easyMultiple Choice
Generative AI Leader Practice Question: Best describes the transformer architecture's key…
Which of the following best describes the transformer architecture's key innovation that enabled modern large language models?
⚠ Common exam trap
The Google Gen AI Leader exam often tests the misconception that transformers are just an evolution of RNNs or that their innovation is about memory or local feature extraction, when the true breakthrough is the parallel self-attention mechanism that eliminates sequential processing.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Self-attention mechanism that captures dependencies between all words in parallel
The transformer architecture's key innovation is the self-attention mechanism, which allows the model to compute attention scores between every pair of tokens in the input sequence simultaneously, rather than processing tokens sequentially. This parallelization enables the model to capture long-range dependencies efficiently and scale to massive datasets, which is the foundation for modern large language models like GPT and BERT.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Recurrent connections that process sequences one element at a time
Why it's wrong here
Recurrent connections process tokens sequentially, preventing parallel training and limiting long-range dependency capture — the exact bottleneck transformers removed. They are tempting because RNNs and LSTMs were the standard sequence models, and recurrence would be correct for streaming or time-series data with strict temporal ordering.
- ✓
Self-attention mechanism that captures dependencies between all words in parallel
Why this is correct
Self-attention computes relationships between every token pair simultaneously, replacing the sequential recurrence of RNNs. This parallelism removes the sequential-processing bottleneck, allowing training on far larger corpora and enabling the scale that defines modern large language models. It directly satisfies the stem's requirement for the innovation underpinning contemporary LLMs.
- ✗
Memory-augmented neural networks
Why it's wrong here
Memory-augmented networks bolt external read/write memory onto a controller; they do not provide the parallel self-attention that scales transformer training. They are tempting because external memory aids long-range recall, and such architectures would be correct for tasks needing explicit storage, such as algorithmic reasoning or one-shot learning.
- ✗
Convolutional layers that extract local features
Why it's wrong here
Convolutional layers extract local spatial features through fixed kernels, giving no mechanism for global token-to-token attention across a sequence. They are tempting because CNNs dominate image recognition, and convolutions would be the right building block for tasks such as image classification or object detection.
Go deeper
Related to this question
About these practice questions
This Generative AI Leader question is part of Courseiva's 1,008-question bank — original exam-style content with full explanations and wrong-answer analysis, never real exam questions or exam dumps. Learn why practice questions differ from exam dumps →
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This Generative AI Leader practice question is part of Courseiva's free Google Cloud certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the Generative AI Leader exam.