AI in Architecture and a Brief Description of GANs-AI of the Future

The Problem

If anyone has ever invested in the building of a home or thought of doing so, they understand the dilemma faced by millions of Americans: the costs to carry out a project of this magnitude are simply too high. No longer can new-home-hopefuls expect to keep an existing home, a car, and maintain the payroll of a builder and an architect without feeling great financial stress.

QuickenLoans estimates that the average cost of building a new home is roughly $150,000. This cost alone amounts to about THREE YEARS of an average American’s salary!

It’s simply not feasible for Americans to be expected to pay this expense.

THIS NEEDS TO CHANGE.

Can New Tech, Like AI Really Change This?

We have all seen the power of even early-stage generative AI solutions. Intuitive computing is a driving force in the development of influential AI, and its applications seem truly limitless. The structure under which many of these intuitive solutions reside is known by the formal distinction of Generative Adversarial Network (GAN).

Welcome to GAN 101

**I will try my hardest not to bore you with the many specific mathematical factors that are included, and are formative in this technology**

The math behind most machine learning algorithms relies heavily on the concept of derivatives.The basis for the GAN mathematical algorithm also includes a generative model known as the ‘generator’ being pitted against an adversary known as the discriminative model. Collectively these two entities work against each other with the purpose of improving the algorithm to create the most optimal output given the training and parameters.

This form of neural network has been transformational in the creation of generative solutions, and has numerous projects displaying the true power of its utilization.

Some examples of recent GAN applications:

The Benefit of Computer-Led Intuition

Humans naturally function under certain limitations of thought. We create solutions by understanding what is being asked of us, and connecting this asking to we have previously thought as a feasible set of solutions.

Although computers have gained the ability create intuition, often, there is a beneficial falling-short of human-like behavior. The more practical explanation: when computers think in a purely black and white fashion, they are able to make informed decisions without a limiting bias. This allows for systems to explore all possible solutions given a set of parameters. Often these solutions are terribly irregular, yet just as often, terribly efficient. They push the standards of large and small scale manufacturing and create new norms.

In addition to the examples shown above, we have seen this advancement at a profound level in the creation of virtual car chassises.

The Story of the Virtual Chassis

After taking a test car loaded with various sensors to the desert, driving, and making a variety of turns, Mourice Conti and his team of AI researchers were able to capture millions of data intricate data points representing specific instances of movement or change.

With these data points, the researchers were able to define the positive and negative aspects of certain car action. These defined sets of data points allowed for the researchers to add thousands of parameters prior to running the algorithm, to optimize results.

The visual results from this experiment were more intricate and efficient designs then previous estimates imagined. In addition to this difference, and similar to the results of other AI research, the computer-generated chassises have drastically differing designs than industry standard. Unlike most large-scale early auto manufacturing plants, this form of slightly supervised(for lack of better terms) GAN has the ability to both take in, and fully-comprehend large pools of intricate, technical analysis, and individually create something meaningful.

Opportunity Presented in Architecture

With this new form of innovation in AI, jobs like those done by architects and designers have the potential to be fully automated. Imagine an unsupervised GAN algorithm that takes input in the form of blueprints, and after recognizing certain similarities between these blueprints, could create its own, similar depiction. The ability to create the most optimal home on a personal computer revolutionary. One is able to avoid the great financial obstacle of paid design/architecture. Imagine the level of opportunity that arises when one is able to create plans for a home or a to-be-built structure via the internet or on-phone applications. The role-out of technology like this is something that will be done in the next 3–5 years, and its widespread adoption will be formative in the development of the architectural industry.

Takeaways:

• The added costs of buying a home are far to much for the average American to pay while maintaining financial stability.
• General Adversarial Networks(GANs)- the broad umbrella under which most generative solutions fall.
• The adversarial aspect of this network title derive directly from the concept of adversaries with the GAN’s algorithm( generator-understands and analyzes accuracy; discriminative model- inaccuracy)
• Many examples of advancement in this particular subset of AI are heavily publicized and gain massive attention for the technology.
• There is tremendous opportunity for adoption within the architectural development space.

Contact:

Email: mmcd.jack@gmail.com

LinkedIn: https://www.linkedin.com/in/jack-mcdonald-a960ab194/

Twitter: https://twitter.com/jackmmcd123