The AI Fork: Levelling up or tuning out
I remember hesitating to ask follow-up questions in school—either due to fear of appearing dumb or knowing the teacher might not be interested in going deeper due to time constraints.
Having a patient teacher was a dream then. With AI, that dream is reality.
I can query: "Teach me like a 5-year-old", "Teach from a different viewpoint", and so on. Now, we have a teacher who doesn't get annoyed and teaches with patience at your level.
So, this brings us to a dichotomy: an AI fork, if you will. We can use AI to advance our knowledge or to delegate our thinking. There are multiple areas where I can see this playing out.
Here are the domains where I frequently use/misuse AI, and I think the correct usage will be a net positive sum for humanity. On the other hand, the incorrect usage might drive us down the value chain.
Learning: AI can help us learn to code, understand ML papers, get us up to speed on a ton of topics. However, an ideal way is to understand the code/context rather than copy-pasting it verbatim. Recent research shows that having AI give you answers actually lowers test scores when the AI is removed (source). Using a Socratic prompt works quite well here.
Peak Performance Coaching: Imagine having a peak performance/therapy coach available to you 24/7. We can query an LLM, maintain an ongoing chat, or even journal our progress. With a custom prompt, we can even give a voice or an identity to it. LLMs have memory and can give us advice based on our chat history.
Communication and Writing: AI is wonderful to bounce ideas. I use it for getting parenting advice from popular books and also to critique my writing. The flip side is I might lose my 'voice' if I use it heavily.
Critical Thinking: Constantly querying AI on how to do something better will give good results. AI can generate multiple ideas that can surface our own blind spots in thinking. Not trusting everything an LLM says—and doing some due diligence—is important.
In this era of job uncertainty, the people who might thrive are those who use AI to increase their overall skill set rather than offloading our thinking.
Caveats : We should not lean on chatbots for support for serious emotional issues. This leads to extremely adverse consequences. Also, LLM is known to exhibit sycophantic tendencies. So, we have to be careful in deciding where it is helpful and where it is not.