Fundamental properties of digital IT have set ons on a road not to a Singularity Point, but towards Complexity Crunch. That has consequences for our strategic (IT) choices and landscapes. A ‘long read’ (sorry) about lessons we can learn by now after half a century of Digital Revolution so far. Written as I have been giving talks about the subject this pas half year.
Author: gctwnl
I work in the trenches of real enterprise architecture and write on the basis of that.
Generative AI doesn’t copy art, it ‘clones’ the artisans — cheaply
The early machines at the beginning of the Industrial Revolution produced 'cheap' (in both meanings) products and it was the introduction of that 'cheap' category that was actually disruptive. In the same way, where 'cheap' is acceptable (and no: that isn't coding), GenAI may disrupt today. But there is a difference. Early machines were separate inventions creating a comparable product. GenAI is trained on the output of humans, their skill is 'cloned' and it is this 'cloned skill' that produces the 'comparable product'. GenAI is not 'copying art', it is 'cloning the artisan'. And our intellectual rights haven't yet caught up.
No-IT. Really. No. I. T.
What happens when your organisation suddenly loses all of its IT? There are enough realistic ways for that to happen. Think: a really successful ransomware attack. As it turns out, first turning ourselves into 'digital organisations', and then requiring a speedy recovery from 'digital armageddon' creates a weapons grade challenge. A story about 'Out-of-Systems', 'Out-of-Sync', and your 'Minimal Viable Organisation' (MVO), and a 'fix' that may only make matters worse.
When ChatGPT summarises, it actually does nothing of the kind.
One of the use cases I thought was reasonable to expect from ChatGPT and Friends (LLMs) was summarising. It turns out I was wrong. What ChatGPT isn't summarising at all, it only looks like it. What it does is something else and that something else only becomes summarising in very specific circumstances.
Microsoft lays a limitation of ChatGPT and friends bare
Microsoft researchers published a very informative paper on their pretty smart way to let GenAI do 'bad' things (i.e. 'jailbreaking'). They actually set two aspects of the fundamental operation of these models against each other.
Don’t forget all the things that a core team performs to a tee, but that you never see
The third 'fragmentation wave' of the IT-revolution is upon us, it seems. Fragmentation/encapsulation is a repeated pattern in the IT-revolution for managing complexity. First as object oriented programming (for code) and later as agile (for IT landscape change). Now, it is the organisation’s turn to fragment. How strong is your mission, your ‘why’? You might soon find out, thanks to IT.
Ain’t No Lie — The unsolvable(?) prejudice problem in ChatGPT and friends
Thanks to Gary Marcus, I found out about this research paper. And boy, is this is both a clear illustration of a fundamental flaw at the heart of Generative AI, as well as uncovering a doubly problematic and potentially unsolvable problem: fine-tuning of LLMs may often only hide harmful behaviour, not remove it.
Will Sam Altman’s $7 Trillion Plan Rescue AI?
Sam Altman wants $7 trillion for AI chip manufacturing. Some call it an audacious 'moonshot'. Grady Booch has remarked that such scaling requirements show that your architecture is wrong. Can we already say something about how large we have to scale current approaches to get to computers as intelligent as humans — as Sam intends? Yes we can.
The Department of “Engineering The Hell Out Of AI”
ChatGPT has acquired the functionality of recognising an arithmetic question and reacting to it with on-the-fly creating python code, executing it, and using it to generate the response. Gemini's contains an interesting trick Google plays to improve benchmark results. These (inspired) engineering tricks lead to an interesting conclusion about the state of LLMs.
Memorisation: the deep problem of Midjourney, ChatGPT, and friends
If we ask GPT to get us "that poem that compares the loved one to a summer's day" we want it to produce the actual Shakespeare Sonnet 18, not some confabulation. And it does. It has memorised this part of the training data. This is both sought-after and problematic and provides a fundamental limit for the reliability of these models.
R&A IT Strategy & Architecture 