|
|
|
|
|
|
New AI Technique Speeds Up Language Models on Edge Devices
|
|
Transformers are big models that have historically been difficult to use on small devices. They are state of the art for many inference tasks, and so the sooner we can use them on edge devices, the better. IBM thinks it has come up with a significant step in the right direction with Hardware-Aware Transformers (HAT).
|
|
|
|
|
|
|
|
|
AI Skills
|
|
|
|
Explainable AI: Neural Nets
|
You need to learn how to explain the predictions your models are making, otherwise it's hard to take them seriously. This blog is an essential how-to in explainability. Read more
|
|
|
|
Tools & Libraries
|
|
|
|
OpenAI Debuts Gigantic GPT-3 Language Model with 175 Billion Parameters
|
GPT-3 was announced last week, and OpenAI published a paper on the model's SOTA results. It's 175 billion parameters may be a milestone, or it may be an empty monument. Read the paper and let us know your thoughts. Read more
|
|
Your New Remote Data Science Toolkit
|
COVID-19 is keeping most data science teams out of the office, but that does not mean you have to give up being productive. IBM has launched a Remote Data Science site to help smooth your delivery process. Read more
|
|
|
|
Solutions & Products
|
|
|
|
Economics of Scaling Machine Learning Workloads (II) : Architecture Lessons from Data Science Engagements
|
This is the second post in a blog series detailing lessons in architectures from conversations with clients. It reviews the best practices in building a scalable and economical pipeline. Read more
|
|
|
|
Research
|
|
|
|
Making Decision Trees Accurate Again: Explaining What Explainable AI Did Not
|
This novel architecture from Berkley Artificial Intelligence Research (BAIR) tries to match the accuracy of Neural Nets to the interpretability of Decision Trees. This could be an important milestone in learning to justify predictions. Read more
|
|
|
|
|
See all upcoming community events here. |
June 4, 6–7 pm PT | Online Meetup
Join us for our fun, informal office hours to get you up and running on the community. Get a quick guided tour, meet other data science enthusiasts, share your questions and ideas. Perfect for new community members!
|
June 9, 8–9:30 am PT | Virtual Meetup
|
|
Virtual Meetup | June 10, 5–7 pm London, UK (GMT+1)
This webinar will show how deep learning can be used to help align developers and data scientists to a particular framework.
|
June 11, 1 pm EST | Webinar
Python Slow? 3 Ways to Speed Up Your Code with Data Science Engineer Joseph Gibli
|