Amazon SageMaker continues to expand machine learning (ML) use in the cloud
Amazon Web Services (AWS) adds new multi-model engine capabilities to SageMaker as users continue to unlock hidden value in data.Read More
Report: U.S. loses AI leadership to India despite a 6-year head start
Peak’s DI Maturity Index found that India is now a more mature market than the U.S. or U.K. in readying their businesses to adopt AI.Read More
Got generative AI FOMO? Keep calm and carry on | The AI Beat
Enterprise leaders dealing with severe feelings of AI FOMO should focus on their business problems and what they are trying to solve. Read More
If Musk Starts Firing Twitter’s Security Team, Run
What’s next for the social network is anyone’s guess—but here’s what to watch as you wade through the privacy and security morass.
Non-Autoregressive Neural Machine Translation: A Call for Clarity
Non-autoregressive approaches aim to improve the inference speed of translation models by only requiring a single forward pass to generate the output sequence instead of iteratively producing each predicted token. Consequently, their translation quality still tends to be inferior to their autoregressive counterparts due to several issues involving output token interdependence. In this work, we …
Read more “Non-Autoregressive Neural Machine Translation: A Call for Clarity”
Prompting for a Conversation: How to Control a Dialog Model?
Dialog modelling faces a difficult trade-off. Models are trained on a large amount of text, yet their responses need to be limited to a desired scope and style of a dialog agent. Because the datasets used to achieve the former contain language that is not compatible with the latter, pre-trained dialog models are fine-tuned on …
Read more “Prompting for a Conversation: How to Control a Dialog Model?”
A Treatise On FST Lattice Based MMI Training
Maximum mutual information (MMI) has become one of the two de facto methods for sequence-level training of speech recognition acoustic models. This paper aims to isolate, identify and bring forward the implicit modelling decisions induced by the design implementation of standard finite state transducer (FST) lattice based MMI training framework. The paper particularly investigates the …
Vernetzte Versorgungsunternehme: Der Weg zu einer nachhaltigen Wasserwirtschaft
(An English-language version of this post can be read here.) Während sich das Klima verändert, die Bevölkerung zwar langsamer, aber zumindest mittelfristig unaufhaltsam weiter wächst und die globalen Ansprüche an den Lebensstandard steigen, ist sauberes Wasser eine der kostbarsten natürlichen Ressourcen. In den Industrieländern ist sauberes, fließendes Wasser eine Selbstverständlichkeit, und um das auch weiter zu …
Read more “Vernetzte Versorgungsunternehme: Der Weg zu einer nachhaltigen Wasserwirtschaft”
How Newcomp Analytics partners with IBM to advance clients’ supply chain insights
When Newcomp Analytics started working with chocolatier Lindt Canada more than 15 years ago to support their supply chain, Lindt had no full-time IT personnel for analytics. Lindt now has a team of 10, including a business intelligence (BI) manager and BI developer analysts. Yet Newcomp continues to be an essential and trusted partner, helping …
Read more “How Newcomp Analytics partners with IBM to advance clients’ supply chain insights”
Improve price performance of your model training using Amazon SageMaker heterogeneous clusters
This post is co-written with Chaim Rand from Mobileye. Certain machine learning (ML) workloads, such as training computer vision models or reinforcement learning, often involve combining the GPU- or accelerator-intensive task of neural network model training with the CPU-intensive task of data preprocessing, like image augmentation. When both types of tasks run on the same …