Would it be possible to use machine learning to create new digital networked services and improve existing ones? Is it possible to create services that understand what we really want?
Maybe some 10 years ago I looked at Wolfram as I began teaching at MediaLab at Aalto University (The TAIK) the course Dynamic Visualization Design 2. 2 Stands for 2nd semester and the first semester was and still is taught by Markku Reunanen. At the time Wolfram didn’t open up to me even though I understood this should be a powerful tool for our students to learn and use.
10 years later I have not progressed much more in regards to learning, let alone teaching Wolfram but my interest spiked again after writing my post yesterday on Service Design and AI. There was a possibility of me running a service design department at a school here in Finland but that unfortunately did not materialize, yet. I have what I think are very interesting and future proof courses in mind that circle around service design and AI.
During my Caloom time (version 1, version 2, version 3 and version 4) I especially in version 3 Nick and I spoke about intelligent services. Most of our quest was about creating a smart service that allows you the user to figure out what data correlates to what other data and information. The result could be a better understanding of the information and especially now in the hyped “fake news” era this could have been a good tool for some. What I learned was that our human choices influences how the system would represent the data and information. For example, the choice of how to structure categories and what to include in categories would impact what is seen in which context. We humans often tend to confirm what we think, you will see this yourself when you observe your search engine queries.
To break this I apply my “180 degrees” tactic where I first type what I look for (confirmation bias) and then turn it 180 degrees around in another query. This gives very interesting results but above all it changes my attitude towards “facts” and “truth”. I am not saying everything is subjective but what I do say is that too often we only see a piece of the bigger picture whereas the whole picture would give a good idea about the reality and a small piece not. So also for interacting with digital and automated services, garbage in = garbage out?
So can we create services that broadens our understanding rather than pushing us deeper inside the echo chamber or search tunnel?
Now back to Wolfram and machine learning. With Caloom version 3 we always stayed at a very high level of using only what+when+where+who+worth. This worked and now my question is: can Wolfram be used in a similar way but then better, faster and smarter?
- Beginners can write web applications with only a basic knowledge of HTML and Mathematica
- Built-in tools for website deployment—no additional toolboxes required
- Free-form linguistic input for learning new functionality while getting results
Unique to Mathematica
- A large set of customizable data mining and graphing functions lets you analyze website traffic in the way that makes sense for you
Unless I am mistaken it sounds as if with Wolfram I can write browser based applications with basic(?) knowledge of HTML and Mathematica. The first I have, I can write HTML and CSS in a simple text editor and it even works. Mathematica I would have to learn but that is doable I suppose. This combination should allow me in theory to connect to vast amounts of data, somehow work with that data, somewhere, and present this back inside the browser.
If the presentation can change based on the data it receives this could be interesting indeed. In theory I could begin to customize services for each individual user and leave this dreadful Web 2.0/3.0 time frame. Back in 1999 I gave a lecture at the University of Tampere, department of Information Sciences or something about how in the future we would need a personal interface to the online world. In my lecture I outlined how lot’s of interaction with this new online world would be repetitive and personal.
Repetitive in for example the use of username/password, filling in addresses, emails and so forth. Even though we had autofill in some browsers you still had and have to take care of what information you actually want to supply. A service that helps you with this is still desirable I think. But these are details as what I find most fascinating is the opportunity to turn the table and let us, humans, take control of how we use data instead of a handful of mega corporations. It is good these mega corporations exist as they bring us the massive amounts of data and also tools. What worries me is the 1-way mirror through which they can look at us and we cannot look back at them.
The data and logic is too often a black box and we must assume that their “recommendations” are actually based on our preferences. Having seen way too many times this not matching up for me I have of course began to wonder about the logic that is used. You see, every click I make on my keyboard is sent to my computer and directly acted upon in the form of changes on my screen. In many cases my clicks result in changes on computers far away and these clicks are stored and analysed. All fine but why can I not analyse my clicks too?
My click database would outnumber any collection of clicks of mine at any given mega corporation, unless they monitor all I do which I doubt. So I should be in the best position to capture these clicks and other interactions, store them, analyse them and present them back to me with recommendations. Now these recommendations should be a mix of my data combined with other, external, data. So the mega corporations do have an important role to play.
I do believe that my 1999 idea could be turned into an AI service possibly built with Wolfram. This would also put me as a non-designer in the seat of creating my own services. Will I be the “service designer” then as well or merely someone who shuffles the pre-designed options? I do not know, possibly a bit of both. This is what I would like to work on if I would venture into AI and SD…