Hacker News new | ask | show | jobs
by mkolodny 1176 days ago
When a problem can be solved mindlessly - with a repeatable set of steps no matter the situation - imperative/declarative programming makes sense.

Most real world situations are unique and require unique solutions. That's where AI really shines. You just describe your target, attempt to solve the problem, and pay attention to how far off you were from your target. The learning happens naturally.

Neural networks are too complex - sometimes billions of variables - to decide what each neuron should do. We as a species have evolved to develop brains that are extremely adaptable. AI mimics our own natural learning process. And it's proven to be far more effective at solving unique problems.
2 comments
amarshall 1176 days ago
> [AI] has proven to be far more effective at solving unique problems

An extraordinary claim, I think. Source or evidence?

link
mkolodny 1176 days ago
https://chat.openai.com/
link
lostcolony 1176 days ago
When you've built a product using only OpenAI, and not a programming language, come back and tell us.

While at it, come back and tell us when you've implemented a neural net using AI, and not a programming language, come back and tell us too.

link
mkolodny 1176 days ago
https://github.com/features/copilot
link
lostcolony 1176 days ago
You don't think code is involved in that my dude? The input wasn't a description into an AI model, with copilot the model. It's code that calls to an ML model. Which is my point. The model was created with code. The model is deployed with code. It runs on infrastructure that executes a bunch of other code. Code takes in user input, puts it into the model, takes the resulting response, does something with it. There is no ML model that is a product in and of itself. ML is not a compiler. It is not a runtime environment. It does not understand business needs, it does not take direction. It certainly may be transformative, for good or ill, but it hasn't suddenly deprecated the need to code; far from it.
link
astrange 1176 days ago
Neural networks explicitly don't work the same way as human learning; they don't have online learning, humans definitely don't learn through backprop, humans have memory and compute in different parts of the brain, etc.

Also, training a neural network can make it worse; it's actually the combination system of the model and its engineers that makes it improve. (https://en.wikipedia.org/wiki/Catastrophic_interference)

link