I listened to this interview between Tim Ferriss and Kevin Kelly and he described ChatGPT in a way I really resonated with: as the Universal Personal Intern. His analysis is that ChatGPT is effectively culminating the wisdom of the crowd into a series of tools that is capable of providing the same level of work you could expect from an intern: highly adept at producing a high volume of mediocre ideas that then need to be tuned and adjusted before they are ready for primetime. He then described it as "not bad", or the universal not bad personal intern (UNBPI).
I really like this assessment and it aligns with my own (limited) experience with using ChatGPT.
I've been astounded as the proficiency and usefulness of ChatGPT as a whole, and it really is a step-change in the level of functionality that is publicly available (for free!) to anyone who wants to use it, but I must agree with KK on this: the quality of output is just not that high.
There are countless stories of it "hallucinating" and confidently presenting false information, or writing complete essays that are built on false pretenses and misunderstood sources. This is dangerous for those who don't understand what they are doing, and expect the system to be more reliable than it is. It's like giving a power tool to a pre-teen: it's incredibly powerful, but only useful if used correctly.
However, the way I like to think about ChatGPT is as a first-year intern, or something that is fully capable of producing elementary analysis of complex topics and providing that information back to me to make a decision. I consider it an assistant who is able to quickly search the internet for what currently exists, and provide options for review. I would never allow this system to publish without review, hence my comments above, but it is incredibly effective as a generative tool.
So here are a few things I've used my UNBPI for in the last month:
- Creating recipes that align with my target diet
- Creating a workout plan based on my goal of running a marathon this year, taking into account my schedule
- Creating templates for standardized documents, like product requirements documents and executive proposals
- Researching competitive products and their positioning within the market
- Producing a list of blog post ideas related to various topics, and headlines on those topics
- Writing a SQL query to accomplish a specific task I couldn't figure out on my own (side note: ChatGPT as a paired coding tool is unbelievably powerful, this is worthy of its own blog post)
- Producing ice breaker questions for our team summit and fun interview questions for candidates
- Reviewing, editing, and summarizing blocks of text to condense them and make it easier to read
Going forward, I believe AIs (plural, because they are different kinds of machines) will influence our world in new and unexpected ways, but I tend to agree with KK that they are going to be incredibly adept at a limited range of things, requiring a more holistic review by humans.
Great as an intern, poor as an IC.