The LF Deep Learning Foundation on Wednesday announced the availability of the first software from the Acumos AI Project. Dubbed “Athena,” it supports open source innovation in artificial intelligence, machine learning and deep learning.
This is the first software release from the Acumos AI Project since its launch earlier this year. The goal is to make critical new technologies available to developers and data scientists everywhere.
Acumos is part of a Linux Foundation umbrella organization, the LF Deep Learning Foundation, that supports and sustains open source innovation in artificial intelligence, machine learning and deep learning. Acumos is based in Shanghai.
Acumos AI is a platform and open source framework that makes it easy to build, share and deploy AI apps. Acumos standardizes the infrastructure stack and components required to run an out-of-the-box general AI environment, freeing data scientists and model trainers to focus on their core competencies, and accelerating innovation.
“The Acumos Athena release represents a significant step forward in making AI models more accessible for builders of AI applications and models, along with users and trainers of those models and applications,” said Scott Nicholas, senior director of strategic planning at The Linux Foundation. “This furthers the goal of LF Deep Learning and the Acumos project of accelerating overall AI innovation.”
The challenge with AI is that there are very few apps to use it, noted Jay Srivatsa, CEO of Future Wealth.
“Acumos was launched to create an AI marketplace, and the release of Athena is a first step in that direction,” he told KERNEL.
The Acumos AI Platform
Acumos packages toolkits such as TensorFlow and SciKit Learn, along with models that have a common API that allows them to connect seamlessly. The AI platform allows for easy onboarding and training of models and tools.
The platform supports a variety of popular software languages, including Java, Python, and R. The R language is a free software environment for statistical computing and graphics.
The Acumos AI Platform leverages modern microservices and containers to package and export production-ready AI applications as Docker files. It includes a federated AI Model Marketplace, which is a catalog of community-distributed AI models that can be shared securely.
LF Deep Learning members contribute to the evolution of the platform to ease the onboarding and the deployment of AI models, according to LF Deep Learning Outreach Committee Chair Jamil Chawki. The Acumos AI Marketplace is open and accessible to anyone who wants to download or contribute models and applications.
“Acumos Athena is a significant release because it enables the interoperability of AI, DL and ML models and prevents the lock-in that usually occurs whenever projects are built using disparate configurations, systems and deployment techniques,” explained Rishi Bhargava, cofounder of Demisto.
It will ease restrictions on AI, DL and ML developers by removing silos and allowing them to build standardized models, chain each other’s models together, and refine them through an out-of-the-box general AI environment, he told KERNEL.
“The efficiency of learning models is hugely contingent on the quality and uniqueness of data, the depth and repeatability of feature engineering, and selecting the best model for the task at hand,” Bhargava said. “Athena will free developers of extraneous burdens so they can focus on these core tasks, learn from each other, and eventually deliver better models to businesses and customers.”
Athena Release Highlights
Athena’s design is packed with features to make the software quick and easy to deploy, and to make it easy to share Acumos AI applications.
Athena can be deployed with one-click using Docker or Kubernetes. The software also can deploy models into a public or private cloud infrastructure, or into a Kubernetes environment on users’ own hardware, including servers and virtual machines.
It utilizes a design studio graphical interface that enables chaining together multiple models, data translation tools, filters and output adapters into a full end-to-end solution. Also at play is a security token to allow simple onboarding of models from an external toolkit directly to an Acumos AI repository.
Models easily can be repurposed for different environments and hardware. This is done by decoupling microservices generation from the model onboarding process.
An advanced user portal allows personalization of marketplace view by theme and data on model authorship. This portal also allows users to share models privately or publicly.
“The LF Deep Learning Foundation is focused on building an ecosystem of AI, deep learning and machine learning projects, and today’s announcement represents a significant milestone toward achieving this vision,” said LF Deep Learning Technical Advisory Council Chair Ofer Hermoni of Amdocs.
The Acumos release is significant for the advancement of AI, DL and ML innovation, according to Edgar Radjabli, managing partner of Apis Capital Management.
The AI industry is very fragmented, with virtually no standardization.
“Companies building technology are usually required to write most from scratch or pay for expensive licensed cloud AI solutions,” Radjabli told KERNEL. “Acumos can help bring a base (protocol) layer standard to the industry, in the same way that HTTP did for the Internet and Linux itself did for application development.”
LF Deep Learning members are inspired and energized by the progress of the Acumos AI Project, noted Mazin Gilbert, vice president of advanced technology and systems at AT&T and the governing board chair of LF Deep Learning.
“Athena is the next step in harmonizing the AI community, furthering adoption and accelerating innovation,” he said.
Open Source More Suitable
Given the challenges of growing new technologies, open source models are better suited to the development process than those of commercial software firms. Open source base layer software is ideal. It allows greater adoption and interoperability between diverse projects from established players and startups, said Radjabli.
“I believe that Acumos will be used both by other open source projects building second-layer applications, as well as commercial applications,” he said.
Today, the same situation exists in other software development. Open source base layer protocols are used across the industry, both by other open source/nonprofit projects and commercial operations, he explained.
“Athena clearly is geared to an open source environment, given that it already has about 70 or more contributors,” said Future Wealth’s Srivats.
Benefits for Business and Consumers
The benefits to be gained from AI, DL and ML are very significant. Companies across the industry have been making progress in the development of unique applications for AI/DL/MO. More growth in this space will result from Acumos, according to Radjabli.
One example involves a company that uses neural networks for predictive healthcare analytics. This system allows it to diagnose breast cancer with zero percent false negatives simply from patient data correlation analysis. This does not involve any invasive testing or imaging, according to Radjabli.
“The correlation is comprised of over 40 variables, which means it would have never been found through traditional medical research data analysis and was only made possible through the use of convolutional and recurrent neural networks working in combination,” he said.
AI, DL and ML are all geared toward businesses understanding and predicting consumer behavior, added Srivatsa.
“Both will benefit,” he said.
What’s Next for Acumos AI
The developer community for Acumos AI already is working on the next release. The company expects it to be available in mid-2019.
The next release will introduce convenient model training, as well as data extraction pipelines to make models more flexible.
Additionally, the next release will include updates to assist closed-source model developers, such as secure and reliable licensing components to provide execution control and performance feedback across the community.