AI/ML Roundup November 2018
Amazon Web Services announces enhancements to Amazon SageMaker, plus a slew of new AI/ML services. Enhancements to SageMaker include:
- Amazon SageMaker Ground Truth, a service that helps customers build training datasets for machine learning.
- Support for reinforcement learning
- Amazon SageMaker Neo, a capability that enables developers to train machine learning models once and run them on many different hardware architectures.
Other new services available immediately include:
- Amazon EC2 P3dn Instances, designed to support distributed machine learning, with up to 4x the network bandwidth of P3.16xlarge instances.
- Amazon Elastic Inference allows customers to attach low-cost GPU-powered acceleration to Amazon EC2 and Amazon SageMaker instances to for low-cost deep learning inference.
New services in Preview include:
- Amazon Forecast, a managed service for time series forecasting.
- Amazon Personalize, a service that enables developers to create individualized recommendations for customers.
- Amazon Textract, a service that automatically extracts text and data from scanned documents.
Janakiram MSV, one of the sharper analysts in this field, opines that AWS is now an unstoppable juggernaut.
A busy month for H2O.ai. The company announces H2O Version 3.22. Enhancements include the addition of XGBoost GBMs to the AutoML feature and a new algorithm, Isolation Forest for unsupervised learning.
H2O.ai changes the Driverless AI roadmap again and postpones delivery of multi-node training to Version 2.0 next year. Driverless AI currently runs on single machines only. Multi-node training is a key enhancement for strategic partners NVIDIA and IBM, who want to sell clusters of GPU-accelerated machines.
H2O.ai announces plans for its big event, H2O AI World San Francisco, for February 2019. (The company originally planned this event for November 2018). The company continues to expand its partner ecosystem, announcing partnerships with BlueData and CirraScale. H2O.ai also launches an academic program. The company will provide non-commercial Driverless AI licenses and other benefits for qualified university classes. More information here.
- H2O Driverless AI
- H2O-3 rebranded as H2O Artificial Intelligence
- H2O AutoML
- H2O Generalized Linear Models
- H2O Gradient Boosted Machines
- H2O Deep Learning
The Driverless AI service is Bring Your Own License (BYOL). The other five services are available free of software costs; users pay AWS infrastructure costs. AutoML, GLM, GBM, and Deep Learning are also available in AWS SageMaker (at pricing that depends on the EC2 instance the user selects.) Driverless AI and H2O AI are available as Amazon Machine Images (AMIs).
SAS tries to reposition itself as a player in AI. SAS reorganizes its Banking and Financial Analytics messaging, replacing the Advanced Analytics category with AI and Machine Learning. SAS also shuffles its featured banking products and adds links to SAS AI Solutions, which is neither a product nor a service but a marketing umbrella that points to existing ML and NLP software.
Pro tip: if you want to be a player in AI, develop some AI products and services.
Based on personal contacts, there seem to be quite a few SAS sellers heading for the exits. Sellers exiting at this point in the sales year is not a bullish sign for 2018 revenue growth.
Dataiku releases Data Science Studio Version 5.0.3, with minor enhancements and bug fixes. Separately, Dataiku publishes a white paper from Ovum and sponsors the O’Reilly ebook on machine learning interpretability.
The company holds its “sold out” EGG New York in its own offices. There is absolutely no truth to the rumor that attendees rolled their chairs to the conference room.
Dataiku blogs about the importance of automated machine learning and releases a white paper on the same topic. The paper includes a table of features and shows which ones Dataiku automates. Presumably, Dataiku thinks it has a competitive offering, or it wouldn’t bang this drum.
Pro tip: if you automate some things but not others, you’re selling half a bridge.
Dataiku also adds two Marketing Directors and sheds an Alliance Director.
Alteryx announces Release 2018.4, with minor enhancements and bug fixes.
Google Cloud launches AI Hub, a service designed to help customers design, launch and keep track of their machine learning projects.
Gartner publishes its PeerInsights Customers’ Choice awards for Data Science and Machine Learning Platforms. PeerInsight aggregates product reviews from verified customers. For 2018, Gartner awards Customers’ Choice distinctions to Alteryx, IBM, and RapidMiner. KNIME, Dataiku, Databricks, and H2O.ai also perform well. The PeerInsight rating correlates (to some degree) with scores on the customer experience survey that Gartner uses for its Magic Quadrant. That bodes well for these vendors in the 2019 MQ.