Transforming architectural design through AI/Generative design technologies - Dr. Mazharuddin Syed Ahmed

 Notes from a presentation to Ara kaimahi (staff) and ākonga (students) by Dr. Mazharuddin Syed Ahmeh, our Building Information Management (BIM) tutor.

Mazhar presents on the fundamentals of AI, LLMs and Generative design. He also covers AI governance (privacy, ethics, hallucinations, and misinformation); and emerging AI and LLM tools / technologies of relevance to architecture. https://www.theb!m.com/BIM-For-Beginners https://bim-in-nz.squarespace.com/bimtools

Began with overview of he journey to BIM via BS in Civil Engineering and through to PhD in education at University of Canterbury.

Proposed that technology disruption is increasing in speed and the need to keep up with the fundamentals underlying technology. A disruptive technology is one that displaces an established technology and shakes up industry or a ground-breaking product that creates a completely new industry. Used the mail service as an example - moving from pigeon post/pony express, to the postal system and then digitally into email and across social media.

Summarised the technology revolutions across the last few centuries - industrial/steam (1760-1820), electricity (to 1900), computing (1900S), digital (today), and artificial intelligence (2025-2030?). For computing, it has shifted from mainframes in the 1960s to mini, personal (1980s), desktop/internet (1990s), mobile (2000s) and wearable/everywhere/cloud (2014+). A key would be increased computing power along with progress in computer science. Gardner hype cycle for 2024 indicates the innovation triggers, peak of inflated expectaions, trough of disillusionment, slope of enlightenment and plateau of productivity (when citizen developers are able to utilise the technology). Humans adoption patterns can be summaries through the technology integration diffusion curve - innovators (techies), early adopters (visionaries) - the chasm - early majority (pragmatist), late majority and laggards (skeptics). 

Adoption of ChatGPT was the fastest - 1 million in 5 days, 100 million in 5 months, almost 7000 prompts a minute! - raising awareness of AI's potential into the mainstream.

Overviewed 'what is data' - presently much of data is unstructured and has had exponential growth, doubling every year. In comparison, traditional data, pre-digital, took 8000 years to double! Present human capability, makes it impossible for individuals to keep up with this volume of data being generated. The human brain has to take 'shortcuts' to help make decisions, leading to implicit or unconscious bias - of which are there many - see visualcapitalist for example!! One way to make sense of things is to use DIKW model - data, information (who, what, when, where), knowledge (how) and wisdom (why). Access to the internet (especially mobile access) is a precursor of individuals drawing on the knowledge of many - although there are implications if we move to 'onemind'.

Implications of AI on how technology is adopted and on jobs/ the world of work discussed. The need to attain data literacy is now paramount. In architecture, everything is data, every data follows a pattern, and every pattern can be modelled and predicted. Explained the concept of big data and data science principles. Defined and provided examples of LLMs (around since 2010) - large language models and their ability to predict 'the next word' based on word structure and sentence construction. Tokens serve as the fundamental units of text in LLMs. A token does not always represent a single word; it can also be made up of a group of characters. As a general rule, one toke is roughly 4 characters. 

ChatGPT/Copilot/Claude/Gemini prompts are more effective if they provide context, task, instruction, clarify, and refine. 

In human learning, we learn by observation, practice etc.to increase muscle memory and cognitive networks. Al-ML-Dl-Gen AL learnings through machine learning - recieve data, analyse, find patterns, make predictions, send answer. Provided examples of how ML is trained through supervised learning, with a 'reward model' used to refine the output. 'Transformers are used to interpret these outputs and convert to the type of response (text, pictures, multimodal etc.) required. AI moving into the near future able to undertake many of the functions of humans. Artificial General Intelligence (AGI) still only able to undertake some functions, so no worries!! - for the moment. AI is still prone to generating mis-information, is somewhat unreliable and may create 'hallucinations'. Pluses of AI need to be balanced with some of the disadvantages of ethics, dependency on data quality, risk of bias, complexity of development and maintenance, lack of emotional intelligence. AI governance is important.

