First Principles or First Prompts?

Imagine a public servant gently nudging someone fast asleep on a park bench meant for people to sit and relax. He'd use a stick, not a knife or a gun. In their earliest stage, guns with all their charm would have positioned themselves as the best weapon under all circumstances. But slowly, people would have realized their folly.

SQL is the stick. Python is the knife. AI agents are machine guns.

Learning has never been more fun:

I like ChatGPT and frequent it regularly. I find it remarkably good at putting up with my petty questions - one on top of another - until I understand a concept from first principles.

It has made learning fun, easy, and effective.

But in rare cases, without learning the concept deeply, I piece together a solution with a couple of prompts just to get the task done and please my subjects. Not surprisingly, I feel hollow when I have to ask about it again at the onset of a similar task in the future - not a single input but a handful of them - time-consuming and building no knowledge.

Learn, then do.

Original thoughts only:

Art is under attack. Authenticity is questioned.

Any work sounding similar to an AI-generated one is under the radar of naysayers who are quick to label the artist as being 'unoriginal'.

I made a rough draft of this work on my OneNote over the weekends, and I have been itching ever since to post it on social media. My fear harbors around a secret API that fetches my words and passes them to an AI model as training data. Any later than a few days (or hours), my words become part of the mainstream internet and I the guy who stole from it. Better to post it ASAP!

Surely it must be in terms and conditions - but it’s a lot to read - except for people in high buildings in big cities in their neatly fitted clothes - talking about those lawyers in Suits. Anyway, terms and conditions are a good use case - read through terms and conditions and elaborate on the points that are potentially overlooked and dangerous - without having to copy-paste the whole thing into a prompt.

Palace made of wax:

Well known argument against AI is the accountability problem. Litigation could be drawn against lawyers or doctors for poor work, but not against a product that could potentially do some of their work?

It doesn’t help when the very people building it, hoping to appear realistic and trustworthy, caution against its potentially destructive effects, only to add that its rise is inevitable.

AI use cases are aplenty - some of which feel like darts thrown into the unknown, hoping to hit a target of essence. An app supposedly claims to predict how successful an individual will be based on facial structure. As you might have guessed, it often predicts incorrectly - the same person scores better in front of a bookshelf wearing glasses than without them. Simplistic enough. What Artificial Intelligence Can Do, What It Can’t

Withstanding its accuracy, it's novelty is captivating and the possibilities are endless. It's akin to wax clay in the hands of a kid. Can build cute things with it. But palace? Cautionary tale: the palace made of wax in the Mahabharata. It burned.

I may sound like an AI skeptic right now. But, I change my opinions faster than Elon patching with Trump.

Orginally published on LinkedIn