Recently, Ethan Mollick published a piece called On-boarding your AI Intern. If there’s one thing you read about artificial intelligence this week, this should be it.
TL:DR; (but please do read the article!):
- LLM-powered bots are essentially gifted interns, with a wealth of capability and knowledge that they don’t fully understand how to use. It’s on us to figure out how to guide and leverage them to help us do our jobs better.
- A well-chosen intern can drastically improve your productivity and happiness.
- Choosing the right kind of intern for you is hard work! You want the right personality, cost, strengths/weaknesses to complement your own, etc.
Ethan Mollick is a professor of innovation and entrepreneurship at Wharton Business School (UPenn), and a regular in the AI community of conversation.
His article stood out to me for a few reasons:
- I think it’s wise to start thinking of LLM-powered applications like people, because they will likely play the roles of real people in your lives, we already know how to interact with people, so there’s no need to come up with a “how to talk to robots” vernacular, and;
- At some point, sooner than we think, we will create sentient AIs that are intellectually and emotionally indistinguishable from people. Let’s practice kindness and politeness with them early and often.
The most important question is what roles AIs will play in our lives (point A above).
- Are they going to do our jobs for us?
- Are they just better Googles or Alexas?
- Somewhere in between?
This will be partly about capabilities (are AIs good enough to do X?) but more about the roles we ask them to play.
For now (maybe for a limited time), we still have the privilege of defining their roles, and ours too.
My hot take based on “On-boarding your AI Intern”
The age of “the specialist” is over.
This has been said many, many times over the years (here’s an example from 2019), but this time it’s actually true.
Who’s more powerful – a laser-focused data analyst, or a generalist product manager who a) understands the bigger picture and b) is paired with an AI intern playing the role of…a laser-focused data analyst?
I may be biased, given my background in product, but give me the product manager every day and twice on Sundays.
The real question is whether traditional roles will survive.
I am a former coder who’s been doing product for 15 years; the last time I pushed production code was in 2007 on a Palm Treo.
Until…about 3 months ago, when I forked a Github repo, asked ChatGPT to write me a few new Typescript functions, used my former-coder common sense to plug it in, and voilá! LLMs make it much easier for anyone to be “full stack” and tackle multiple dimensions of a build on their own – especially if you’ve done the job before, just not recently or in this context.
Full stack may not scale, so you still want to have teams of people who work together well, but generalists who can do a little bit of everything and lean on an AI when they need to go deep will be incredibly valuable teammates.
That said, we should all expect a lot of role confusion as teams figure out how to use these AI tool sets efficiently and effectively.
Dan Mason
