Three of our senior architects at Lang, Shelley & Associates, Adele, Beyonce and Cher, are working on Projects A, B, & C, respectively. None of the architects has any coding expertise.
Each project can be broken down into 10,000 discrete steps, in a particular order.
All three projects have the same steps.
For this exercise, we’re going to hit the middle of the task chain, and examine the following sequence of steps:
·
Step
3,443: Go
through the BIM Model and tag any items that are manufactured more than 500
miles from the site
·
Step 3,444: Generate
a csv list of all found items, along with their tag number
·
Step 3,445: Generate
a list of possible substitutions that are manufactured within 500 miles of the site
·
Step 3,446: Verify
that the possible substitutions comport with the specifications previously
written for that item
·
Step 3,447: Run
a price check on all possible substitutions.
·
Step 3,448: Create
an Excel file and sum the price differentials of all possible substitutions
called "Net Value of All Substitutions"
·
Step 3,449: If
the Net Value of All Substitutions is positive (i.e.
substituting all products would represent a net contract add, subtract items
from the list that represent a cost increase until the value of the Net Value
of Substitutions is =< zero.
·
Step 3,450: Create
a separate Excel file "Proposed Changes" of all subtracted items
detailing their name, their SKU number, their price, the net increase over the
previously specified item, their distance to the site, the projected carbon
cost of transport, and the net savings on embodied carbon.
·
Step 3,451: Create
a third Excel file "No Cost Changes" of all non-subtracted items
detailing their name, their SKU number, their price, the net increase over the
previously specified item, their distance to the site, the projected carbon
cost of transport, and the net savings on embodied carbon.
·
Step 3,452: Replace
any No Cost Changes in the BIM model with their locally sourced equivalents,
log those changes as a no-cost change.
·
Step 3,453: Draft
an email to the client, proposing that the subtracted items in "Proposed
Changes" be substituted, even if it represents a cost increase, to bring
the project into line with the client's ambitious environmental goals. Describe how the proposed substitutions fit
within the client's stated goals, and indicate that
you have already made substitutions on the non-subtracted items. Detail the projected carbon savings of those substitutions, and emphasize that they were done at zero net
cost to the project.
·
Step 3,454: If
response is positive, replace any "Proposed
Changes" in the BIM Model with their locally sourced equivalent, and
prepare a change order for the contractor.
Last Thursday, Adele received permission from her client to make some changes to the design of Project A. She had suggested a few substitutions, because certain products were manufactured far from the site, and she thought that client’s environmental goals could be better served if all products were manufactured within 500 miles of the site. Adele now must go through the BIM model and manually make those substitutions, and then prepare a change order for the contractor (Step 3,454). She doesn’t feel like doing Step 3,454, so she asks her Autonomous AI “AutoAdele” to do it for her. She merely asks it, in natural language, to go through the BIM model and make the substitutions based on the itemized list that she had included in her email to the client, and then subsequently prepare a change order for the client’s signature, using the firm’s standard template. AutoAdele executes the task perfectly, and logs the required steps, and required code, into LSA’s central bank of “MiniAIs” as MiniAI 3454, so that it can be used in the future.
On Friday, Beyonce, who is working on a different project, on a schedule slightly behind Project A, is trying to work out a different task. She has similar environmental values as Adele, and asks her Autonomous AI “AutoBeyonce” to go through the BIM model and identify any products manufactured more than 500 miles from the site, generate a csv file of all those items, and then generate a list of possible substitutions that comport with the specification previously written for those items (Steps 3,443 through 3,446). AutoBeyonce executes these tasks, and logs the required steps, and required code, into LSA’s central bank of “MiniAIs” as MiniAI 3443_3446, so that it can be used in the future.
The following Monday, during the weekly LSA staff meeting,
Adele and Beyonce share their creations during the regular “AIs I Made Last
Week” portion of the meeting. Cher
is stoked to hear this news, since she’s coming up on a similar phase in her
Project C. On a previous project, she
had “AutoCher” develop a MiniAI to execute Step 3,447
“Run a Price Check on all Possible Substitutions” and knows that it’s
still in LSA’s MiniAI central bank.
Sensing an opportunity, she directs her Autonomous AI “AutoCher” to ‘connect the dots’ and develop a MiniAI to
complete Steps 3443 through 3454, utilizing MiniAI 3443_3446, MiniAI
3447, and MiniAI 3454.
At the next staff meeting, Cher relates to the LSA staff that she has built MiniAI 3443_3454, to much applause. Going forward, those steps can be automated by anyone, and anyone’s Autonomous, Personal AI can recruit MiniAI 3443_3454 as part of an even larger process.
A few important observations about MiniAI 3443_3454:
1. It’s not project specific. It can execute its task list on any BIM model, for any project.
2. It’s not mandatory. In this example, we envisioned the exact same steps, executed in the exact same order, for 3 different projects. In any real world example, it’s unlikely that any two projects have exactly the same number of steps, or the same steps, or the same order. But that’s beside the point. The staff at LSA (or your firm) are constantly generating new MiniAi’s to facilitate their work, and as long as those are stored in an accessible format, Autonomous AIs will be able to recruit and deploy them in different combinations, different sequences, etc.
The process of creating MiniAI 3443_3454 should sound familiar: it’s exactly how we currently create code, macros, parametric modeling techniques, etc. The ubiquitous Grasshopper works on a similar, albeit visual logic. The difference here is that MiniAis are being generated by code-illiterate staff, with hardly any effort at all.