Keeping up with Machine Learning — (February 2020 — Week 2)
Welcome to your weekly dose of Artificial Intelligence, Machine Learning, and Data Science. Sign up for our weekly newsletter to stay up to date on the latest news.
Gain Some Insight
Artificial Intelligence is Replacing Musicians
Google isn’t only building tools for lyrics and vocals; the tech behemoth also has a variety of AI songwriting tools under the umbrella of open-source research project Google Magenta, including NSynth.
Another Google Magenta project is Piano Genie, an intelligent controller that maps 8-buttons of input to a full 88-key piano in real-time.
Is AI going to eventually replace human musicians? This article shares an interesting take:
Artificial Intelligence is changing Business
While AI is still in its early stages, there are multiple areas of business that can benefit from AI right now.
As described in the next article, these areas include:
Sales: AI will virtually engage with existing customers, and explore customer graphs to find new customers.
Marketing: AI will identify lucrative product areas and capture trending concepts.
IT: AI will predict and prevent maintenance requests. Constantly test deployments with intelligent tooling.
Human Resources: AI will Explore social/business networks such as LinkedIn and compile candidate lists with graph-processing AI.
See this article for further detail on business areas that will tap into AI:
How will your career be impacted by AI?
AIs are currently slated to handle basic tasks and administrative work, leaving humans to tackle more complex tasks that require problem-solving, critical thinking, creativity, emotional intelligence, judgment and decision-making, and cognitive flexibility. While AI handles basic tasks and administrative work, people will demonstrate complex thinking to propel organizations forward.
This article looks at how AI is affecting — and will continue to affect — people’s career opportunities:
Some New Releases
In a significant boost to 3D deep learning research, Facebook AI has released PyTorch3D, a highly modular and optimized library with unique capabilities to make 3D deep learning easier with PyTorch.
PyTorch3d provides efficient, reusable components for 3D Computer Vision research with PyTorch.
Researchers and engineers can similarly leverage PyTorch3D for a wide variety of 3D deep learning research, whether it be, 3D reconstruction, bundle adjustment, or even 3D reasoning to improve 2D recognition tasks.
For smooth integration of deep learning and 3D data, PyTorch operators:
- Are implemented using PyTorch tensors
- Can handle mini-batches of heterogeneous data
- Can be differentiated
- Can utilize GPUs for acceleration.
3D deep learning researchers can easily import the loss functions using the modular differentiable API.
Microsoft DeepSpeed
Microsoft Research today announced DeepSpeed, a new deep learning optimization library that can train massive 100-billion-parameter models. In AI, you need to have larger natural language models for better accuracy. But training larger natural language models is time-consuming and the costs associated with it are very high. Microsoft claims that the new DeepSpeed deep learning library improves speed, cost, scale, and usability.
Improve your skills
This article contains a list of IBM Data Science and Artificial Intelligence programs with 30 days of free access. This list is updated whenever there are new additions.
Be careful and only enroll in the program that you really want to take, as learners can claim the offer one-time only.
This next tutorial displays transfer learning for diverse NLP tasks with Universal Sentence Embeddings. It explores models such as ELMo, Universal Sentence Encoder, and ULMFiT. Ultimately, it showcases that pre-trained models can be used to achieve state-of-the-art results on NLP tasks
Video Of The Day
Data analysis using R’s data.table is very easy to use, high performant and convenient. According to the comprehensive Comprehensive R Archive Network, data.table adheres to
- concise and consistent syntax irrespective of the set of operations you would like to perform to achieve your end goal.
- performing analysis fluidly without the cognitive burden of having to map each operation to a particular function from a potentially huge set of functions available before performing the analysis.
- automatically optimizing operations internally, and very effectively, by knowing precisely the data required for each operation, leading to very fast and memory-efficient code.
So, what’s the Python version for this. Datatable! which also prioritizes big data support and high performance. How does it compare to pandas in terms of speed? Datatable outperforms pandas when reading large datasets. See this video for more info!
Book Of The Day
Pattern Recognition and Machine Learning
If you’re looking to formally start your machine learning studies, it typically boils down to two books. “Machine Learning — a probabilistic perspective” by Kevin Murphy or “Pattern recognition and Machine learning” by Christopher Bishop. Personally, I think they are both great books, but I highly recommend reading Bishop’s book first as each chapter is designed to build off previous chapters in a very coherent and intelligible manner. It’s pretty self-contained and rarely expects prior knowledge in machine learning or probability theory (unlike Murphy’s book, where the ordering is harder to follow but compensates in coverage).
Enjoy!
https://www.amazon.ca/Pattern-Recognition-Machine-Learning-Christopher/dp/0387310738
https://www.amazon.ca/Pattern-Recognition-Machine-Learning-Christopher/dp/0387310738
That’s all for this week’s AI updates. Make sure to subscribe below (if you haven’t already) to stay up to date on the latest developments!
