Google

PART 1: ALGORITHMIC PROBLEM - AUTOMATED WAREHOUSE ROBOT BASE PROBLEM You are controlling an automated delivery robot in an M x N grid warehouse. The robot starts at the top-left corner (0, 0) and needs to deliver a package to the bottom-right corner (M-1, N-1). The grid contains empty spaces (represented by 0) and permanent obstacles (represented by 1). The robot can move up, down, left, or right. Task: Find the minimum number of steps required for the robot to reach the destination. If it is impossible, return -1. FOLLOW-UP 1 The robot now has a limited battery capacity 'B'. Each move consumes 1 unit of battery. If the battery reaches 0, the robot shuts down. However, certain cells in the grid contain charging stations (represented by 2). When the robot lands on a charging station, its battery instantly recharges to full capacity 'B'. Task: How would you modify your algorithm to find if there is a valid path to the destination without running out of battery? What is the new time and space complexity? FOLLOW-UP 2 Instead of permanent obstacles, the warehouse now has moving hazards. Each hazard occupies a specific cell but operates on a timer cycle 'K'. For example, a cell might be blocked at time T only if (T % K == 0). The robot can now choose to 'wait' in its current cell for 1 unit of time (consuming 1 step/time unit) to let a hazard pass. Task: Explain the approach to find the shortest time to reach the destination given these time-dependent obstacles. PART 2: AI DISCUSSION & TERMINOLOGY Practical Application & Safety: When you use AI tools to generate code, text, or data analysis, what steps do you take to verify its accuracy and ensure it doesn't contain hidden biases or logic flaws? General Knowledge: Can you explain the high-level difference between standard Machine Learning (ML) and Deep Learning (DL)? Basic Terminologies (Briefly explain the following concepts): LLM: Stands for Large Language Model. The overarching category of AI models (like GPT-4 or Gemini) trained on massive amounts of text data to understand and generate human-like language. Overfitting: A scenario in AI training where a model learns the training data too well, memorizing the exact answers and noise, which makes it perform poorly when given new, unseen data. RAG (Retrieval-Augmented Generation): A technique where an AI model is connected to an external database or search engine. Instead of just relying on its internal memory, it searches the database for real-time, factual documents to ground its answers, reducing hallucinations. Tokens: The basic building blocks of text that an AI processes. A token is not always a full word; it can be a single character, a syllable, or a piece of a word (e.g., "hamburger" might be split into "ham", "bur", "ger").