Me, My Sanity and AI: Some Gibberish
Recently, I listened to a lecture on AI, and the speaker provided use cases for AI not just in research but also in day-to-day life. The talk included a demonstration of how wonderful AI is, and how we can use it to reply to emails, post on social media, create videos of ourselves as avatars, write research papers, and finish work that would otherwise have taken us 6 months in an hour. I am an early adopter of AI and have used it for a variety of tasks, read and thought deeply about AI disruption, but the talk was the spark I needed to write on AI, but as a skeptic.
Definitely not another blog to say all the wonderful ways to ‘automate’ your life, but to ponder on what it means to be human before allowing AI into different spheres of life. My only hope as I write this is not to be remembered as a generation that outsourced sanity to AI.
What Not to Outsource: Six Critical Use Cases
Do not use case 1: AI for relationship advice of any kind. Humans are social animals, and a big part of being human is interacting with the people around them. With the internet and instant messaging apps, the way we communicate has already shifted from face-to-face to instant messaging. However, the new trend is to feed these chat messages into AI and ask it to advise on how to respond. In any type of relationship, including friendship, arguing, misunderstandings, fights, etc., are part and parcel of it, and that is how Emotional Quotient develops. If we decide to outsource this even, then we need to redefine what it means to be a Human. I can't imagine living in an era where I have to question whether the message I got from my friend is actually from the person or the AI in disguise.
Do not use case 2: AI for medical advice. Geoffrey Hinton predicted radiologists would be replaced by AI, but that prediction missed how the job actually works. But the opposite has happened in many places: demand for radiologists has grown. That is because radiology is not just about reading scans or X-rays; it also involves clinical judgment, communicating findings to other doctors and patients, making judgment calls in ambiguous cases, and showing empathy in difficult situations. Even medical science is not just about symptoms and cures. Medical Practitioners are trained for years to observe patients, take visual cues, communicate, understand, and diagnose. AI can provide shallow, most-likely-to-be-true answers, but such answers in the medical field could be very harmful. Self-medication, with or without AI, could be dangerous.
Do not use case 3: AI in autopilot mode for writing. Writing is a form of thinking, and as you write, you get clarity. I agree that spell-check and Grammarly are forms of AI, but with chatbots, we often see the entire writing process outsourced to AI. "Write a 2000-page essay on X or Y" kind of prompts will produce a grammatically correct article, but it is too verbose, robotic, and often fails to engage with readers. Apart from the written diarrhea it can produce, the person who wrote it is unlikely to remember even a sentence. Doing this over and over again kills the critical thinking, often the most valuable skill in the era of AI!
Do not use case 4: Using AI for reading. I have seen people use AI to read papers or even read novels! It really bemuses me to the core. I do use AI to engage with papers: I read the paper, upload it to AI, use chat to test my understanding of it, or to identify its limitations. But never on autopilot, summarising the entire paper to me. With AI, human expertise will be more valuable than ever before, because verification is the limit. Reading is important for building critical thinking skills, and outsourcing it to AI may feel like saving time for now, but in the long run, it will be disastrous for our ability to comprehend and think.
Do not use case 5: Updating sensitive data to AI chatbots. With AI, we never know what information is encrypted and what gets used to train the model. So never use AI to draft a confidential document (say, for a government or a company) or to handle confidential data that contains personal information.
Do not use case 6: Relying completely on AI for trading advice. It is okay to learn stock market trading skills using AI, but asking AI to give stock recommendations or to trade futures is a really bad idea.
Final Thoughts
In summary, AI is just a very powerful tool. The tool is always neutral, neither good nor bad. It always depends on the person who uses the tool. Before asking if the AI is capable of doing a given task, ask yourself if I should be using AI for this task or not. Or else, just like AI-written text, we risk becoming hollow, lacking any depth of character. Natural stupidity in the era of Artificial Intelligence worries me, and I hope I am proven wrong.