3 ms·

I was trying to keep it an apples to apples comparison. With 10+ more years the dev might have machine learning experience (or something else) which would qual

I was trying to keep it an apples to apples comparison. With 10+ more years the dev might have machine learning experience (or something else) which would qualify for more.

Judgmentality 8y ago
It doesn't take 5 years to teach yourself Machine Learning. While it is definitely a skill, it is not black magic. With a 5 year delta you could easily have ML experience, and a lot more as well to differentiate yourself.
- gaius 8y ago
  It doesn't take 5 years to teach yourself Machine Learning. The low-hanging fruit in ML is absurdly low these days, the tools are very good and there is loads of sample code available. If you were already good with Python or R, and already had some nice clean data, you could be doing something useful with Keras in a day, no joke.
stevesimmons 8y ago
My point is, it is not an apples to apples comparison. Older workers are most valuable when they can leverage more from their broader and deeper experience of the business domain. For instance, you would probably say I am only half as productive as the data scientists I manage. I only get to spend an hour or two a week doing hands-on modelling, so I just don't have the same familiarity with the techniques and tools. But in terms of influencing output, I can have a five minute chat about the goals and context to their work and save them a week of fruitless effort. So in practical terms, my productivity is far higher. I could easily spend a bit of time to improve my modelling skills a bit. But the law of comparative advantage suggests our junior engineers should build their skills while I coach, mentor and set strategic direction. In fact, to give another very practical example, this weekend my company Oakam is hosting the European Pandas core developers (https://python-sprints.github.io/europandas2019/ https://python-sprints.github.io/europandas2019/) in London. My own personal contribution to Pandas this weekend will be minor. But I am far more effective by being able to provide the venue and fund travel grants so 30 core developers can collaborate together.

I was trying to keep it an apples to apples comparison. With 10+ more years the dev might have machine learning experience (or something else) which would qualify for more.

Judgmentality 8y ago

It doesn't take 5 years to teach yourself Machine Learning. While it is definitely a skill, it is not black magic. With a 5 year delta you could easily have ML experience, and a lot more as well to differentiate yourself.

gaius 8y ago
It doesn't take 5 years to teach yourself Machine Learning. The low-hanging fruit in ML is absurdly low these days, the tools are very good and there is loads of sample code available. If you were already good with Python or R, and already had some nice clean data, you could be doing something useful with Keras in a day, no joke.

stevesimmons 8y ago

My point is, it is not an apples to apples comparison. Older workers are most valuable when they can leverage more from their broader and deeper experience of the business domain. For instance, you would probably say I am only half as productive as the data scientists I manage. I only get to spend an hour or two a week doing hands-on modelling, so I just don't have the same familiarity with the techniques and tools. But in terms of influencing output, I can have a five minute chat about the goals and context to their work and save them a week of fruitless effort. So in practical terms, my productivity is far higher. I could easily spend a bit of time to improve my modelling skills a bit. But the law of comparative advantage suggests our junior engineers should build their skills while I coach, mentor and set strategic direction. In fact, to give another very practical example, this weekend my company Oakam is hosting the European Pandas core developers (https://python-sprints.github.io/europandas2019/ https://python-sprints.github.io/europandas2019/) in London. My own personal contribution to Pandas this weekend will be minor. But I am far more effective by being able to provide the venue and fund travel grants so 30 core developers can collaborate together.