📅 Posted 2020-08-26
Everyone uses email. It kinda works, but it’s super janky and would never take off if someone came up with the idea today. Somehow, we just put up with it.
Typical IT roles which design, build and support email systems could include IT Support Analyst, Messaging Engineer, Messaging SME, Solution Architect or IT Systems Administrator.
Typical marketing roles which design, build and support email marketing campaigns include Campaign Manager, Marketing Manager CRM Marketing Specialist, Marketing Automation Specialist, Digital Marketing Specialist.
These people are totally specialised and work with email as part of their core role and without them… well email wouldn’t be quite the same as it is today.
In the ML field, there might be data scientists, data engineers, infrastructure engineers, and other emerging roles which will slowly reshape workplaces, and these are equally important in keeping the ML dream alive.
But regardless of the role today, you’re using email every day. Even my builder uses email (or more likely pesters his wife to send invoices for him).
In 99.99% of roles, email is an essential tool of the trade. It’s not even that good for communication and collaboration, but it’s hard to argue with the pervasiveness of email in every process. We just make do with it, grit our teeth, and carry on.
OK, so I’m not defending email much here. I wouldn’t ever try doing that.
But take the most important feature of email: pervasiveness. Now imagine a world where machine learning is the new email. I mean the good bits. Not the bad bits.
Lets see what Google Trends has to share
Look at the Google Trends from 2004 onwards for Machine Learning:
Massive! Now let’s look at Email and then compare it to Machine Learning:
Oh dear. There is much work to do here. ML doesn’t compare at all but the room for growth is immense. Yes, this is a facetious comparison. But…
Imagine every process imbued with machine learning technology.
Every operator, in every role, as part of every organisation, at every level, doing their thing with machine learning.
A reference definition from Wikipedia is:
Machine learning (ML) is the study of computer algorithms that improve automatically through experience. It is seen as a subset of artificial intelligence. Machine learning algorithms build a mathematical model based on sample data, known as “training data”, in order to make predictions or decisions without being explicitly programmed to do so. Machine learning algorithms are used in a wide variety of applications … where it is difficult or infeasible to develop conventional algorithms to perform the needed tasks.
What if we all ran a ‘machine learning client’, just like we might run an email client (or have gmail open in a browser tab), and we dip in and out of it throughout the day to get our jobs done. What if there was a standard, an app and an expectation that people know how it works, at least in basic terms.
There could be many things you’re doing with machine learning. It’s a large area and it’s changing all the time. But I feel like we’re not taking advantage of it enough, we’re just still thinking it’s about the boffins and the academics and we haven’t dialed it up to 11 and that’s a shame.
Say we’re using this machine learning thing like it was email. Doesn’t matter what our roles are or where we work.
We might be pulling in some new data.
Running some training.
Measuring the output.
Deploying a model.
Sharing the model.
Building on someone else’s model.
Pulling down that latest codebase from the latest hotness in the open source community and capturing cause-and-effect relationships.
Maybe we’re supervising the learning, maybe we’ve gone deep. Like, real deep.
We’ve got infrastructure set up, so that everyone can log onto it.
There’s a support process, so you can call the IT helpdesk if you’re having trouble with your model. Something just isn’t working quite right and normally it’s OK.
There’s access permissions, shared models and open data sets.
Groups can be created, with open or closed membership.
People from outside the organisation can contribute where it makes sense.
You can set up your own profile, with it’s own settings and preferences and this will follow you around the organisatiton, depending on what tool you’re interacting with at the time.
So everyone’s going to be a data scientist now?
No, not exactly. Just like how everyone isn’t a Microsoft Exchange engineer or someone who knows how to integrate Office 365 with an Electron app, the same applies here. I’m looking at this from a pretty hands-on perspective because that’s how we’ve grown to love and loathe email. To be clear, I’m not expecting the janitor to come up with a new algorithm for cleaning floors, but maybe there are accessible ML tools which make it just as easy to get going as writing an email about the mess the yobs on level 3 made last week.
What if the CEO of a company was able to take advantage of machine learning just like how they can jump on their computer, type a few words and send an email to the whole company?
What’s holding us back?
It’s a new(-ish) area. Organisations will take a lot of time to gain confidence in an area which is growing and still finding its feet.
Some big IT corporations have of course jumped onto the bandwagon, touting their incredible platforms which so often fall short of expectations. A lot of bang and a lot of fizzle, sadly.
A lot of it seems like magic and when it works, because the algorithm said so, it’s difficult for some people to reason with.
Organisational structures aren’t typically geared up for this. Imagine if it was 1972 and you wanted to use this new email thing and had to go to the messaging department every time you wanted to write up a new one, only to be blocked by competing priorities because everyone else also wants to send their letters? OK, a facetious example again as I’m sure email was never designed to be so limited in accessibility.
I can see organisations setting up discreet teams who ‘own’ the ML space, rather than focusing on the broader idea that all jobs and all roles are going to be impacted by this radically different approach. These smaller teams are probably easier to get buy-in and easier to remediate, should things need to change radically.
A lot of machine learning experiments will fail and this doesn’t build confidence. But it’s a lot like the new way to do R&D and (generally speaking) companies don’t spend enough time and energy on R&D due to the high failure rate, forgetting that companies who don’t do enough R&D risk becoming the next Eastman Kodak.
And then there’s always those 2 column PDF-based printer-optimised academic papers which are pretty tough reading!
Imagine a world where machine learning is ubiquitous like email. I think it would be an amazing time and place to be.