OpenAI’s seven unsolved problems, NVIDIA integrated GPU to IBM cloud, open-sourcing Psychlab, InterSystems IRIS data platform generally available, and more in today’s top stories around machine learning, deep learning,and data science news.
OpenAI states that it is releasing new batch of seven unsolved problems, which came up during the course of their research. These questions will pave a meaningful way for new people to enter the field, as well as for practitioners to hone their skills. It is also a great way for people to get a job at OpenAI.
Let us now have a look at the seven unsolved problems:
- Implement and solve a multiplayer clone of the classic Snake game as a Gym environment. (One can refer slither.io for inspiration)
- Explore the effect of parameter averaging schemes on sample complexity and amount of communication in RL algorithms.
- Transfer Learning Between Different Games via Generative Models
- Use linear attention for the Transformer model (which uses soft attention with softmax)in order to use the resulting model for RL.
- Use a learned VAE of data, to perform “learned data augmentation”.
- Experimentally investigate (and qualitatively explain) the effect of different regularization methods on an RL algorithm of choice.
Excited? Have a detailed read on OpenAI blog.
DeepMind open-sourced Psychlab, a platform built on top of DeepMind Lab, for others to use.
Psychlab allows direct application of methods from fields like cognitive psychology to study behaviours of artificial agents in a controlled environment.
Alongwith open-sourcing Psychlab, the DeepMind team have also built a series of classic experimental tasks to run on the virtual computer monitor, which has a flexible and easy-to-learn API, enabling others to build their own tasks.
Read more about Psychlab and the added tasks on DeepMind’s blog.
IBM announces availability of NVIDIA Tesla V100 GPU on its Cloud, which aims to accelerate enterprise efforts in mission-critical artificial intelligence (AI), deep learning, and HPC workloads.
The V100 GPU is NVIDIA’s fastest and most advanced GPU accelerator on the market, says John Considine, general manager of cloud infrastructure services for IBM Watson and Cloud Platform. Users can now integrate individual IBM Cloud bare metal servers with up to two NVIDIA Tesla V100 PCle GPU accelerators. This combination of IBM’s high-speed network connectivity and bare metal servers with the V100 GPUs will provide a major boost to compute-intensive workloads.
In a blog post, John Considine, general manager of cloud infrastructure services for IBM Watson and Cloud Platform said,”With the Tesla P100 GPU accelerator, you can leverage up to 65 percent more deep learning capabilities and 50 times the performance than its predecessor”.
For details, visit IBM’s blog.
DeepMind has come up with a new paper in the Journal of Artificial Intelligence Research(JAIR). The new paper showcases how Deep Neural Networks can be extended for generalizing visually and symbolically.
This paper proposes a Differentiable Inductive Logic framework, which can solve tasks which traditional Inductive Logic Programming (ILP) systems are suited for. It can also show robustness to noise and error in the training data which ILP cannot cope with. Further, as it is trained by backpropagation, it can be hybridised by connecting it with neural networks over ambiguous data. Thus, this provides data efficiency and generalisation beyond what neural networks on their own can achieve.
InterSystems announced the general availability of InterSystems IRIS Data Platform, the first data platform to deliver multi-workload and multi-model data management, native interoperability, and an open analytics platform in a single product.
The IRIS is a complete unified data platform that makes it faster and easier to build real-time data-rich applications. It allows organizations to combine event and transactional data with large sets of historical and other data for capturing untapped business opportunities and also to improve operational efficiencies.
InterSystems IRIS Data Platform aims:
- To delivers concurrent transactional and analytic processing, and multiple data representations (including relational and non-relational models which are always synchronized) in a single database;
- To provide a complete set of interoperability capabilities for integrating disparate data and applications and create seamless real time business processes
- To include business intelligence and natural language processing capabilities, and an open analytics platform that allows best-of-breed, third-party analytics to be easily incorporated through dedicated connectors and industry standards
- To support flexible deployment options for public and private cloud, on premises, and hybrid environments.
Have a detailed read at the official press release.