Over the past few months, we have read a growing number of articles describing the increasing use of artificial intelligence in CAD and BIM (c.f. 1, 2, 3). Designers and engineers are asking where the division of labor lies, in a world where access to magic wands, genies in lamps, and even Tony Starks (with all the help caring Iron Man constantly provided to Spider-Man) is pretty much universal. AI or human. Which way is the relevant way?
Let us illustrate the issue with examples from familiar professions. Skilled teachers, translators, and journalists sense a rise in the number of people around them who are accessing AI and then sincerely mistake the labour of others for their own. The AI copiers deem themselves worthy of the same rewards as original thinkers.
The problem is whether people should receive equal pay and recognition for either
- translating text using their foreign language knowledge, or using an AI-powered translation
- preparing lectures based on their own learning, or lecturing students from text generated by large language models
- writing blog posts or articles based on their own research, or publishing text generated by AI
People might not notice the significant reduction AI provides to their efforts, while the compensation for such work seems, for now, to remain the same.
Wisdom From Fairy Tales
The story unfolding around us does not differ too much from those of classic fairy tales. There is a vast number of fishermen who have wives, and yet the tale "The Fisherman and His Wife" offers exactly one fisherman who comes across exactly one magic fish capable of fulfilling his wishes.
In another fable, The Wishing Table (found in French and German folk tales) turns out to be an object that feeds people, but can only be activated by those who know the magic words (a.k.a. modern-day AI prompts). Which raises a question: How many fairy-tale characters in today’s AI world know the magic words?
We are not against AI; we use it ourselves regularly and have no intention of stopping. We understand, for instance, that after writing an article, it can be a good idea to ask our favourite AI to spot weaknesses in the text. It can catch typos and more significant flaws.
But exercising the brain is important. For instance, automated spell-checking saves time and effort. Deliberately turning it off can become a good stimulant to concentrate on the text and be attentive to splng misteaks. The unaided approach maintains – and even raises – the bar for creators who dare take such a step, bringing benefits to them over the longer run.
So why do we make comparisons with folk tales? It’s because they are treasuries of human wisdom, some centuries-old. In many tales, the hero at some point is forced to realise that he is not the architect of his own good fortune. It’s not entirely his efforts that seat him on the throne, or grant him untold wealth, or free his future bride from the dragon. All of these actions are the result of wondrous tools that he used.
His contribution is his desire to apply a seemingly-randomly-acquired artefact to carry out his own ends. After that, the miracle object must be guarded from theft or misuse. As it turns out, in these tales the instrument is stolen, the hero has an epiphany, and heroically wins it back.
Let's return to today. Have modern Cinderellas faced the fact that the original fairy godmother and her miracles are no longer available to them? The new replacement, AI, is more expensive, behaves differently, and remembers nothing of what happened yesterday, such as at the prince's ball. Has it occurred to us that the magic table could refuse to prepare the dishes it made yesterday (or did the inn keeper perhaps steal the magic table and switch it with a regular table?) — even after the eldest son took deposits and promised everyone meals by evening? Well, yes.
For now, juggling questions between AI models goes smoothly, and switching from one solution to another is more or less manageable. As in our fairy tales, everything goes along briskly and cheerfully – for a time. We are like the main characters who continue to believe, almost all the way through, that the text generated, as we delicately put it, "with the help of AI," is the fruit of our labor. We might not even be bothered to read AI’s riposte before publishing it – we have become ‘meat proxies’.
We’d like to conclude this assortment of moderately incorrect comparisons between apples and oranges with these points:
The development, optimisation, and application of AI is undoubtedly an extremely important endeavour. By the way, some of our team at LEDAS began at a university AI lab that already existed in 1999.
From the standpoint of a final result, there is nothing wrong with an AI user is rewarded as before, while thanks to AI grows accustomed to exerting far less effort. It could be that millions of years of evolution have taught us to strive precisely for this. After all, those who preferred to exert more effort for the same outcome simply died out, according to the theory.
We, however, see potential problems:
- The magic friend who grants us wishes is available to not just one protagonist, but to hundreds of millions. This means AI inflation and slop.
- This new friend can be taken away at any moment. This will be painful for those who have grown accustomed to seeing themselves authors, yet are unable to produce on their own what they have been successfully "producing" for months with the aid of AI.
To conclude, we want to recommend a couple of articles by Martin Day. In "Autodesk's journey to AI", he describes how Autodesk's leadership gradually began to articulate the company's strategy for integrating artificial intelligence into its AEC and BIM software. In his analyses of the emerging strategy, Day speculates what it might mean for the future of BIM and CAD. He covers a spectrum of possibilities, from AI building fundamental geometric models to a new concept called "project intelligence."
In "Dawn of the Judgement Stack", he asserts that AI is becoming another critical component of BIM software. This is evident from the fact that all major BIM vendors are striving to create increasingly sophisticated AI assistants and pilots.
Day argues that when CAD firms focus exclusively on such features, they may, however, miss the bigger picture. He believes that the future of AI in the AEC industry may depend less on the vendors themselves, and more on their clients – the design firms who are turning their decades of experience into assets that they may well deploy with in-house AI systems, independently of the AI added to CAD software.
The wisdom that age-old fables hold for us moderns is that we ought not make ourselves too dependent on the convenience of new technology. Dependence could make us weaker, especially when it could even be taken away from us.