Artificial Intelligence and the evolution of Being Digital
Next week, at the annual Web Summit conference in Lisbon, there is going to be one subject in front of mind – the rise of Artificial Intelligence. There is little doubt that the next phase of digital transformation will be AI-driven, and the conference will be right to draw a spotlight on the subject. Indeed our very own Sairah Ashman will be speaking on the subject of the unique design challenges which we face as we enter into this new era.
However, as I’ve read the increasingly frantic thought pieces and encountered clients who seek to understand this evolution, I’ve been concerned by the seemingly loose use of the term “Artificial Intelligence” as a singular concept.
Despite often being thrown into the ‘disruptive technology’ ring with the likes of blockchain, AI isn’t really a single technology per se, more a goal which a variety of research is exploring. There are some specific technologies linked to this which I’ll expand upon below, but in most cases it’s more accurate to talk about “AI-related technology” (call it AIRT if you really want something snappy for your presentation) than AI as an entity unto itself.
The Wikipedia page for AI is actually pretty good and a very useful starting point for deeper exploration of the field. I would particularly call out this bit in the intro:
[The term AI can be applied] when a machine mimics “cognitive” functions that humans associate with other human minds, such as “learning” and “problem solving”.
Intelligence is obviously a loaded term. Is a plant ‘intelligent’ because it turns to the sunlight? Is a computer intelligent because it can parse human speech or process millions of records in a fraction of the time a human can? Real intelligence is the macro goal of AI, but even if this still remains well out of reach (and it really does – the complexity of the problem is only just beginning to be explored), the research into the very complex systems which lead to human intelligence is throwing up some interesting areas of study and some potentially transformative applications. However, let’s not confuse advances in, say, sensor technology or general computing power with the more specific meaning of AI as a research goal.
By way of example, here are a few things that AI is not:
Natural language parsing – although a deeper understanding of language structures and things like sentiment analysis would be.
‘Chatbots’ and virtual assistants (Alexa, Siri etc) – whilst the endgame of these is some semblance of self-directed intelligence, and many may even pass the classic ‘Turing’ test, in their current form they are certainly not intelligent. This is a simulation of real interaction at best, and there is a big difference between the ability to parse language and the ability to understand it.
Computation and algorithms – this is often confused with AI, when what they really refer to is just a very fast, very capable computer running clever code. Whilst different approaches to computer and code design will aid in reaching AI’s goals, computation in of itself is not AI, no matter how fast or advanced it gets.
Robotics – most robotics work at present is centred around improving sensors and the processing of this data. Again, AI will be an endgame for robotics, and arguably the ability to build up complex models of the real world and set goals will be important, but at present that is still the stuff of sci-fi. See also autonomous vehicles, drones, etc. The ability of a machine to sense and respond to its environment does not constitute intelligence (see aforementioned plant example).
So why is AI important right now?
If the last 20 years of digital transformation have been about humans learning to “be” digital, then the next digital revolution will be about a more symbiotic relationship with technology. The research goals of AI (e.g. machine learning, knowledge representation, planning, social intelligence) are really about giving our machines a better understanding of how the human world works.
Perhaps a better way to look at what AI will really represent for humanity is to think of the technology-driven goals in more human terms – evolutionary technologies which will superpower us.
1. Sensing us, understanding our world
The kind of tech being applied to autonomous vehicles will also help us build detailed, real time models of our environments. Technology can become more passive and present at the point of need, rather than requiring our direct attention. What place will the smartphone have in that world?
2. Understanding our language, understanding us
Language is the base unit of communication in human society, but it isn’t alone. Sentiment, emotion, body language and the structure of our speech are vital in conveying our complex needs. If our technology can start to understand this, then important context will help our daily interactions with the digital.
3. Learning from us
One of the building blocks in any evolutionary process is the ability to learn. Whether through trial and error, or by building complex models, humans test and learn without even being conscious of it. The AI goal of machine learning helps our technology learn from us and with us.
4. Reasoning and solving our most complex puzzles
A computer’s ability to process lots of data really quickly has been its biggest utility to us, but generally, it’s following a human’s encoded logic. The AI goals around reasoning and problem solving might help us gain unique perspectives on the world.
5. Planning and modelling in the human world
Whilst this can be the most contentious area of AI’s impact in the world (particularly the fear that automation may lead to a loss of agency), there are also promising applications. The building of predictive models against everything from digital security vectors of attack to health conditions will be vital tools in helping us see problems before they manifest.
6. Creating with us
Finally, perhaps technology will create new things with us. It may even help super-charge humans to create new ideas, new perspectives and perhaps even new expressions of what it is to be human.
You can already see the diversity of goals, capabilities and approaches that current AI research is bringing to the table. The value of this technology is far more complex than just a loose application of the term to denote “cool new computing technology”.
Perhaps more importantly, like any moonshot, the real prize isn’t actually the end goal of general machine intelligence. This may not even be possible (we are only scratching the surface of what we understand about human intelligence, let alone building machines to recreate it). But in the course of this next evolutionary phase of “Being Digital” we will learn and invent all kinds of things along the way, and it’s these things that will certainly be changing our world over the coming decades, even if true AI remains in the pages of science fiction.