The standard model of particle physics and technical architecture (framed around AI)

Science has being great at building knowledge on top of hypothesis, ideas and even points of views. Mathematics and science have found a way of co-existing to complement each other. This has been key to help our knowledge of the universe advance and nothing represents this better than the standard model which showcases the particles that make up the universe. One of the most impressive things is that several particles that appear there were thought in theory as part of a large mathematical equation and then with the help of the scientific method plus the progress with particle accelerators they were discovered to be true. Actually an amazing book from 2009 written by Brian Cox and Jeff Forshaw (2 well known professors of physics) finishes talking about how amazing it would be to discover the Higgs field/higgs bosons (sometimes called the God particle as it's key for mass to exist) and 3 years later at the CERN Large Hadron Collider it was discovered. By the way Peter Higgs proposed this in 1964.

Why does the above matters for Technical architecture? I imagine this should be the first question that will come on anyones mind as they read this. As someone that has been working in envisioning Technical architecture from 20+ years I can certainly say that we should find ways of creating better standards, transparency, cross discipline collaboration and embrace failure. Let me cover each individually and why they matter today. I will purposely frame it around (AI) as a topic which is currently hot.

Better standards

A lot of AI applications today are mostly leveraging someone's API. The responses from the APIs are several times quite mind blowing on how accurate they are and the great engineering behind them. One thing where we have room for improvement is very clearly knowing how long it should take to respond, how many concurrent queries, size of the data obtained on each call. For example what if there was an agreement of Restful API response not taking longer than 0,5 seconds (random number with lack of standards). There will be of course an impact through the lens of today but at the same time then there would be an agreed baseline to works towards to that will force devops, development, etc to better work together towards a common goal. Additional to this is naming conventions and structure of the APIs. For example, did you know that Quarks (elementary particle) were named by Murray Gell-Mann in playful way but unfortunately makes Quantum Physics harder than it needs to be just because the name has nothing to do with their properties. I see constantly the basic principles of naming conventions on AI APIs getting dropped which then create unnecessary complexities.

Transparency and Cross discipline collaboration

Humanity has advanced when knowledge gets shared. Getting your work criticised is important and learning to criticise is even more important. Newton once said that "If I have seen further than others , it is by standing upon the shoulders of giants". I'm genuinely happy with AI reaching the masses not because of the commercial value but because I like fresh perspectives. For some of us that live in 1s and 0s it feels like it has been there forever but honestly I feel we have done less than I imagine we would have done 20 years ago and I'm grateful when I see humour getting enhanced with AI. My point being is that architecture and technology should be written in a way where it's better understood and shared. There's a lot of craft on creating a good architecture and technical documentation but equally many times it can be written in such a complex way that it defeats the purpose. For AI to continue progressing we need to better collaborate and share our architectural ideas with people outside of the discipline. We will get surprised on how it can only make the technology better.

Embrace failure

My son plays chess and people constantly say to him that in chess you either win or you learn. Loosing teaches humility and humility is key to progressing. Eratosthenes (275-194 B.C) predicted the earth circumference to be 24000 miles (today we know it's 24,901 miles) using mostly angles and trigonometry but later others used different methods and got to 17000 (very inaccurate) but thanks to this then people like Columbus decided to sail the sea and try to reach India. Technology is full of these mistakes that have a knock-on effect. When computer science is taught most of the time courses around architecture creation focus on patterns, models, etc but they should take the time to show real life examples of when it went terribly wrong. AI will get things terribly wrong many times and there will be complete architectures that fail and we need to find the right way to openly share more our experiences on this matter. This needs to all the way form the methods of teaching to the way companies behave and act.

One final thought is on the back of an anecdote is that we need to take the bull by the horns when it comes to AI being biased. 7 years ago we created a web app that leveraged AI to understand if someone loved or hated a product while they tasted it. This was done for a campaign with a very loved or hated product in the UK called Marmite. This was something not done before and the technology was very new. We quickly found out that it was racist and the levels of accuracy went down dramatically due to colour or skin and shape of eyes. We made an executive decision to make the algorithm more forgiving which then had a detrimental impact on the people where it worked well just to level the ground. The product became a viral sensation for the right reasons. My learning here is that if we know something can fail then we shouldn't hide it but instead make the right decisions to test, learn and progress.

//This article has been produced with no use of AI and no editors. I'm embracing writing without checking so much how I write to avoid loosing the original intention. I feel this is becoming more relevant now with Internet getting full of AI generated tex. I'm hopefully that my grammar mistakes and anyones grammar mistakes are valued and appreciated more. Isn't it nice to play with languages when we speak and communicate! I'm hopeful that AI will help us appreciate more the nuances of incorrect use of words for the sake of more meaningful interactions.