Machine Learning News - January 2018
Facebook’s Expanding Machine Learning Infrastructure. Thenextplatform.com, January 8, 2018.
Facebook has made notable progress in fine tuning it data centers. Facebook is focusing on infrastructure right from its distinct split network backbone, neural network based viz system , to large scale upgrades being made to its server farms.One of the major developments in Facebook infrastructure is its own server designs that use latest GPU packed open compute based approach and serve more than 2 billion accounts . The company’s infrastructure Big Basin system that was unveiled last year has been a great success as it can train machine learning models that are 30% larger because of the availability of larger arithmetic throughput and increase in memory from 12GB to 16GB. The shift to Big Basin system has helped Facebook gain 300% enhancement in throughput over Big Sur system. However, the machine learning infrastructure at Facebook still consists of CPU and GPUs only as they are sticking to its Nvidia and Intel guns for the time being.
(Source : https://www.nextplatform.com/2018/01/08/facebooks-expanding-machine-learning-infrastructure/ )
If you would like more info about DeZyre's Certified Machine Learning Course, click on the request info. button on top of this page.
Verizon acquires machine learning-based threat detection startup Niddel. Siliconangle.com, January 8, 2018.
Verizon acquired a startup Niddel Corp. that offers a threat-hunting platform that is based on machine learning technology. This acquisition will help Verizon enhance the security of its enterprise network products. Niddel has subscription based service known as Niddel Magnet that makes use of machine learning and identifies threats in customer networks.It uses machine learning to improve information accuracy considerable and reduce false positives to significantly improve detection and response capabilities of Verizon’s enterprise network products.Niddel’s capabilities will be added to Verizon's network and gateway security , incident response and security monitoring tools .
(Source : https://siliconangle.com/blog/2018/01/07/verizon-acquires-machine-learning-based-threat-detection-startup-niddel/ )
Why Machine Learning Needs GPUs ? Motherboard.vice.com, January 9, 2018.
There is no magic to machine learning, it is all math. The basic idea behind machine learning is very simple and it is all about optimization. Given a long equation with several variables, you need to come up with reliable way of tweaking these variables in such a way that the equation gives out reliable predictions.It might appear to be a conceptually simple question to ask but the computing is labor-intensive. The important thing to understand with machine learning is that it involves crunching big matrices of numbers and this is what happens with graphics processing where the matrices instead refer to pixels.Machine learning algorithms tend to work faster just with the addition of more processor cores within a GPU.This is the reason GPU’s are extremely important to machine learning, and increasingly, vice versa.
(Source : https://motherboard.vice.com/en_us/article/kznnnn/why-machine-learning-needs-gpus )
Workday buys SkipFlag to bolster machine learning capabilities.Zdnet.com, January 16, 2018.
Workday announced the acquisition of SkipFlag , makers of the so called AI knowledge base that builds itself from company’s internal communications. SkipFlag’s AI system extracts huge amounts of enterprise data from employee profiles, support tickets, messages and corporate documents and then builds a database from which answers to all employee questions are pulled automatically. This deal marks another step in Workday’s effort to invest in areas like advanced search, deep learning, machine learning and NLP. SkipFlag’s team is all set to join Workday and its technology will be incorporated into Workday’s core platform.
(Source : http://www.zdnet.com/article/workday-buys-skipflag-to-bolster-machine-learning-capabilities/ )
India moves to address AI talent supply gap, gets a leg-up from Google, Microsoft, Intel.FactorDaily.com, January 18, 2018.
Talent is the biggest entry barrier in the field of AI and Machine Learning not just in India but across the globe. According to a New York Times report ,there are only 10,000 people who can actually tackle serious AI research and command six figure dollar salaries.India is a country that supports more than half of the tech outsourcing requirements and the limited supply of AI and Machine Learning is alarming. Employers are finding it difficult to hire skilled talent.The Indian government, universities and companies like Microsoft and Google are launching several learning initiatives to train people in AI and Machine Learning. Google recently partnered with Pluralsight and nanodegree training provider Udacity to give out 1.3 lakh scholarships in emerging technologies including Machine Learning to Indian developers. India has the largest number of developers after US and is expected to overtake US by 2021. The Indian government is also looking at wide range of certificate oriented professional program that will professionals take skills to the next level.
(Source : https://factordaily.com/india-ai-talent-gap-google-microsoft/ )
Google lets developers add machine learning to their apps.ItPro.co.uk, January 18, 2018.
Today there are very few organizations across the world that have access to the required talent and budgets to fully leverage the advancements in AI and Machine Learning.There are very limited number of people who can develop advanced machine learning models. Companies having limited access to AI and ML engineers, still need to manage compute and time intensive complicated process of building custom machine learning models.
Google has launched a simpler product known as Cloud AutoML that will businesses integrate machine learning capabilities into applications in less than a day.Business without the expertise to build machine learning models can use Cloud AutoML to transfer learning from Google.The cloud AutoML tool will be released in various stages.
(Source : http://www.itpro.co.uk/machine-learning/30307/google-lets-developers-add-machine-learning-to-their-apps )