The question is regularly thought upon… will AI replace me?
Well in short, no. We will always need a human mind to both create problems and solve them!
I often wonder, will AI help to unlock our creative spirit? Can all those mundane tasks of analyzing data be taken care of by AI, allowing us to finally ditch our screen addiction and interact with one another on a more human level. Why not step away from our computers and let AI do its thing whilst we converse, generate ideas and innovate?
This would be fantastic, but many of us don’t have our houses in order to even be able to use AI effectively! So, what do we need for AI?
Can the chef cook me something to my taste? Hmm no or at least not very well! Not if they have no information about which meal it is, what types of flavors or cuisines I like, for how many people, where I'd like to eat my meal or when. Providing context and being specific about your question and expected outcome will lead to much better results!
Start with the problems you’re experiencing, what are those tedious tasks you think AI can help with? Further develop these to be specific, measurable goals for AI. The classic, ‘who, what, where, when, how’ serves as a great baseline, however your goals will also need to consider any inputs (data types/ sets), outputs (formats/ end use), performance (accuracy/ speed) and constraints (software/ compliance/ budget).
Think of data as the essential ingredient AI needs to cook up to results and without it, there's really nothing to serve! But not just any data...
Imagine a supermarket which has all the various types of products located randomly on it’s shelves every time you visit - it would much trickier to find what you needed! Whereas food products that have been categorized, labelled and stored in an allocated place would make your shopping experience much more efficient.
A critical (very critical!) step for sorting your BIM data is to implement naming conventions and categorizing data appropriately. It helps us to achieve consistency, which not only avoids confusion but allows for automation, classification and AI processes to easily find and work with the data.
We may already have our own individual set of standards, but adopting any global standards such as, ISO 19650 or IFC, provides an opportunity for AI to increase its learning, accuracy and speed, as a result of the consistency with data across our software, projects and companies.
You’re at an event that includes lunch, you head over to the table but can only see a basic sandwich and two types of soft drinks. Unluckily, the sandwich contains meat and you’re a vegetarian. Do you think your participation for the afternoon session of the event will be to your best? Probably not!
AI is constantly learning from the input data it is fed with. The more you feed it with data, the higher increase in its performance and accuracy. Better still, is if you feed it with a diverse range of data, so that it is able to handle multiple scenarios, recognize patterns and predict results.
Depending on the use cases of implementing AI within your BIM processes, data can be collected from various sources for example: drawings, models with meta data, schedules, reports, documents, photos, videos, scans, sensors, etc. generated from the lifecycle of the project.
Fresh ingredients with balanced flavours give us a safe, delicious and nutritious dish. Bad quality or missing ingredients however, may result in an unsavoury or inedible dish!
Good quality data is essential in order to be able to rely on the outputs and predictions from AI. If AI is fed with poor quality data, it will throw up incorrect or biased answers.
Implementing quality control into your BIM processes will go a long way. Clean your data by fixing any errors and inconsistencies, removing duplicates, completing any missing attributes, performing cross checks and checking for code compliance. Update any data that may be out-dated and ensure relevant data is being used to the intended purpose of AI.
Setting up the right environment for your data will improve your results. If your kitchen is equipped with unreliable or weak appliances; or it is not large enough to even store what you need, it can be very frustrating to cook in and you are more likely to be slower, make mistakes or run out of ingredients.
High performance servers and networks will really help AI to process large amounts of data more quickly, reducing the risk of crashes or errors. Cloud-based and hybrid servers will allow real-time data access and integrations to take place.
BIM data is often generated through collaboration and comprises of large datasets. A server that is powerful, can be scaled up if needed and accessible remotely would be appropriate for AI systems. Security requirements should also be considered in your selection to prevent any data breaches.
Humans are the chefs who are skilled, knowledgeable and experienced. When adopting new systems, there will always be some learning and small changes in one’s roles and responsibilities.
With the adoption of AI for BIM, team roles will need to enhance their flavour profiles to include: delivering high quality data, clearly defining problems that may be solved by AI, being able to choose and operate the right AI models for achieving tasks and recognizing inaccuracies and errors through design and construction knowledge.
Let’s not chuck our random ingredients into the pot for this, but rather take a breath and prepare intentional, high quality data; combine with collaborative, skilled teams; and a dash of innovation for a delicious and healthy application of AI for BIM (or other).
Bon appetit!