Скачать или смотреть The End of Finetuning? A Conversation with Fast.ai's Jeremy Howard in Telugu

Is one of the most fundamental techniques in modern AI—finetuning—actually obsolete? According to Jeremy Howard, the co-founder of Fast.ai and the man who helped pioneer it, the answer is a resounding yes. In this video, we're diving deep into his provocative conversation on the "Latent Space" podcast.

We explore Jeremy Howard's incredible journey, from creating the groundbreaking ULMFiT technique (which became foundational for today's LLMs) to his current belief that traditional finetuning is a flawed approach. We'll break down his argument against it, focusing on the critical problem of "catastrophic forgetting," and explore his proposed solution: a more effective method he calls "continued pre-training."

In this discussion, we explore:

The story of ULMFiT: The transfer learning technique that changed NLP.

Why the pioneer of finetuning now believes it's the wrong approach.

The critical problem of "catastrophic forgetting" in AI models.

The future of model training: Understanding "continued pre-training."

This is a must-watch for any AI engineer or machine learning practitioner who wants to stay ahead of the curve and understand the future of training large language models.

🤝 Connect with Me
Instagram:   / wisdom.ai__  
LinkedIn:   / teja-rachakonda  
X (Twitter): https://x.com/rachakonda12403

#JeremyHoward #Finetuning #AI #LLM #Fastai #ULMFiT #MachineLearning #DeepLearning #AIResearch #AIHyderabad