After hearing hype around machine learning eating up tech jobs, I've wondered if I just have my head in the ground about what kind of work I'm doing. So far, I haven't seen a straightforward way to apply machine learning to my work goals.
However, today I realized I can start a mind shift toward my work that will be both more healthy for me personally, and will allow me to see opportunities for machine learning that I've been missing.
There are 4 main levels of work I can see:
- Initiative
- Project
- Task
- Micro-task
As an individual contributor or team lead, the aspects of the work are scoped down to levels 2-4. Most of the opportunities I've seen for applying machine learning have been at the Initiative level, where statistical methods can be applied to solve novel problems or automate whole classes of new/existing work.
Now, imagine you already work at levels 2-4. Imagine you have a computer with infinite intelligence. Imagine that if you describe a piece of work that needs to be done down to the 80% level of precision, that this intelligent computer could get it done and take care of the details. This is what I've found impossible to imagine in advance.
The best way I've found to imagine this is as follows:
- do a Task the way I've always done it (level 3)
- break the task up into atomic pieces by creating self-standing git commits as I go (one commit per level 4 Micro-task)
- when I write each git commit, imagine that I had asked a computer to do what I just did
- write the git commit in imperative form, as if I had commanded the computer to accomplish the work at the 80% level of precision
- imagine that I had spent the last 30-60min doing something other than solving the problem, and ask myself, "What else would I have been able to do while the computer was spending 15-30min on this?"
This helps me wire up the neurons in my own brain to start thinking about the computer as an intelligent agent that can assist me. And it helps me to imagine how I can use my attention and energy more effectively, rather than just solving micro-tasks all day.
It appears to me that applying machine learning effectively requires stepping back at least to the Project or Initiative level and delegating more work to an "intelligent" computer. And if such an agent doesn't exist, perhaps it can be built.
In a weird sort of way, I'm starting to use the
Tell,
Don't Ask principle in my own thinking to enhance my ability to imagine solutions coming together more quickly. I know that's taking the Law of Demeter way outside of its traditional scope. But I'm trying to break out of the box here. ;)
If 2018 was a year I didn't write any blog posts, it was certainly a year of great personal growth. I look forward to writing here more during 2019.