AI cannot replace programmers yet.

 I had gpt 4o/3.5 try to create a c program neural net library to process the mnist data set to recognize hand written digits.  It came close to writing the code, it created the framework, but the code didn't actually train a neural net.   So ai has a long way to go before it can replace programmers

You're right; while AI models like GPT-4 and GPT-3.5 can generate code, they sometimes struggle with complex tasks like implementing a complete neural network library. In this case, the model could create a basic framework, but it didn't manage to produce a fully functional program that trains a neural network.

There are several reasons for this:

Complexity: Writing a complete neural network library requires a deep understanding of various concepts and techniques in machine learning, as well as expertise in the programming language being used. Current AI models have limitations in handling such complex tasks.

Context: AI models may not fully grasp the context and requirements of a task, especially if it involves multiple steps and dependencies. This can lead to incomplete or inadequate code.

Lack of feedback loop: Human programmers can test their code, identify bugs, and iteratively improve their work until the desired result is achieved. AI models don't have this feedback loop and may not be able to refine their output effectively without explicit guidance.

While AI has made significant strides in code generation, it still has a long way to go before it can truly replace human programmers in handling complex projects. For now, AI models can serve as helpful assistants or tools for accelerating certain aspects of the programming process.