten different potential routings, who's going to determine which one to use?  You??  The client??  Or me?  [It won't be me, because I'm not going to use it.]


And if AI got good enough that I decided it was better than me at routing golf courses, what would I then do with my life?  Learn to code??  If our society is going to replace artists of all sorts with AI, what is left for humans?

[Matt Schoolfield]

I just think any statement that "AI will never be able to do ___ as well as a person because the human element is irreplaceable" is really missing the reality of what it can already do and how rapidly development is proceeding.

Again, while I agree with you that we should never say never (mainly because there is no reason to presume the human mind is particularly sacred), and I think it's perfectly sensible that we might find the model that produces Artificial General Intellegence, I think it's extremely important to reiterate that existing AI/LLMs are not magic, nor are they AGI. They are based on straightforward math, and LLMs are still very dumb because they don't actually have what we would consider a mind behind them.

https://humsci.stanford.edu/feature/study-finds-chatgpts-latest-bot-behaves-humans-only-better

https://www.pnas.org/doi/10.1073/pnas.2313925121

You're not going to find a bigger fan of Alan Turing than me (the movie they made about him was unfortunately absolute garbage and literally slander), but the Turing test itself is philosophically problematic. Firstly, there should be no defined beginning and ending of a Turing test, and any existing prescribed test can be obviously built around. It also ignores the philosophical problem of qualia and consciousness, though it is arguable that those might be illusions.

A real test would be literally any question. We know that Chat-GPT's LLM can't handle spacial locations very well, and it is easy to exploit the tokenization system it uses to get wrong answers, so it's not really close to passing the Turing test, however it is obviously, extremely impressive.

Again, it is not magic. It's a designed model. The model does things very well, but it's not anything approaching AGI, which is what everyone keeps conflating into the conversation with LLMs. That isn't to say that it couldn't happen one day, but even if it did, we don't actually know how much better it could create a golf course than a person. There are a lot of philosophical issues about the limits of intelligence that are just not known. I generally think it's reasonable to believe that the speed-limit of scientific knowledge is empirical research, not the human mind. That said, here's hoping (assuming it doesn't accidentally or on purpose, kill us all).

[Jason Connor]

I just don't understand the point of AI in this instance.  As I said, the work of routing the golf holes is the FUN part of the job -- the creative part.  Why would I want to pass that off to an AI, any more than I want to pass it off to one of my associates?  The whole point of the exercise is, what do I find in the land?

Tom, I'm your biggest fan and enormously grateful that you respond to this board.  I always find your posts education and informative.


And I 100% agree with you, artists want to do art.


I think the "why?" is because less talented & creative artists than you MIGHT be able to create the art that currently only you and a few others can.  Then because there are more such people, they can do it cheaper.


Or someone with a piece of land could use AI and not even require an architect, just an experienced builder.  (Numerous great courses, as we all know, were 1-off designs by amateur "architects")


I'm not saying it's right, I love and admire the art, that's why I'm here.  I want to discuss the ART of GCA.  That's why I own every book you've ever written.


But these are the risks / questions / things we need to be thinking about.


So I appreciate everyone's discussion.

[Simon Barrington]

Seriously, though, why would any designer turn over the fun part of the design work to AI ?

Absolutely. Same for Historical or Academic Research Tom; it is the hunt, the search, the tangential rabbit-holes that get you to answers...that's the fun. Why use an approximation engine that uses incomplete data to guess and fill in the gaps of our knowledge...it's the path to mediocrity.


It seems to me most humans are inherently lazy and always seeking approximation, homogenisation and shortcuts.


That is why those few (in whatever field) with genuine talent and who have spent incomparable hours absorbed by their vocation/calling/passion, fly the highest...long that may be the case

[zachary_car]

Seriously, though, why would any designer turn over the fun part of the design work to AI ?

Absolutely. Same for Historical or Academic Research Tom; it is the hunt, the search, the tangential rabbit-holes that get you to answers...that's the fun. Why use an approximation engine that uses incomplete data to guess and fill in the gaps of our knowledge...it's the path to mediocrity.


It seems to me most humans are inherently lazy and always seeking approximation, homogenisation and shortcuts.


That is why those few (in whatever field) with genuine talent and who have spent incomparable hours absorbed by their vocation/calling/passion, fly the highest...long that may be the case

I'm very much with Simon here, at least for now. What the future will bring in regards to AI, and its development, is a different story.


For the time being, however, AI has given artists a stopwatch, so to speak, against which they can be measured and have their growth tracked. Basically, a form of standardized testing for writers, musicians, etc.


It'll destroy the lower class of the various fields and, in turn, force the cream to rise to the top. Those who aren't good enough at their crafts will simply be rendered obsolete.

