[mm007emko]

LLMs like (Chat)GPT won't take your job. Software engineering is a creative job. Language models are not creative. Imagine that instead of using SQL to query a database of knowledge you use plain English. They can't create anything on their own. If they interpolate from their knowledge, it's usually fine. If they extrapolate, the outcome is shit. For instance, ChatGPT has enough knowledge (training data) about Python and Java. It can generate unit tests for Python and Java functions. I threw a less-known language, Common Lisp, the other day on it. It failed miserably (calling functions which didn't exist, calling with wrong number of parameters, it even imagined a mocking library which doesn't exist at all!). The more I know about machine learning and artificial intelligence (I study that stuff in a Ph.D. programme) the less I am worried. AI isn't dangerous. The only "intelligence" in AI is in its name. Nowadays machine learning really boils down to matrix multiplications (and set of (usually quite static) rules which can change the numbers in the matrices). People not understanding it and misusing it are. Does the vast majority corporate managers understand AI? We know the answer.

There is a ChatGPT craze right now but it will cool down eventually. It's not the first time this is happening. Google changed the way we search the internet, suddenly everything can be at your fingertips. Teachers didn't lose their jobs and people still go to school since you can't google things when you don't know the fundamentals of the topic. Chatbots changed the way we approach customer support. Support people didn't lose their jobs (at least entirely), since chatbots are subpar experience. If you are 20 years in, you can probably remember Visual Basic 6 - a tool which everybody can learn and write programs easily. We, software engineers, didn't lose our jobs.