Episode
Jeremy Howard: fast.ai Deep Learning Courses and Research
1:44:17
Published: Tue Aug 27 2019
Description
Jeremy Howard is the founder of fast.ai, a research institute dedicated to make deep learning more accessible. He is also a Distinguished Research Scientist at the University of San Francisco, a former president of Kaggle as well a top-ranking competitor there, and in general, he's a successful entrepreneur, educator, research, and an inspiring personality in the AI community. This conversation is part of the Artificial Intelligence podcast. If you would like to get more information about this podcast go to https://lexfridman.com/ai or connect with @lexfridman on Twitter, LinkedIn, Facebook, Medium, or YouTube where you can watch the video versions of these conversations. If you enjoy the podcast, please rate it 5 stars on iTunes or support it on Patreon.
Chapters
This podcast discusses the practical application of Airtable and Fast AI.
00:00 - 05:13 (05:13)
Summary
This podcast discusses the practical application of Airtable and Fast AI. Airtable provides easy access to subsets of data and Fast AI enables hands-on exploration of cutting-edge deep learning that is both accessible to beginners and useful to experts.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
This podcast delves into the world of array-oriented programming languages, their history, and how they differ from traditional programming languages.
05:13 - 16:25 (11:11)
Summary
This podcast delves into the world of array-oriented programming languages, their history, and how they differ from traditional programming languages. The host explains how these languages can increase productivity and discusses how the community for these languages is niche but passionate.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
The development of domain-specific languages and higher level programming languages, such as Swift, aim to make tensor computations easier and enable more creativity with things like RNNs and sparse convolutional neural networks.
16:25 - 23:02 (06:37)
Summary
The development of domain-specific languages and higher level programming languages, such as Swift, aim to make tensor computations easier and enable more creativity with things like RNNs and sparse convolutional neural networks.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
This episode discusses the potential benefits of using AI for diagnostic and treatment planning in developing countries, particularly in areas with a shortage of healthcare professionals, such as India, China, Indonesia or Africa.
23:02 - 32:35 (09:33)
Summary
This episode discusses the potential benefits of using AI for diagnostic and treatment planning in developing countries, particularly in areas with a shortage of healthcare professionals, such as India, China, Indonesia or Africa.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
Experts discuss the importance of collecting data in order to build models for impactful products, such as recommender systems, without relying on individual data.
32:35 - 41:21 (08:45)
Summary
Experts discuss the importance of collecting data in order to build models for impactful products, such as recommender systems, without relying on individual data. They also ponder on the balance between data collection and privacy concerns in various fields, including medicine and social media platforms.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
Professor Jeremy Howard discusses the challenges of active learning and how transfer learning can make deep learning more accessible and efficient for people without large amounts of data and resources.
41:21 - 46:25 (05:04)
Summary
Professor Jeremy Howard discusses the challenges of active learning and how transfer learning can make deep learning more accessible and efficient for people without large amounts of data and resources.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
In this podcast, the host discusses an experiment that teaches AI to "see" low quality images and identify them clearly after training with high quality versions of the same image.
46:25 - 52:03 (05:38)
Summary
In this podcast, the host discusses an experiment that teaches AI to "see" low quality images and identify them clearly after training with high quality versions of the same image. Despite its limited relevance, this experiment could pave the way for better image recognition software.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
Lex Fridman discusses the accessibility of deep learning and the use of multi-GPU or multiple machine training to speed up code.
52:03 - 56:27 (04:23)
Summary
Lex Fridman discusses the accessibility of deep learning and the use of multi-GPU or multiple machine training to speed up code. Many people believe that deep learning is inaccessible to individuals outside of Google for useful work.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
This episode discusses the potential for cheaper lenses combined with intentional movement and how deep learning is evolving to allow for better subpic and resolution.
56:27 - 1:01:27 (04:59)
Summary
This episode discusses the potential for cheaper lenses combined with intentional movement and how deep learning is evolving to allow for better subpic and resolution. The guest also shares his findings on super convergence and the possibility of automatically combining audio from multiple sources to remove noise in devices like Alexa.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
Interpreting machine learning models can assist in quickly becoming a domain expert by focusing on important features and misclassifications, allowing domain experts to do more work rather than deep learning experts.
1:01:27 - 1:06:26 (04:59)
Summary
Interpreting machine learning models can assist in quickly becoming a domain expert by focusing on important features and misclassifications, allowing domain experts to do more work rather than deep learning experts.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
Theano and TensorFlow, being static graph frameworks, were harder to teach and research on, while PyTorch offers a more dynamic approach.
1:06:26 - 1:11:13 (04:46)
Summary
Theano and TensorFlow, being static graph frameworks, were harder to teach and research on, while PyTorch offers a more dynamic approach. Google Cloud, Salamander and PaperSpace offer fast and easy ways to access a GPU for Deep Learning development.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
The speaker compares writing TensorFlow code from scratch versus using pre-existing code and finds that pre-existing code can be inefficient and less programmable.
1:11:13 - 1:18:47 (07:34)
Summary
The speaker compares writing TensorFlow code from scratch versus using pre-existing code and finds that pre-existing code can be inefficient and less programmable.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
Jeremy Howard emphasizes the importance of running experiments to gain an intuitive understanding of machine learning, regardless of one's level of expertise in coding or AI.
1:18:47 - 1:24:32 (05:44)
Summary
Jeremy Howard emphasizes the importance of running experiments to gain an intuitive understanding of machine learning, regardless of one's level of expertise in coding or AI.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
The key to creating a successful AI recruiting startup is to have prior experience as a recruiter and be pragmatic about the startup journey.
1:24:32 - 1:32:38 (08:05)
Summary
The key to creating a successful AI recruiting startup is to have prior experience as a recruiter and be pragmatic about the startup journey. It’s crucial to keep costs low, save up some money beforehand, and choose a domain or dataset that you care for.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
The speaker reveals that they spend at least half of every day learning or practicing something new, which makes them more efficient in their work.
1:32:38 - 1:40:09 (07:31)
Summary
The speaker reveals that they spend at least half of every day learning or practicing something new, which makes them more efficient in their work. They have a specific approach to learning, which involves deeply understanding concepts and incorporating them into their work.
EpisodeJeremy Howard: fast.ai Deep Learning Courses and Research
PodcastLex Fridman Podcast
Host Rachel Thomas and guest Jeremy Howard discuss the importance of data scientists recognizing their role as educators and acknowledging the potential societal impact of deep learning and AI-driven technologies.
1:40:09 - 1:44:07 (03:57)
Summary
Host Rachel Thomas and guest Jeremy Howard discuss the importance of data scientists recognizing their role as educators and acknowledging the potential societal impact of deep learning and AI-driven technologies.