Massacre of the Dilberts – and ant colony optimisation

The massacre of the Dilberts

A phrase coined by Governor of the Bank of England Mark Carney back in 2018 to describe the impact of automation on millions of routine white-collar workers – Tech Upheaval Means a ‘Massacre of the Dilberts,’ BOE’s Carney Says.

So what does that mean for all of us ‘Dilberts’ in the lighting professionals?

Phil Bernstein’s fascinating and timely book, Machine Learning, sets out the threats and the opportunities for architects, although the thrust of his arguments apply to all of us in the business of designing and delivering buildings.

Phil uses Mark Greaves’ hierarchy of AI capabilities, broadly based on Bloom’s taxonomy of learning,  to describe the evolution of digital technologies over the past 20 years: from basic understanding, to evaluation, simulations and generation.

CAD, BIM and current ‘big data’ machine learning algorithms are powerful number-crunchers.  But, as psychologist and entrepreneur Gary Marcus points out, most AI tools currently focus on ‘narrow intelligence’ strategies that are fatally flawed as they lack a rudimentary understanding of how the world works, or anything remotely resembling common sense. 

Luckily this leaves (just) enough space for us Dilberts to add enough value to earn our place at the table. 

But this won’t last long. Pedro Domingo’s influential book, the Master Algorithm, set out a vision of an ‘artificial general intelligence’, informed by logic, neural networks, evolutionary psychology and probability as well as pattern recognition that will learn and reason about the world in context’.  The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World. We are already seeing this silent rip current at work in the AI-driven script generation platforms that triggered the first strike in the film-making sector since 1960 – ‘Embrace it or risk obsolescence’: how will AI jobs affect Hollywood? While the Writers Guild Of America may have won this battle, the war is far from over.

‘Data is the oil of machine learning’ – and yet, as Bernstein points out, most of the data in our sector is fragmented and siloed with no commercial incentive to disclose, let alone collaborate to optimise performance for our clients and colleagues or the environment. We have no equivalent of the legal sector’s case law –  British and Irish Legal Information Institute or scientific journal directories – Google Scholar to learn from each other’s successes and our mistakes. Even our industry standards are locked behind paywalls, making it harder for clients and independents to check specifier claims against legal requirements and guidelines for themselves. Warranties and post-occupancy verification let alone ongoing adjustments based on real world data are the first casualties of ‘value engineering’, Besides, they open up a minefield of liability that most are keen to avoid.

Bernstein also tackles the thorny issue of skills, pointing out that most architects (like lighting designers) sign up for a ‘creative’ profession and have little interest or natural abilities in technical skills such as programming or data analysis.

And yet there are plenty of opportunities to divert the Dilbert doomsday prediction. Lighting infrastructure is uniquely-placed to provide a distributed backbone for an integrated approach to adaptive building management, transforming the role of lighting professional from afterthought to foundational partner in a shared goal of delivering comfortable, productive and sustainable spaces for all. At the detailed level, the potential for machine learning to assimilate billions of data points from spectral curves to optimise visual comfort and circadian stimulus could serve, rather than replace the craft of the lighting designer if we challenge the false dichotomy between science and art.

Bernstein’s book concludes with an invitation for us all to work together ‘to improve the quality of the built environment and to enhance the relevance of the human architects who are best-suited to deliver these improvements through design’. To paraphrase:

1. Proactively define and guide the technologies that will frame our practice in the future.
2. Expand our remit beyond design and specification to include performance in use.
3. Create the data infrastructure for greater transparency and integration.
4. Change the relationship between different players in the value chain to improve cost and energy efficiency overall.
5. Shift the value proposition to include simulation, licensing and other asset management models.

We ignore this tide at our peril. There is still time to build the raft of data and skills that could help all our ships to rise.

New Designers 

If you’re tired of AI’s uncanny ‘cute’ aesthetic, head over to the Business Design Centre in London to download a bundle of gloriously analogue texture, talent and enthusiasm – TICKETS FOR NEW DESIGNERS 2024, or surf some of the trends online- here’s a taster video from Swansea College of Art – SWANSEA COLLEGE OF ART UWTSD.

I’m privileged to be one of the judges for the Colour in Design Awards this year, sponsored by Dulux – will be a tough decision I know! – The Judging Process.

Awkward or advocate?

Physical workplace adjustments to support neurodivergent workers: A systematic review.

Walk into any workplace, even one in a garden shed, and you’re likely to find an adjustable chair, an ergonomic mouse and potentially a sit-stand desk.

But it’s rare to see similar attention paid to the lights.

And yet, even between ‘neurotypical’ people preference for light levels varies by up to 50% – Lighting preferences in office spaces concerning the indoor thermal environment, while low-glare screens and balanced task lighting can cut risk of painful dry eye syndrome in half – Digital Eye Strain- A Comprehensive Review. Flicker from fluorescents or poorly-controlled LED’s combined with refresh rates on screen trigger eye strain, headaches and exhausting low-grade stress, especially in people on the autism spectrum – Influence of ambient-tablet PC luminance ratio on legibility and visual fatigue during long-term reading in low lighting environment. Visual clutter is distracting and for everyone, but especially for people wired to notice the details – ‘Visual clutter’ in the classroom: voices of students with Autism Spectrum Disorder, Increasing the load on executive working memory reduces the search performance in the natural scenes: Evidence from eye movements.

So you may feel as though you’re the only one ‘making a fuss’ asking for adjustments to the lights, you’re almost certainly not alone.

Here are five simple reasonable adjustments you can request without breaking the bank:

1. The ability to switch off overhead lights to reduce sources of flicker and glare
2. Access to adjustable task lighting to increase light levels for detailed tasks
3. Anti-reflectance screen covers and adjustable monitors for head and neck posture
4. Manual control over window shading and blinds – and replacement of ‘venetian-style’ blinds that generate high-contrast ‘stripes’
5. Portable screens or shields for visual privacy and focus – Autism and Visual impairment: A First Approach to a Complex Relationship.

Ant intelligence – or artificial intelligence?

You might find ants a pest – especially if they get into your garden or kitchen –  but Ant Intelligence is fascinating.

These incredible creatures use simple rules to optimise individual behaviour that solve complex problems for the group.

That’s the inspiration for a whole category of machine learning algorithms called Ant Colony Optimisation.

But that’s not all.

Ants ability to extract patterns from crude monochrome visual data is the inspiration for many of the computer vision algorithms in the facial recognition software that gives you access to your banking app or identifies an anomaly in a routine scan.

Ants have compound eyes – an array of individual sensors, like pixels in a camera – that captures a very low-resolution image, usually in black and white.

Their tiny brains use simple rules to extract edges, movement, and light polarization from that simple data stream. They can even learn to find their way through a maze with this simple data – that’s why they keep coming back for more.

Ant Intelligence – or Artificial Intelligence?

It’s definitely worth a look!