[Tom_Doak]

It'll destroy the lower class of the various fields and, in turn, force the cream to rise to the top. Those who aren't good enough at their crafts will simply be rendered obsolete.

I disagree.  I think it will empower IMG Design, or Charlie Woods, or some 15-year-old future golf pro who would otherwise be a mail-it-in mediocre designer.


That was the problem when I entered the field - all of the jobs went to ex-jocks and second-generation designers, just like in sports broadcasting.  We’ve finally gotten away from that, but really good AI would take us back in a flash - not in my generation, but in the next one.


Not to mention that building a nuclear power plant to power the thing is a lot more of an environmental concern than anything we’ve argued about in permitting golf courses!

[zachary_car]

I disagree.  I think it will empower IMG Design, or Charlie Woods, or some 15-year-old future golf pro who would otherwise be a mail-it-in mediocre designer.


That was the problem when I entered the field - all of the jobs went to ex-jocks and second-generation designers, just like in sports broadcasting.  We’ve finally gotten away from that, but really good AI would take us back in a flash - not in my generation, but in the next one.


Not to mention that building a nuclear power plant to power the thing is a lot more of an environmental concern than anything we’ve argued about in permitting golf courses!

We'll see. I tend to think that the really first-rate, creative human mind will always be able to offer an element that AI won't but maybe I'm just naïve in that stance

[jeffwarne]

Wait until AI starts following animal trails around a property....
UGH...


Zachary,
There were "first rate creative human minds" available in the 80's and 90's, but that didn't stop owners from hiring "jocks and second generation designers".
It took a long time for talent, skill and experience- not golf skill and family name recognition-to be the recognized as qualities that will produce a superior product.
A lot of well heeled potential future owners in last night's TGL audience probably thought that virtual set was clever design.

[zachary_car]

Wait until AI starts following animal trails around a property....
UGH...


Zachary,
There were "first rate creative human minds" available in the 80's and 90's, but that didn't stop owners from hiring "jocks and second generation designers".
It took a long time for talent, skill and experience- not golf skill and family name recognition-to be the recognized as qualities that will produce a superior product.
A lot of well heeled potential future owners in last night's TGL audience probably thought that virtual set was clever design.

I'm not a proponent of it, and it's going to replace me at my desk job within 5 years, but that's the reality that people in the creative arts are facing with AI. The artists whose productions will fall below the ever-increasing baseline set by AI art simply will not survive in the market. I wish that this wasn't the case but that's the reality we're facing going forth.

[Matt_Cohn]

Tom, I hope you'll enjoy this part of ChatGPT's response when I asked it to speculate about AI designed courses in the next 10-20 years:

Challenges and Limitations [of AI designing golf courses]

• Creative Vision
• • Golf course design is as much an art as it is a science. The "soul" of a course often reflects the vision and personality of the designer. AI might struggle to create the kind of unique identity associated with architects like Alister MacKenzie or Tom Doak.

Anyway, I enjoy taking a fairly extreme position in AI discussions of all types—in part because I think it might be true, and in part because it opens up some thought-provoking questions. "What if AI produces better art than humans do?" is a much more interesting question than "What if AI can never replicate what humans do?" I also note that I live San Francisco, where I'm practically surrounded by AI engineers and where I share the road with self-driving cars on my way to the gym or the grocery store every single day.

Basically, I look around and find it easy to believe that if AI can't do something now, it'll be able to not too long from now. And routings seem like a good place to start—an architect might consider 10 or 100 possibilities, certainly not a million like AI could. And I just think that ability—at least a high baseline, as Zachary refers to it—isn't so far away.

[Terry Lavin]

Can somebody post a topographic map for a site (or a few) in this thread and I'll put them in ChatGPT with some guidance and see what happens?

Matt:  Sure.  But first send me all the data about your job so I can feed that into AI.


Seriously, though, why would any designer turn over the fun part of the design work to AI ?




It’s probably due to LOFT, meaning Lack Of Fucking Talent!  I can’t imagine that I would respect a design that relied on AI.

[jeffwarne]

Wait until AI starts following animal trails around a property....
UGH...


Zachary,
There were "first rate creative human minds" available in the 80's and 90's, but that didn't stop owners from hiring "jocks and second generation designers".
It took a long time for talent, skill and experience- not golf skill and family name recognition-to be the recognized as qualities that will produce a superior product.
A lot of well heeled potential future owners in last night's TGL audience probably thought that virtual set was clever design.
[/q


I'm not a proponent of it, and it's going to replace me at my desk job within 5 years, but that's the reality that people in the creative arts are facing with AI. The artists whose productions will fall below the ever-increasing baseline set by AI art simply will not survive in the market. I wish that this wasn't the case but that's the reality we're facing going forth.

Interesting-thanks.


I guess we better destroy all copies of Tom Doak's "Getting to 18" before A1 absorbs it