Stealing the show from most big tech companies, Alphabet Inc, for the last quarter of 2021, reported $257.6 billion in revenue, calculating the growth of 41% YoY. Just in the last quarter, the company reported $75.3 billion in revenue with a 32% YoY growth. Withstanding the pandemic and a dry year in 2020 (Alphabet’s sales grew by only 13% in all of 2020, the slowest rate in over a decade), the company upped their tech and business game in the previous year with innovations, pre-pandemic rate of hiring and construction and financial plans.

While on the financial side, Alphabet’s strategy of breaking up their huge $2,640 share price into fractional-share investment boosted investments in the company, on the technical side, Alphabet’s subsidiaries, mainly Google, developed some of the best innovations of 2021. Google, which makes chunks of its revenue from advertisement, earned over $61.2 billion, and Google Cloud services earned $5.54 billion. 

Sundar Pichai also attributed this growth to the company’s investment in AI. “Our deep investment in AI technologies continues to drive extraordinary and helpful experiences for people and businesses across our most important products. Q4 saw ongoing strong growth in our advertising business, which helped millions of businesses thrive and find new customers, a quarterly sales record for our Pixel phones despite supply constraints, and our Cloud business continuing to grow strongly,” he said, in Alphabet’s financial report. Analytics India Magazine has analysed the key innovations by Alphabet that may have led to this success.

Google Cloud

Google’s winning strategy was meeting the customers’ demands and needs of digital transformation and work from home post-pandemic. Cloud has disrupted the technology landscape, and Google Cloud has been on top of its game with updates and features that meet the users’ needs and satisfaction. As a result, Google Cloud’s quarterly revenue increased by 45% to $5.5 billion. In 2021, Cloud expanded to supporting clients exploring blockchain, introducing the Ethereum Blockchain Explorer with Cloud Spanner. 

At the Google Cloud Next 2021 event, Pichai announced Google Cloud’s expansion plan where the company has added cities like Warsaw, Delhi, Melbourne and Toronto to Cloud’s network. In addition, 2022 will add Chile, Israel, Germany, Saudi Arabia and North American regions. “It’s all connected by 19 subsea cables. Our Grace Hopper transatlantic cable will be the first to incorporate novel optical fibre switching,” he said. Further, addressing the growing security concerns with digital transformation, the company announced comprehensive security to support organisations in protecting their data. Additionally, Google Cloud introduced Smart Canvas to help content creators remove friction from workflows and enable teams to stay connected outside of meetings to foster well being and reduce video call fatigue. 

Google has also teamed up with global telecom leaders like Erricson, Nokia and T Systems to deliver 5G edge and hosted computing solutions. In addition, Google Cloud also announced Vertex AI, a managed machine learning (ML) platform for deploying and maintaining AI models. Vertex AI brings AutoML and AI platforms together into a unified API and can build, deploy and scale ML models faster. Deployed by Google Research, Vertex AI required 80 per cent fewer lines of code for custom modelling. In addition, it integrated with open-source frameworks, including TensorFlow, PyTorch and Scikit-learn. 

Google AI innovations

Talking about Google’s innovations, Pichai said, “Over the past year, we have had more than 1500 product and feature releases. That is about four a day.” In addition, Google introduced several successful AI innovations in 2021, surpassing previous benchmarks and models. 

Google’s GLaM that dominated GPT-3

Google introduced their Generalist Language Model (GLaM), a trillion weight model that uses sparsity. GLaM has improved learning efficiency across 29 public NLP benchmarks in seven categories: language completion, open domain question answering, and inference tasks. In addition, GLaM’s performance compares favourably to GPT-3 (175B). 

Google had the highest paper acceptance

The Neural Information Processing Systems conference is a key AI and ML research initiative. In their 35th edition in 2021 (NeurIPS 2021), Google had the highest number of accepted papers, totalling 177 papers. The topics covered in these papers include online learning for latent Dirichlet allocation, the life of a dataset in ML and the tandem learning experiential paradigm. 