Closed with the potential of AI in architecture. Numerical calculations (numbers, abacus, slide rule, calculator, mobile phone, VR), construction documentation (sketch, to plan, CAD (1980), 3D modelling (1990), BIM bringing in may layers of building data (2005) allow for this data to be drawn on for AI. Therefore, physical structures (cars, buildings, machines) can have a digital twin. Digital data can be used not only in BIM but in the internet of things (ioT) - smart buildings, connected constructions sites etc. From concept, through the design, analysis, scheduling etc, all can be digitised through dimensions of BIM. The AI-assisted design cycle - design details, validation design etc. is possible. Shared examples of the application of AI to architechutral work tasks. 

Trainees at work - online presentation for ConCOVE project

 As a follow-up to another ConCOVE project (see this one in May on Supporting Māori learners in the workplace). Jackie Messam presents findings from the 'supporting technical experts to be workplace trainers' project.

Supporting people who learn on-job has been an under-researched topic. Many sectors are involved in construction and infrastructure - over a dozen - the project involved 7 sectors. Many apprentices learn on the job only but there is a range across sectors. Most trainers are supervisors or peers with some support from the training advisor (from an industry training organisation - now the workplace-based subsidiaries at Te Pūkenga).

One of the rationale for the project was that supervisors had diverse views on training apprentices. Some felt apprentices bring fresh perspectives and new techniques and some felt it was a waste of their time. So do we need adult trainers to complete a 'teaching' qualification or should site supervisors put their emphasis only on health and safety and team leadership? Previous literature on the topic losing currency. Important to gather updated data from trainers and trainees.

Last year, 27 interviews with trainers (9), trainees (7) and strategic roles (11) to establish motivation, knowledge and skills in training. 

Found lots of different people doing the training - the uninformed, the buddy, the recent graduate, the technical expert, the chosen one, the supervisor and the learning/ development team member. 

The reasons they chose to train depended on perspective of trainees (grateful for trainers), trainers (want to see trainee suceed, contribute to team work, better team cohesion), and strategic role (technical experts want to pass on skills and knowledge). From Te Pūkenga Te Rito reports, trainers needed to be responsive to learners and help make learning accessible; provide hands on learning experience; build relationships and connections; and take a future focused approach. The current project identified the attributes of trainers including having emotional intelligence, being experienced, engaging with the qualification and making the workplace a good place to learn and work. 

The project came up with a framework to help workplaces put in place processes to support workplace training. The categories are trainer readiness, focus on the trainee, training skills and strategies, assessment and feedback, and building independence. 

Summarised capability development solutions proposed through survey. There are many trades-relevant resources that can be used to support trainers and examples provided - Spotify podcast (The Tradie show); Tik Tok videos, various courses etc. The project findings are useful to vocational education and training - employers, educational providers, qualification developers, programme designers, industry bodies, procurement decisions, other trade-based industries. 

A good project to update the data from industry perspectives. Workplace learning is always a complex and contested environment. Therefore, important to keep tabs on what is happening at the 'coal face' to support the work of workplace trainers as they are the people who develop the next generation of industry experts. 

Student perceptions of Gen AI - JISC report

 From JISC, comes a report on student perceptions of Gen AI.

Students are from the UK HE and colleges sector. Of note is that this is a follow-up report from one completed in 2023. The changes noted are of interest. These are the shift towards collaborative learning using AI as coach to support learning activities; students' reporting of the need to attain Gen AI skills for their future work; concerns around ethics, equity and accessibility; and an expectation from students that their educators should be competent in integrating AI and have policies to ensure the learning environment is fair.

A stock take is also reported on how students use Gen AI for communications and content creation, learning, researching, programming, to support creative ideation, assist with productivity and task management, personal and emotional support.

Students also provided concerns on the need for Gen AI literacy, better information and clarity on academic integrity, the responsible and ethical use of Gen AI, equity and accessibility, and employment connections. 

All in a good summary of student concerns and recommendations. 

The power of peer learning - open access book

 Here is an open access book, published by Springer in 2023. The power of peer learning: Fostering students' learning processes and outcomes is edited by O.Noroozi and B. De Wever.

The book has 17 chapters, organised into four sections;  conceptual contributions, methodological contributions, technological contributions,  and empirical contributions to peer learning.

Most of the chapter feature work undertaking in higher education with several in secondary education. 

The first chapter, The four pillars of peer assessment for collaborative teamwork in HE, by B.Sridharan, J. McKay & D. Boud, sets the scene and provides an overarching framework to plan and structure the peer assessment process. 

