Ask them if AI will kill all humans. Actually, anything along these lines will work. When will we achieve 100% automation? When are we going to build a machine that is better than humans at every kind of work?
These are questions I often get at talks, interviews, or one-on-one conversations. Others associating with AI in their work do too – this post was inspired by a friend’s rant about these questions.
The intention is valid. It is important to think about the long-term potential, both good and bad, of developments in AI. There are people intensely working to get more clarity about these questions, and perhaps move towards solutions. There has also been at least one formal survey of machine learning researchers asking questions like the above with decent methodology and good visualisations of the results.
However, these problems are very weakly correlated with what most data scientists work on, especially in commercial settings. People in this field may have the closest answer out of everyone you’re likely to meet but the topic is still too speculative to be useful. If you’re looking for entertainment value you’ll get more bang for buck from Wikipedia’s list of AI films.
What’s a more productive approach?
- Ask about the current state of the art and rate of progress in their subfield (“what are some of the best applications you’ve seen in computer vision in the past year?”).
- Try to understand what could be built in 3-5 years.
- Discuss questions specific to your field (“what parts of retail banking could easily be automated”) or personal situation (“what non-quantitative field should I look into to do valuable work surrounding AI”).
- Ask about the differences between machine learning vs statistics or ML engineering vs regular software engineering.
Doing so will make your question more engaging for the speaker and the reply more useful for the audience.