Working memory for current task state in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Role of Working Memory in Task State Management

Which statement best describes the role of working memory in managing the current task state in an AI agent?

A It temporarily holds relevant information needed to complete the current task.

B It stores all past experiences permanently for future reference.

C It deletes all previous task information to avoid confusion.

D It predicts future tasks without using current information.

💡 Hint

Think about what information an AI needs right now to perform a task.

Working memory temporarily holds information relevant to the current task, allowing the AI to focus and update its state as needed.

❓ Predict Output

intermediate

Output of Working Memory Update Code

What is the output of this Python code simulating a working memory update for task state?

Agentic AI

working_memory = {'step': 1, 'info': 'start'}

# Update working memory for next step
working_memory['step'] += 1
working_memory['info'] = 'processing'

print(working_memory)

A {'step': 1, 'info': 'start'}

B {'step': 1, 'info': 'processing'}

C {'step': 2, 'info': 'processing'}

D {'step': 2, 'info': 'start'}

💡 Hint

Check how the dictionary values change after the update.

The 'step' key is incremented by 1, and 'info' is updated to 'processing'.

❓ Model Choice

advanced

Best Model for Maintaining Working Memory in AI Agents

Which model architecture is best suited for maintaining and updating working memory of the current task state in sequential decision-making?

A Autoencoder

B Feedforward Neural Network

C Convolutional Neural Network (CNN)

D Recurrent Neural Network (RNN)

💡 Hint

Consider models that handle sequences and remember past inputs.

RNNs are designed to process sequences and maintain information over time, making them ideal for working memory tasks.

❓ Hyperparameter

advanced

Hyperparameter Affecting Working Memory Capacity

Which hyperparameter most directly affects the capacity of an RNN to maintain working memory over longer sequences?

A Dropout rate

B Number of hidden units

C Batch size

D Learning rate

💡 Hint

Think about what controls the size of the memory in the network.

The number of hidden units determines how much information the RNN can store and process at each step.

🔧 Debug

expert

Debugging Working Memory Loss in Agent Code

Given this code snippet for updating an agent's working memory, what error causes the working memory to reset unexpectedly?

class Agent:
    def __init__(self):
        self.working_memory = {}

    def update_memory(self, key, value):
        working_memory = {}
        working_memory[key] = value

agent = Agent()
agent.update_memory('task', 'start')
print(agent.working_memory)

A The update_memory method creates a new local dictionary instead of updating the instance variable.

B The print statement is outside the class and cannot access working_memory.

C The working_memory dictionary is never initialized in the Agent class.

D The key and value parameters are not passed correctly to update_memory.

💡 Hint

Check variable scope inside the method.

The method creates a new local dictionary named working_memory, which does not affect the instance's working_memory attribute.