Are Bigger Language Models Better? | DeepMind Gopher and RETRO

What is the parameter size of DeepMind's Gopher model?

• 530 billion

• 208 billion

• 175 billion

• 100 billion

Which model is larger than Gopher in terms of parameters?

• BERT

• RETRO

• Megatron Turning

• GPT-3

In which areas did Gopher show improved performance?

• Art and Design

• STEM and Medicine

• Sports and Entertainment

• Finance and Economics

What issue does Gopher still face despite increased size?

• Limited data storage

• Increased computational cost

• Higher toxicity and bias

• Decreased accuracy

How does the CO2 emission from training Gopher compare to a transatlantic flight?

• Less than a flight

• No emissions

• Equal to a flight

• More than a flight

What is the parameter size of the RETRO model?

• 530 billion

• 7 billion

• 175 billion

• 208 billion

How does RETRO improve the quality of its output?

• By increasing parameters

• By using a retrieval database

• By reducing training time

• By using more data