The Cult of G.E.N.I.U.S. Pt. 1
A Speculative Exploration on the Possibility of a Full-Stack, AI Architect
Part 1: Introduction
In this Post[1]:
Why Design an A.I. to Replace Yourself?
A.I.’s Tumultuous Present
A Few Rules
A Few Assumptions
A Few Preemptive FAQs
Why Design an A.I. to Replace Yourself?
This series describes a speculative process for replacing 95-99% of the work currently done by human architects with artificial intelligence – in a sense, replacing myself, and most of my friends. The protocol described in this series serves as an extension of the previous technical addendum to the Design Intelligence Quarterly article “In the Future, Everyone’s an Architect (and why that’s a good thing)” parts 1 and 2. I broke off this particular explanation because it was more speculative in nature. The article and the processes described in the article focused on replicating the role of an architect during the earliest parts of the design cycle, since those seemed most challenging to replicate. Here, we’ll go further.
This series is an attempt to investigate the entire design cycle (DD’s, CD’s, CA, etc.) and how AI could displace the work of architects in the future (or in the present).
There are parts of this series, which will be released in 10 parts, that human architects may find insulting, enraging, and technologically preposterous. That’s part of the reason I wrote it. Architecture has such a dismally poor record of keeping up with technology, often embracing the latest technologies from 30 years ago and celebrating itself as revolutionary for doing so. In contrast to previous technological revolutions, Natural Language, Generative AI (NLGAI) represents a unique threat to an architect’s livelihood. We must adapt, and quickly. The biggest risk factor in architects’ possible displacement is the fact that they believe they can’t be replaced. I spoke about this at length in the article, so won’t rehash all those thoughts here.
“You gotta let reality hit you in the face”
-Greg Brockman, cofounder of Open AI
In truth, I don’t believe that architects will be replaced. But I think their job description will change significantly. A current architecture student is unlikely to ever run a firm of 100 people. They will likely run a firm by themselves, supervising 20 to 30 Intelligent Digital Architects (IDAs) who execute most of the work that used to be done by human architects, while doing the same volume of work that used to be done by a 100-person firm. It sounds fantastical, but all technology does, until it becomes ubiquitous. It has always allowed one person in the present to do the work of a dozen people of the past. Architects are surrounded by Low-FTE/High-Output firms, and most don’t know it. Midjourney AI only has a staff of 11, Rhino has a team of 80, and David Rutten basically developed Grasshopper by himself. And they did all that without AI. In the future, with conversational AI as a keystone technology, every architect will be a CEO. I’ll expand on that claim in a future post, already being drafted.
A.I.’s Tumultuous Present
Before we get to that abundant future, we have to go through a tumultuous present. So let’s begin, and see how we can replace architects with A.I. Since so many of my friends are architects, it may seem insensitive to design a digital replacement for architects. But I provide one here for the same reasons I wrote the article. If we properly anticipate the onset of technological innovation, we have an opportunity to adapt to it.
I love architecture. More for its promise than its delivery, though. And I believe that the release and sudden global embrace of GPT[2] is a technological moment where architecture can shift its focus to more meaningful problems. I’ve spent most of my career working at nonprofits and on humanitarian issues, and trying to convince other architects to do the same. But those arguments are hard to make, since nonprofit architecture pays even worse than regular architecture, and architects are already notoriously underpaid.
Natural Language, Generative AI (“NLGAI”) platforms like Chat GPT, DALL-E and Midjourney could make things worse, and lead to a catastrophic unemployment crisis in architecture.
I believe that the best way to hedge against this outcome is an immediate reinvestigation of architecture’s core value propositions – what we do, and why it is important. And I believe that if we put our heads together, we can think up some things to do (other than what we do now) that will be more lucrative, better for society, and more fun than our current job description.
For that reason, and as part of my due diligence in writing the article, I crafted a plan for (almost) entirely replacing an architect with Artificial Intelligence. In these pages, I’ll use myself as an example, and show you how I would replace myself, if it were me. I’m planning to release this protocol in ten parts, on the following schedule, and hope you’ll tune in for all of it:
4/26: Part 1: Introduction
4/26: Part 2: Setting up G.E.N.I.U.S.
