Google’s Ethical Dilemma, Cloud leaders & Maths

This Week's top AI stories

AIMResearch is a dedicated market research arm of AIM, specializing in the AI & Data Analytics industry in India. Besides syndicated research on how AI & data science industry is shaping in India, we also execute various customised research for organizations including surveys and benchmarking.

Recently, we released our annual  Analytics India Attrition Study. The insights were interesting.  The report found the voluntary attrition rate of analytics professionals in India dropped from 30.7% in 2019 to 16.0% in 2020.

Though India’s attrition rate of 16% in 2020 is a significant improvement year on year, the rate is slightly higher than the world average of 14.1%.

There are a lot of deeper research reports to come. Reach out to us in case you are looking for any specific data point or insight into the industry.


Google’s Ethical Dilemma

Google removed “don’t be evil” from its code of conduct in the first half of 2018. In retrospect, the move anticipated the sequence of events culminating in the firing of Google’s leading ethical AI researchers, Timnit Gebru and Margaret Mitchell.

It all started with Gebru & Mitchell’s paper on the dangers of large language models like GPT-3 and BERT.

Gebru popped the critical question, “how big is too big”, much to Google’s disgruntlement. The conversations around ethics are gaining traction as AI and ML usher in the fourth industrial revolution. 


What Separates Cloud Leaders From Others

A handful of cloud service providers, namely AWS, Azure and Google Cloud, rule the market. Many cloud players have burst into the scene along the way but are yet to make a mark. What separates the top three from the rest is their constant revival of niche services.

The big three forayed into ML-based services quite early. Now they have custom options to create chatbots, deploy AutoML, recommendation engines, and many other applications that power most companies.

So what makes them unique?


Do Large Machine Learning Models Struggle At Maths?

A new study by the researchers at the University of California, Berkeley, have now introduced the MATH dataset. The team said the dataset provides a detailed assessment of a model’s mathematical ability across difficulties and subjects.

When the MATH dataset was tested for large language models, including GPT-3, the accuracies were found to be abysmally low, ranging from 2.9 percent to 6.9 percent.  


Featured Video | Analytics India Guru - BERT क्या है, आइए जानें हिंदी में। What is BERT (in HINDI)

In This Video, we will introduce you to BERT in Machine Learning. Google released Bidirectional Encoder Representations from Transformers or BERT in 2018. It offered a new ground to embattle the intricacies involved in understanding the language models.

Pre-training a binarised prediction model helps understanding common NLP tasks like Question Answering or Natural language Inference. BERT allowed improvement of NLP applications through Search, smart compose, and many more.


Hugging Face And Its Tryst With Success

Recently, Hugging Face raised $40 million in Series B funding. ‘Its entire purpose is to be fun’, a media report said in 2017 after Hugging Face launched its AI-powered personalised chatbot. Named after the popular emoji, Hugging Face was founded by Clément Delangue and Julien Chaumond in 2016.

What started as a chatbot company, has transformed into a "GitHub for machine learning". Hugging Face has become one of the fastest-growing open-source projects, especially in the area of machine learning. Across four rounds of funding, Hugging Face has raised over $60 million till now. 


Hands-On Guides for ML Developers

PyVista: A Python Package For 3D Plotting And Mesh Analysis

DeLighT: Deep and Light-weight Transformer

STRIPE: Shape and Time Diversity in Probabilistic Forecast

Kornia: An OpenCV-inspired PyTorch Framework

Real-time Voice Cloning: Neural Network System for Text-to-Speech Synthesis



Commoditization Is The Biggest Problem In Data Science Education: Prof. Raghunathan Rengasamy, IIT Madras

You Got To Have A Strong Backbone To Be In This Career: Olivia Gambelin, Founder, Ethical Intelligence

Government’s Proposal For Semiconductor FAB Facilities Will Boost India’s Manufacturing Ecosystem: Sanjay Gupta, NXP

Interview with Dr Inder Gopal, CEO at India Urban Data Exchange

Talking Pains And Promises Of GPT-3 With Sahar Mor

How Decimal Technologies Uses AI To Empower Digital Lending Platforms

Tech Behind Test Preparation Platform

How This Bangalore-Based Startup Provide Automation For Media Organisations


Is AI Adoption Going Way Too Fast?

The COVID-19 pandemic has accelerated the pace of AI adoption, but many industry insiders find the speed of adoption a bit overwhelming, according to a KPMG survey.

According to KPMG, industries are experiencing a COVID-19 ’whiplash’ with AI adoption skyrocketing due to the pandemic. Meanwhile, experts have reposed faith in AI’s ability to solve significant business challenges.


Why Do Companies Prefer Pre-Trained Models?

Building a model from scratch requires a great deal of time and effort. When innovations are happening at a break-neck speed, concentrating all the efforts on building models could prove counterproductive.

Enter pre-trained models. Put simply, it is a model created by a third party to solve a business problem. While it may not be 100 percent accurate, it saves time and effort in building a model from the ground up.



Here's what all happened last week.

Solar Analytics Startup Prescinto Raises $3.5 Million In Seed Round Introduces Open-Source Automatic ML Package For Wave Apps

Microsoft Launches Azure Migration Program And FastTrack For Azure In India

CBSE Partners With Intel To Launch AI Learning Platform

Coursera Launches World’s First Online Data Labelling Course For Free