Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit

Problem Set 1

AI and ML

BUSOBA 7247

Spring 2023

Note: You do not need to turn in your answers to the problems from

HOML. For solutions, see homl.info.

– Chapter 1: Exercises 1, 3, 4, 8, 12, 14, 16, and 19

– Chapter 3: Exercise 1

Assignment Questions

Please turn in an HTML export of the Jupyter notebook you use to

solve these questions. Include answers to the specific questions as

markdown cells or Python comments.

Building on the application of k-nearest neighbors to the MNIST

handwritten digits dataset (see 02_knn_mnist .ipynb):

1. Use automation to find a good value of k for multiclass classification via cross-validation.

2. How would you characterize the performance of the model you selected (i.e., the nearest-neighbors model with your value of k)?

3. Would you expect your model to generalize to new instances similar to those from the MNIST handwritten digits database?

4. Describe the objective of the narrow AI you’ve created.

5. Going beyond the MNIST example: consider Stuart Russell’s ar- guments regarding the definitions of objectives and intelligence  (see, e.g., the passage under the heading“What Went Wrong?” on page 9 ofHuman Compatible). Discuss a situation wherein an  AI could achieve its objective —i.e., the objective the AI’s creators  specify —yet the AI’s actions do not align with the creators’or soci-  ety’s goals.