I were 17, I'd learn how to build LLMs from scratch (twitter.com)

Article summary

Paul Graham suggests that if he were 17, he would learn to build large language models (LLMs) from scratch and train them with available hardware. He believes this foundation of knowledge would be beneficial for future startup ideas. Graham advises against trying to start a startup at 17, instead focusing on building a strong understanding of LLMs. This approach would allow for better startup ideas to emerge later on.

Main themes

• LLM development
• career advice
• startup strategy
• AI impact
• education and skills
• job market and opportunities

What commenters say

• Learning to build LLMs from scratch is a valuable skill, but it may not lead to many career opportunities.
• The field of LLMs is already saturated, and training another model may not be impactful.
• Understanding the fundamentals of LLMs is essential for making progress in the field, but it requires a strong foundation in math and programming.
• The advice to learn LLMs from scratch is overly simplistic and ignores the complexity of the field.
• The value of learning LLMs lies in understanding the underlying technology, not just in building a model.
• The job market for LLM experts is limited, and most companies do not need or cannot afford to train their own models.
• Learning LLMs can be a valuable educational experience, but it should not be seen as a direct path to a successful career.
• The emphasis on LLMs may be overhyped, and other skills, such as learning a trade or understanding business, may be more valuable.