4/26: Part 3: Firm ‘Strategy’
4/28: Part 4: Schematic Design
4/28: Part 5: Preparing a Schematic Design Presentation Package
5/01: Part 6: Design Development
5/01: Part 7: Construction Documents
5/03: Part 8: Bidding & Negotiation
5/05: Part 9: Construction Administration
5/05: Part 10: Conclusion
We’ll begin by imagining that I am currently Founder and CEO of a 25-person architecture firm called ‘Lang, Shelley & Associates.’ (LSA) I’ll walk you through how I would design a full-stack, Generative AI architect called “G.E.N.I.U.S.” (Generative Environment for Natural Intelligent Urban Solutions) that will allow me to quit my day job, in a manner of speaking. G.E.N.I.U.S. is going to handle the day-to-day practice of Lang, Shelley & Associates so that I have more time to solve all the world’s humanitarian problems, which is how I would prefer to spend my day. We’re going to create G.E.N.I.U.S. with the process described below – a process which would not entirely exclude humans. You would still need a human to oversee the process and ensure the result, which was a central conclusion of my article. But it should cut back the human workload at Lang, Shelley & Associates by 90 – 99%, I think.
A few rules:
I only included software that either exists at the date of publication, or is currently scheduled for release. I excluded all speculative technologies.
I’ve included the specific names (and links) of different software platforms to ground Point No. 1 in some reality that the reader can verify for his or herself. This should not be taken to mean that that software is the only software for that role. The AI space is quite dynamic, and depending on when you read this, some of the software platforms listed below may be out of date, superseded, or discontinued for other reasons. The software platforms should be understood as examples only, and not definitive endorsements.
I have not used all of the software platforms described in this piece, for differing reasons. In most cases, it was merely that they cost money and I did not have a budget for this project.
A few assumptions:
We’re going to assume that all computers and programs in the near future will work off of natural language processing. Even in the six short months since Chat GPT was introduced to the world, we have seen this trend explode. It was briefly mentioned in the article, but to me, the most revolutionary aspect of NLGAI is that it provides a common means of communication for all software.
We’re going to assume that all industries are automating at a roughly similar rate. However fast (or slow) we imagine architecture automating, we will imagine that engineering, construction, and the world at large are automating at the same pace.
The AI field, for the near future, will be an evolving, contested space, similar to the 90’s dotcom boom. With any new technology, we can expect a flurry of startups, mergers, acquisitions, busts, etc. It is unclear how all this will shake out, and will remain so for a while.
A few preemptive FAQs:
Q: Is this real?
A: Yes, it’s very real. All technology described in this series is either available today or scheduled for release by Fall of 2023. There is no imaginative sci-fi navel gazing. There is no Rosey the Robot. What’s described in this series isn’t Vision, or Ultron. But it may be Jarvis.
Q: Will I lose my job?
A: Maybe. Maybe not. Your job will certainly change, though.
Q: What should I do to prepare for this future?
A: That’s a contextual question. But I think the first thing to do is to educate yourself about what’s real, and what’s just hype. After that, call me.
I hope that frames the conversation adequately. Read on to Part 2: Setting up G.E.N.I.U.S. to begin setting up our artificial architect.
[1] Reading Notes:
1. References to ‘the article’ refer to “In the Future, Everyone’s an Architect (and why that’s a good thing)” parts 1 and 2, available on Design Intelligence. If you haven’t read them, you should probably begin by doing so, as they give important context to what follows.
2. References to the ‘technical addendum’ refer to the previous technical addendum to “In the Future, Everyone’s an Architect (and why that’s a good thing” which was written to provide a detailed record of how the original AI architect/client exchange was developed.
3. References to ‘the video’ refer to the video included in “In the Future, Everyone’s an Architect (and why that’s a good thing),” which is also available on my youtube.
4. I developed the initial language model on March 12th, 2023 and developed the video in the days after. I submitted the article for publication on April 7th, 2023. The article was published on April 26th, 2023. The speculations are current as of date of publication, however, I expect them to be rapidly outmoded. I’ll be updating my substack periodically to track new developments at the intersection of AI and architecture, but will leave these articles in their original form.
[2] “GPT” refers to both Chat GPT (Chat GPT 3) and GPT 4 since both were used in the construction of the model. Chat GPT and GPT 4 have different capabilities, and wherever the use of one was specific, it is noted as such. Wherever “GPT” is used, it could refer to either Chat GPT or GPT 4.