Why building big AIs costs billions - and how Chinese startup DeepSeek dramatically changed the calculus
Briefly

DeepSeek, an AI startup from China, is shaking up the AI landscape by demonstrating that substantial financial investments are not the sole path to creating advanced AI models. Instead, they emphasize efficiency as a key strategy, challenging expectations set by industry giants like OpenAI and Google. Most large language models require extensive pretraining using massive datasets and powerful computing resources, predominantly utilizing GPUs. However, DeepSeek's approach suggests a possible shift in how AI can be developed, proving impactful results can emerge from more economical methods.
DeepSeek's disruptive debut comes down not to any stunning technological breakthrough but to finding efficiencies in AI development, challenging the vast investments by major players.
Developing powerful AI systems starts with building a large language model that predicts the next word based on given words, requiring substantial data and computing power.
Read at Inverse
[
|
]