Other chapters are useful with contemporary presentation of processes that are relevant in the age of digitised / online learning. A book to dip in and out of as the need arises. 

The future of LMS in an AI world

 The work of  Professor Stephen Marshall and Professor Michael Sankey are always important to keep abreast with. They provide scholarship on the strategic purposes and direction in how technology is managed and applied across higher education. 

A recent piece of work 'the future of the learning management system in the virtual university' summarises their thoughts on how the current LMS, needs to move from being a 'single system' towards being part of (and perhaps a primary part of) a learning ecosystem. As AI creates more disruption within education, the morphing of AI into the offerings of LMSs to create personal learning environments) PLEs for learners becomes more pressing. 

Many institutions promote constructivist learning but LMSs structure, often make the organisation of resources / learning activities to make this happen, clunky. There are some specialised systems exampled by OB3 but in the main, many educators use LMSs as a giant resource repository, rather than a means by which learners are able to archive their learning. ePortfolios could be one way for learners to construct their own 'learning hub' but requires time and effort. Learners are instead welded to the strictures imposed by whatever LMS is used in the institution they study at and for many, little or no ability to modify what is offered in their course LMS.

Personalised learning which is supported by AI is not new. However, to put it in place, requires institutions either purchasing the platforms and integrating this into the existing LMS, or developing their own. Gen AI creates opportunities to democratise the creation of 'chatbots' or similar with tight parameters to help learners attain specific learning outcomes. However, these may be too specialised and again, not usually customisable by learners themselves. 

Last year, Professors Sankey and Marshall, wrote on 'the learning management system of 2028'. The article provides a good overview of LMSs, where they came from, where they are now and what should happen to make them more relevant to learners going into the future. They propose not only the integration of AI into the LMS ecosystem, but also the importance of being aware of what happens to LMSs beyond education. That is how professional learning development across work, use LMSs. The other important take away from the paper, is to align the way LMSs work, to the productivity tools used across businesses or corporations, especially the ways that communication, knowledge sharing and team work occur.

All the above for some indications as to how to progress beyond the current way LMSs are constituted and what should happen, going into the future, to provide a more authentic and learning-centred learning environment. 

Higher education for good - impressions of the book

 Here is an open access book - Higher Education for Good - with a few relevant chapters. The book takes an optimistic look at higher education and its potential to contribute to the common good of society and humanity.

The current economic and political climates. have been challenging for tertiary education as a whole, across many countries. The book has a series of essays, discussing the challenges and future for higher education. Sections include - finding fortitude and hope; making sense of the unknown and emergent; considering alternative futures; making change through teaching, assessment and learning design; and (re)making higher education systems and structures. 

Several chapters are of interest to those beyond higher education. Including: 

- Artificial intelligence for good? Challenges and possibilities of AI in HE from a data justice perspective.

by E.  Pechekina, The focus of the chapter is on how AI can be used to support students and learning but also undertakes an discussion on the ethical use of AI, especially in the 'prediction' of outcomes based on learner demographics, performance and other data. The issue of equity is also presented.

- Humanising learning design with digital pragmatism - by K. Malloy and C. Thomson. This chapter provides a learner designer view, with the practitioners located in a 'third space' of being neither management or academics. It argues for the need to champion the needs of learners, as one way to navigate the power structures in higher education.

A book that will have chapters that will be of interest to all tertiary educators, interested in how their work impacts on social justice and how each is able to play a part in making the world a better place.

 

AI literacy book

 This open source book - Towards AI literacy 101+ - creative and critical practices, perspectives and purposes contains a range of short stories on AI. Examples are provided as to the challenges and ways to define AI literacy and the many interpretations of AI literacy across various educational levels and disciplines. The range of articles show how diverse education actually is and that there is never a 'one size fits all' model. Education, by the grounding of its purpose, must be of relevance to learners and the communities it serves. Therefore, how each context needs to use and apply AI, provides how AI literacy could be introduced and integrated. 

The 'generic' skill or competency to come to grips with AI would be critical thinking. Again, to be able to think critically in various contexts, is variable. The discipline or subject is but a platform from which the development and application of critical thinking is grounded. Care there needs to be taken when designing learning, to ensure that there is a balance between discipline/subject-based 'situated learning' and to ensure that learners are provided with models and skills to 'cross the boundary' when they make shifts across  disciplinary fields.