When I first encountered AI in the general sense, it was through ChatGPT. Obviously, I had known about and been exposed to AI systems(spam filters, algorithms, chatbots, etc) well before that, but with the launch of GPT-3.5, it became possible for the general public to wield it.
In this article, I’m going to use AI, LLMs, and other adjacent terms interchangeably instead of sticking to their technical usage.
In November 2022, I got my hands on ChatGPT 3.5. I got access to it as a research preview. It had yet to be released to the general public. I was fascinated and amazed by it, to say the least. I was prompting it to write haikus, make diet plans for me, and trying to talk to it as if it were a human being. I used it—or rather, overused it—for a few months, and then I really got bored. I understood its limitations, how good it was, and how dumb it could be. By the time GPT-4 was released in March 2023, I was already bored with it.
And that was fine. I had seen a few tech cycles before with crypto, NFTs, and similar trends. So I assumed it was going to fade away in a few years. But what happened next was truly unthinkable.
AI, AI, AI—everywhere AI.
There’s hardly any company that has not, in some way, marketed itself as being “powered by AI” or used some other AI-related slogan. And it’s fine for things that genuinely require AI. For example, it’s very good at detecting patterns in financial transactions and identifying anomalies, which can then be flagged and reviewed by a human. (I don’t think these systems should be given autonomy.) Another example would be sifting through huge amounts of data to find what you need, or scanning network logs to identify anomalies and report them. Basically, anything that requires relatively little critical thinking and is mundane and repetitive.
All of these things can also be done by writing a Python script, but an LLM provides more dynamism. A Python script is hard-bound by the rules written by me, whereas an LLM has much broader knowledge and can recognize things that haven’t been explicitly hardcoded.
My resentment towards AI started growing. For me, it wasn’t about whether AI was good or bad; it was primarily about how hyped-up it was—and still is. People consistently overestimate the current capabilities and capacities of AI. When people like Elon Musk say that work will be optional by 2035, the panic and anxiety among the general public are understandable. Elon Musk isn’t the only one living in a fantasy land—almost all of the tech bros keep amplifying AI-related FUD from time to time.
Now, there’s really no doubt that AI is a revolutionary invention, and it’s definitely going to play a substantial role in shaping the future of humanity. But that future is yet to come. In its current state, many AI ventures are burning through money, resources, and whatnot. It’s almost comical how inefficient AI currently is. I also understand that AI is still in its nascent phase, and that the fear and anxiety associated with it have accompanied other transformative inventions, such as electricity. Eventually, much of this inefficiency will be managed.
But my problem is that it’s incredibly overhyped. Much of this progress could still be achieved without all the FUD. Without driving up RAM and storage prices. (I can’t even buy an SSD right now.) Without making basic necessities more expensive for the general public. Electricity prices and other energy costs are also likely to rise eventually as a secondary effect.
Earlier, I was almost entirely negative about AI and its usage. But since it keeps getting better—if not day by day, then certainly year by year—I have found some of its use cases genuinely beneficial. And that’s why I don’t have a definitive opinion on it.