AI and context: which jobs will it replace?
Directions of the only product innovation still standing.
It seems there is an exodus on the product market: product managers, product designers, data scientists, user researchers were let go en masse. As of this writing, there are more than 250.000 people let go from these roles, including Head Of User Research and similar positions of the biggest players.
On the other hand, even my mom knows about ChatGPT: strange, since I had to introduce her to much more mundane inventions, like email, a few years back. I still remember her frowned-upon face receiving her first tablet for her birthday, only to develop a Netflix addiction on it less than 10 years later.
It seems that AI will take our jobs for real now: we wanted it to clean our homes, drive our cars, on the one hand, and make big, hard, world-defining decisions for us on the other, and instead, it writes poetry and paints art. Sure, not the really extraordinary ones we all yearn for, like Ginsberg’s Howl or Klimt’s Kiss, but the serial ones 99.95% of the art industry made a daily living of, for a few thousand years up until about 2022.
It seems AI can take all the bullshit jobs as well, or we secretly hope so — that’s still a few years away, but not to worry, it will come soon. But what can and can’t current AI technology do?
It’s all about context
Were you ever in a (junior or medior level) job where you could get productive after a day or two? Keep your memories; those jobs will go bust.
The two phases of learning for AI
It all comes down to how the AI technology works: it’s essentially a translator, with two phases: learning and doing. When AI is “in training”, data scientists watch it all day, trying to supervise the process. They teach how not to answer certain questions (like, how to hide a dead body), and in general, this is when it learns how to do its job.
While ChatGPT became famous for remembering past parts of a conversation (somewhat a new phenomenon, and something competitors for the Turing Award… ergh Loebner prize has been trying to shoot for, for over a decade), its “learning abilities” for the average user stop right about there.
So while ChatGPT (and all deep learning AI in general) has a wide knowledge of how the world in general works from their perspective (sometimes measured literally in terabytes), its knowledge of how specific is their user’s world is about the complexity of a profile / personal settings page. And that’s a benefit for us human would-be-employees (and would-be-contractors).
The two phases of becoming a surgeon
AI isn’t the only job where learning has two phases: so it is with surgeons. For years, they study general medicine (about six years in most European schools, US differs), only to be enrolled in essentially a trainee program, where they will face the inevitable differences between theory and practice in a somewhat controlled environment: while the appendix of any high school anatomy model will lay nicely on the left side, just above the lower edge of the torso, unfortunately for would-be-surgeons, real appendixes tend to also appear elsewhere as they see fit.
The difference between theory and practice could be called context.
Fortunately for all of us, most surgeons will understand this difference in a few years of completing various tasks in the operation floor, making appendicitis a pretty harmless issue for our era.
Therefore, surgeons won’t be replaced by AI, for two obvious reasons:
- they need to apply their knowledge against a lot of in situ context
- their job requires a certain level of manual dexterity
- before and after surgery, they have to be the human face
None of which will likely be available for the AI in the coming years.
Context-free jobs at risk
Remember my question? If you were able to become productive while being a junior or medior employee in a few days, that was because your job didn’t need a lot of context to keep track of: it was just about how you were trained, or perhaps you even got a day or two training, like how to operate a register and that was it.
Being a waiter / waitress / Apple Store Genius isn’t at risk: you are the human face for an otherwise faceless service (cooking a meal, or being a multinational technology conglomerate). Being a marketing trainee? That’s a problem, however.
You see, all training and junior programs are at risk, especially if it can be done remotely, via a computer: these are jobs you can do without too much context, as simply, you don’t yet have any. At a certain point, instead of your younger self, a bot will accept the Slack message you have received, only to create that marketing email or slogan you developed in your early days.
However, it poses another problem: how will someone become a senior marketing manager, if marketing trainee and junior marketing manager jobs are taken away by robots?
It isn’t the first time machines replace humans
Replacing humans and animals with machines has been done since the invention of steam engine: who would like to be a carriage of a royal, if we can simply burn ancient trees for that? A lot of jobs have been automated since the dawn of the new age, and this one wouldn’t be too different.
The first exodus in IT: women of the early computers
The word computer originally didn’t mean a machine: instead, they were hairdressers unemployed by the French revolution: this division of labour was the model of the first mechanical computer (an all-metal lovechild of Ada Lovelace and Charles Babbage), then it was applied to women handling the first machines helping to decipher secret messages of world-war-era enemies and calculate parameters of the nuclear bombs. Only later on did the machines become self-programmable, replacing this menial task with pure automation, allowing women to take more meaningful roles in computing, like landing us on the Moon.
When the secretaries were replaced
The exodus of workforce from computing didn’t end with the ENIAC, per se: did you ever have a secretary?
If not, your secretary was likely replaced by a Microsoft product: Outlook. It keeps you organized with inbox and outbox, helps you schedule meetings with your colleagues, and gives you the possibility to remind you of your tasks. Of course, you had to learn to type yourself, albeit Microsoft Office has a feature for that, for a few years already. Of course, some professions require a specific vocabulary, but there are industry players for that.
Some people still do employ secretaries however: nowadays, they are called “office manager”, “project administrator”, “head of cabinet”, etc, and their job is to provide a human interface to an underlying organization. While some of them will never go away completely, it is likely that in a few years, you can make an Airbnb booking with SIRI.
Don’t fight, embrace
Obviously, our first reaction could be becoming a Luddite, despising everything technology — quite a hypocritical, but not totally unheard of behaviour.
On the other hand, our job was nearly always to raise efficiency through automation: all the “user experience”, “ergonomics”, “usability” is about lowering the amount of human labour needed to do for achieving a result.
The way to do this was to learn about context: call it context-of-use, call it personas, call it experience map, customer journey or service blueprint, or simply macro environment, it’s all about understanding the nuances, not encoded in the request, but expected by the client of a service.
The tools are out there, only this time, we will have to aim for one level higher: instead of helping a certain person achieve their tasks easier, we will have to make their bosses achieve the same results — without any human employee involved.
Let’s hope it brings a better world for all of us at the end.
