Eyes
The name of a news network, this type of neural network is used for vision-related tasks
What is a CNN?
Cars to robots and back, this type of model is used for language
What are transformers?
What is deep?
This type of learning is used to train game-playing agents
What is Reinforcement Learning?
Controls how fast you pick up new knowledge
What is learning rate?
What is Pooling?
This is what GPT stands for
What is the Generative Pretrained Transformer?
Most common activation function for intermediate layers
What is ReLU?
Learning from demonstrations; parrots do this
What is Imitation Learning?
Character in The Bible, well-known optimizer.
What is Adam?
Unclear speech, this layer applies a filter to sliding windows of the input image
What is a convolutional layer?
Throw a tennis ball and a dog will bring it back. Search chatbots like Bing are also known as _ augmented LLMs
What is Retrieval?
What are loss curves?
What are Q-values?
French for together; averaging results from multiple models to get better performance
What is ensemble?
Something left - this technique helps to reduce performance degradation from having too deep neural networks
What are residuals?
Base language models are trained with this objective. The cat jumped over the _?
What is next token prediction?
Models that train too long on something might run into this
What is overfitting?
Foreign _, this is the entity that takes actions given states
What is a policy?
This technique is used to fit language models into small amounts of memory.
What is quantization?
Vision Transformers split images into these things, wear over eyes
What are image patches?
This is how you fine-tune with human preference data
What is Reinforcement Learning from Human Feedback?
Predictions often follow this pattern when RELUs die
What is identical?
Rotten Tomatoes; these things offer feedback to actors to further their learning
What are critics?
Plants, watermelons, and VC funding. This is set to ensure reproducebility of experiments
What is a seed?