• 0 Posts
  • 19 Comments
Joined 1 year ago
cake
Cake day: July 3rd, 2023

help-circle





  • It requires continuous expansive improvements. It is like real world. Building a system robust to frauds works on the short term, but on the mid and long term is impossibile. That is why laws change, evolve, we have governments and so on. Because system reacts to your rules and algorithms, making them less effective.

    And these continous expensive improvements are done daily, but it is a difficult job


  • It is not at the moment. Models are built on the assumption of stability, i.e. that what they are modelling doesn’t change over time, doesn’t evolve. This is clearly untrue, and cheating is a way the environment evolves. Only way to consider that, is to create a on-line continous learning algorithm. Currently this exists and is called reinforcement learning. Main issue is that methods to account for an evolving environment are still under active research. In the sense that methods to address this issue are not yet available.

    It is an extremely difficult task tbf





  • People isn’t considering that documentation has greatly improved over time, languages and frameworks have become more abstract, user-friendly, modern code is mostly self explanatory, good documentation has become the priority of all open source projects, well documented open source languages and frameworks have become the norm.

    Less people asking programming related questions can be explained by programming being an easier and less problematic experience nowadays, that is true.






  • The problem of current LLM implementations is that they learn from scratch, like taking a baby to a library and telling him “learn, I’ll wait out in the cafeteria”.

    You need a lot of data to do so, just to learn how to write, gramma, styles, concepts, relationships without any guidance.

    This strategy might change in the future, but the only solution we have now is to refine the model afterward, let’s say.

    Tbf biases are integral part of literature and human artistic production. Eliminating biases means having “boring” texts. Which is fine for me, but a lot of people will complain that AI is dumb and boring