Wikipedia-Based Image Text (WIT) Dataset

Google’s Wikipedia-Based Image Text (WIT) Dataset is a large multimodal dataset created on multiple different text selections associated with an image from Wikipedia articles and Wikimedia image links. Google says this was followed by rigorous filtering to retain high-quality image-text sets. 

Indian Language Transliterations in Google Maps

Mitigating the problem of maps not being available in the native scripts of the languages of India, Google came out with an ensemble of learned models to transliterate names of Latin script POIs into ten languages prominent in India. These include Hindi, Bangla, Marathi, Telugu, Tamil, Gujarati, Kannada, Malayalam, Punjabi, and Odia. 

FLAN Model

Google’s Fine-tuned Language Net (FLAN) model explores instruction fine-tuning. This NLP model is fine-tuned on a large set of varied instructions that use a simple and intuitive description of the task. FLAN uses templates to transform existing datasets into an instructional format instead of creating a dataset of instructions from scratch to fine-tune the model.

MURAL

MURAL, Multimodal, Multitask Representations Across Languages model for image–text matching, uses multitask learning applied to image–text pairs combined with translation pairs covering over 100 languages.

Mitigating the silicon chip shortage

While auto and electronic manufacturers are suffering under global semiconductor chip shortage, 2021 was a great year for the Google Pixel smartphones, seeing the highest-ever quarterly shipments in Q4 2021. Sundar Pichai boasted of these sales despite the supply chain issues, and rightfully so.

While Google majorly drove Alphabet’s revenues, the company’s subsidiaries do not end there. With companies like DeepMind, GV, Waymo and more that also did good in 2021, AIM explores the additional factors that partook in increasing Alphabet’s shares in the year.

DeepMind’s innovations

DeepMind had an interesting year with several innovations and releases that were some of the top AI innovations of 2021. While the company has usually been under debt, the world’s premier artificial intelligence labs turned a profit for the first time in 2021.

Player of Games (PoG)

DeepMind’s Player of Games (PoG) performs well at perfect and imperfect information games such as Chess and Go, Poker, and Scotland Yard. It works by using a single algorithm with minimal domain-specific knowledge. PoG uses growing-tree counterfactual regret minimisation (GT-CFR) to build subgames non-uniformly and expand the tree toward the most relevant future states while iteratively refining values and policies. It also uses self-play that trains value-and-policy networks using both game outcomes and recursive sub-searches applied to situations that came up in previous searches. 

Gopher

Gopher is a 280 billion parameter transformer language model from DeepMind that almost halves the accuracy gap from GPT-3 to human expert performance and exceeds forecaster expectations. It outperformed the current state-of-the-art for 100 tasks and competed with models like GPT-3 (175 billion parameters), Jurassic-1 (178 billion parameters), and Megatron-Turing NLG (530 billion parameters). 

Investments in startups

The top news of 2022 already is Google’s plans to invest $1 billion in Indian telecom operator Bharti Airtel to boost their digital offerings. So it comes with no surprise that a major part of the profit also comes from Alphabet’s investments in startups. For example, Alphabet’s investment arm, CapitalG, generated hefty returns from an early bet on Freshworks after the company’s shares soared by 30% on its opening day. Freshworks is a software company providing IT services to consumers. This was CapitalG’s sixth IPO in 2021, after UiPath, Duolingo, Robinhood and Oscar.

Waymo’s partnerships for driverless cars

Waymo already provides fully autonomous cars seven days a week all around Phoenix. Still, last year, the company partnered with several organisations to further their services and research in driverless cars. The company claims to have built the most advanced sensors and perception systems with quality data. The system dataset contains more than 20 million autonomously driven miles and five generations of development. 

Waymo announced delivering groceries curbside in San Francisco by early 2022, in partnership with Albertsons, one of the country’s biggest grocery chains, in a real-world test for the company. They have also partnered up with Geely, a Chinese automaker, to create a fleet of all-electric, self-driving robotaxis. Believed to be the richest dataset for autonomous tech, Waymo open-sourced their high-quality multimodal sensor dataset for autonomous driving, extracted from Waymo self-driving vehicles and covering various environments during different times of the day.