Development Using Data Lakes and Large Language Models
Oct 20, 2023
auto_awesome
Davit Buniatyan, CEO and founder of Activeloop, discusses topics such as efficient training of large language models, the evolution of deep learning, limitations of large language models, and the future of AI in software development.
Developing with large language models can help understand how the brain works and reconstruct the connectivity of neurons.
Specialized storage solutions for deep learning and AI, like data lakes, optimize data transfer, GPU utilization, and reduce costs.
Deep dives
Large language models for neuroscience research
The podcast episode features David Banyatian, CEO and founder of Active Look, who discusses his background in neuroscience and computer science. He explains the importance of large language models and their applications in understanding how the brain works. David shares his experience of using deep learning and computer vision to reconstruct the connectivity of neurons inside a mouse brain, highlighting the challenges of cost and data processing. This prompted him to develop more efficient methods for machine learning operations, leading to the creation of DeepLake, a data lake specifically designed for deep learning applications.
Reducing costs and improving efficiency in AI development
David explains that training large language models can be costly and often inefficient due to underutilization of GPU resources and bottlenecks in data transfer. He shares how his company helped reduce the training time for a customer's large language model from two months to a week, while also optimizing compute and storage costs. David further discusses the need for specialized storage solutions for deep learning and AI, highlighting the benefits of treating images as multi-dimensional columns in data lakes. These innovations help streamline the data transfer process, improving GPU utilization and reducing costs.
Practical use cases for large language models
The podcast explores various practical use cases for large language models. Examples include image generation, where models like Dalle and Firefly enable the creation of images and videos that were not possible before. Another use case is chat GPT interfaces, which allow conversational interaction with language models for applications such as enterprise search, code understanding, and code generation. David emphasizes the potential value impact of these use cases and how large language models are commoditizing software development, making data collection and protection crucial for maintaining a competitive edge.
Preparing for the future of AI development
David addresses the concerns of professional developers regarding the impact of large language models on their roles. He emphasizes that developers' jobs are not at risk but rather evolving. Developers need to adapt by understanding how to condition models, collect training data, and fine-tune models for specific tasks. He discusses changes in development approaches, such as thinking systematically and collecting edge case data to improve model performance. David also emphasizes the importance of learning Python and JavaScript for AI development, as well as the availability of frameworks and certification courses for developers to upskill in generative AI.
In this podcast Shane Hastie, Lead Editor for Culture & Methods spoke to Davit Buniatyan, the CEO and founder of Activeloop about developing with large language models and AI.
Read a transcript of this interview: https://bit.ly/3rYQ5BW
Subscribe to the Software Architects’ Newsletter [monthly]: www.infoq.com/software-architect…mpaign=architectnl
Upcoming Events:
QCon London
qconlondon.com/
April 8-10, 2024
Follow InfoQ:
- Mastodon: https://techhub.social/@infoq
- Twitter: twitter.com/InfoQ
- LinkedIn: www.linkedin.com/company/infoq
- Facebook: bit.ly/2jmlyG8
- Instagram: @infoqdotcom
- Youtube: www.youtube.com/infoq
Write for InfoQ
- Join a community of experts.
- Increase your visibility.
- Grow your career.
www.infoq.com/write-for-infoq/?u…aign=writeforinfoq
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode