AI0-001 AI Governance and Ethics Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Increase the batch size to the maximum the GPU memory allows

  Why it's wrong here

  Maximising batch size can improve GPU utilisation but often requires more epochs or learning-rate retuning to preserve quality, so energy savings are not guaranteed. It is tempting because larger batches raise throughput per step, and would be correct where raw training speed, rather than energy reduction, is the objective.

• ✗

  Use a larger model architecture to achieve higher accuracy faster

  Why it's wrong here

  Scaling up model architecture increases parameter count and compute, raising training energy rather than lowering it, so the carbon footprint grows. It is tempting because larger models can reach target accuracy in fewer steps, and would be correct where maximising accuracy, not minimising energy, is the priority.

• ✓

  Use mixed-precision training and prune unnecessary parameters

  Why this is correct

  Mixed-precision training halves memory and compute per operation, while pruning removes redundant parameters, cutting training energy and carbon emissions directly. Both techniques preserve model quality, satisfying the requirement to reduce energy consumption without degrading accuracy.

• ✗

  Train the model on CPUs instead of GPUs

  Why it's wrong here

  GPUs deliver the parallel throughput that transformer training depends on; CPUs would stretch training time enormously, raising total energy use rather than cutting it. CPU training suits tiny models or inference at the edge, where GPU cost and power are unjustified.