Larger Language Models, Transformers & a Comparison

Top AI Stories of the Week

Have you checked our Discord Community? We recently started building an AIM community on Discord. Our Telegram community of 16k+ users stay as is but we realize that for our members to collaborate easily, the choice of platform matters a lot.

I personally am loving Discord, so do join here


Rush To Create Larger Language Models

There's almost a gold rush to come out with the “biggest” language model. Turing NLG from Microsoft, GPT-3 from OpenAI, Jurassic-1 Jumbo by AI21 Labs and Wu Dao 2.0 from Beijing Academy of Artificial Intelligence (BAAI) — all these language models have been introduced within one and a half years. It's almost a showcase of "mine is bigger than yours".

But Is it worth putting resources on building just larger language models than previous ones?


What’s Wrong With Limiting Transformers To Just Language Models

Three years back, Google introduced Transformer, a novel neural network architecture. Since its introduction, some of the most popular and groundbreaking language models have been based on transformers. It includes Bidirectional Encoder Representations from Transformers (BERT), Generative Pre-Training Models (GPT-2,-3), and XLNet.

Though invented for processing language models mostly, the scope of transformers has grown way beyond that. Transformers can help workers spend more time doing meaningful work and boost productivity.


Big Data Helps Bust Tax Fraud In Kanpur

This month, the Uttar Pradesh IT department identified over 250 street food vendors of Kanpur to be millionaires. These millionaires include vegetable sellers, small pharmacy shop owners, grocers, rag pickers and sanitation workers- all who have been evading tax for years and manage to hoard Rs 37.5 million.

The IT department has been using surveillance software to identify the tax evaders. The software tracked down several dealers who owned at least three cars, KVP worth more than 30 crores and 650 bighas of agricultural land.


Data Science Hiring Process At Mobile Premier League (MPL)

In the last year, MPL’s data science team has grown from scratch to 16 members. Today, the company operates a central data science team that works across different business verticals. The team includes a blend of data scientists, ML engineers and data analysts, following a flat structure consisting of individual contributors with varying experience levels, led by the head of data science.

The data science team falls within the broader data organisation, which comprises data analytics, data products and data engineering.


How Zomato Uses Machine Learning

Zomato uses ML to predict a lot of challenges in real-time, including last-mile delivery of food orders, allocation of delivery partners, estimation of time for restaurants to prepare food, grooming of delivery partners, assessing fake reviews, etc.  

Zomato uses a simple equation of three variables — input, brain, and output, via ‘model training’ and ‘model prediction’ techniques. Meaning, if the output and input are known, the process of figuring out the brain is called model training. If the input and brain are known, the method of computing the output is called model prediction. 


Hands-on Guides for ML Developers

Complete Tutorial On Image Transformations With OpenCV

Guide To Image Reconstruction Using Principal Component Analysis

My Small Project on Animating Images & Videos Using GANsNRoses

Hands-On Guide To Monte Carlo Simulation With Its Significance

Hands-On Guide To OpenBox Toolkit for Black-Box Optimization



How Gupshup Uses AI

AI Has Helped Us In Improving The Performance Of Our Network: Prashant Gupta, Verizon India

How Banking Tech Unicorn Zeta Uses AI

Interview With Shivam Gautam, Tech Lead @ Aurora

How IIT Guwahati Leads Cutting-Edge Research In AI & Data Science

Security Leaders In India Need To Create A Robust Cybersecurity Ecosystem: Deepak Bhawnani, Alea Consulting


AI For Chip Design

Just two months after Google unveiled its new deep reinforcement learning technique for designing the next generation of Tensor Processing Units, Samsung has revealed that it would be taking a similar route. 

Apart from Google and Samsung (Synopsys), other companies like NVIDIA and IBM are also dabbling in AI-designed chips. If successfully deployed, this could mean a new area of AI applications and possibly a breakthrough in designing chips.


Jurassic-1 vs GPT-3 vs Everyone Else

AI21 Labs, an Israeli AI company, specialising in NLP, has released a language model, Jurassic-1 Jumbo. The tool is released with the idea to challenge OpenAI’s dominance in the “natural language processing-as-a-service” field. 

ML researchers and developers posit that larger models trained on more parameters produce better outcomes. In this article, we compare Jurassic-1 to other large language models that are currently leading the market.



Here's what all happened last week.

IIM Ahmedabad To Establish A Centre For Data Science And Artificial Intelligence

Pankaj Rai Joins Aditya Birla Group As Its Group Chief Data & Analytics Officer

Agri-Tech Startup AgNext raises $21 million in Series A

IIT Hyderabad Establishes AI Research Lab In Collaboration with Honeywell

IIT Roorkee Launches PG Certification In Cloud Computing And DevOps