AI = Automated Interns

Earlier this summer, I gave a talk in London about decision-making, explaining how our minds make decisions, why the environment most of us are in doesn't support healthy decision-making, and also shared some strategies for making good decisions.

One key component of making good decisions is strengthening our Reflective Mind (see the link for more info!), so I shared a journaling technique with the audience that I've used for a few years to help clarify my internal thought processes. The following is an edited transcript of a question I received about this

Question: What are your thoughts on using ChatGPT for journaling and self-reflection? Is that a helpful practice?

Response: ChatGPT basically exists to tell you what you want to think. I'm sure all of us have used ChatGPT - if you actually go back through everything it's said to you, you'll tell it the stupidest things and it will say, "That's a great idea!" You can go back and check! Just check - that's the whole point. It's just amplifying your identity.

Take a random idea, like quitting your job and moving to Peru to become an alpaca wool farmer. ChatGPT will say: "Great idea. I think you're deeply qualified to do this because you have such a passion for textiles." That's exactly what it'll tell you. And then it'll give you options like, "Here's how much flights are to Peru. You can use Google Flights or Skyscanner. Here's how you wanna start, here's how you buy one alpaca. And then you can buy two." It'll tell you here's what alpacas eat. It'll just feed you everything that you want, that rabbit hole that your mind wants to go down. It doesn't actually provide you with any definition of good.  

When it comes to journaling, it's not something that's done with other people involved - that's therapy. ChatGPT isn't a licensed therapist, and neither am I. This journaling exercise is meant for you to get clear on what's happening in your head, and that means time by yourself.

So no, I wouldn't recommend it.

There's been an uptick in news articles talking about the manic psychosis induced by excessive ChatGPT usage, especially because of its programmed default-agree with you. Here's Bloomberg, The Independent, and The New York Times documenting this. It doesn't put a check on your thoughts - it blindly obeys, outside of a few situations that usually involve physical violence.

To that end, I've been thinking about this wonderful article from Benedict Evans. It was published in 2018 and speaks about machine learning (the technique behind ChatGPT), but I think its conclusion still applies to how we can use ChatGPT: treat AI like infinite interns.

Ways to think about machine learning — Benedict Evans

Everyone has heard of machine learning now, and every big company is working on projects around ‘AI’. We know this is a Next Big Thing. But we don’t yet have a settled sense of quite what machine learning means - what it will mean for tech companies or for companies in the broader economy, how to

Benedict EvansBenedict Evans

The things AI does effectively nowadays - summarizing information, searching the entirety of the internet, even making artwork (tangent: nearly all the artwork for my posts is AI-generated, and I want to change that - I would gladly pay a human but simply don't know any artists, so please connect me with your artist friends) - is something that could be done by interns.

Sometimes the results might be surprising, just as an intern might surprise you. But most of the time, the work needs to be checked by someone senior. This is also why the entry-level job market for white-collar work is collapsing - here's The New York Times, Axios, and PBS talking about this. When AI does the intern work, entry-level jobs suffer, too.

So yes, I'm skeptical of AI usage. You'd never trust an intern with deciding everything for you.