An AI fashion event, Fashion X AI, featured several designer-led AI fits created by an AI-based Interactive Design Assistant, or AiDA. With over 80 outfits and 14 designers, the event focused on creating outfits using software developed by Ph.D. students and experts from AiDLab, Hong Kong.
AiDLab’s CEO, Calvin Wong, said the idea was to educate people and fashion designers about artificial intelligence as a “supporting tool.” With AiDA, designers can benefit from several technologies like image recognition, image generation, detection, and more. The software’s algorithm also generates blueprints for future reference and enables the designers to add their sense of taste to the outfits.
Wong added that AiDA could generate more than a dozen fashion blueprints within ten seconds, saving a lot of designing time.
Cynthia Tse, a renowned merchandiser, said that she felt this is what the future of fashion looks like. She added that facial covering looked alien-like, yet intriguing. Another Hong Kong-based fashion designer, Mountain Yam, has been using AiDA for over six months and feels that it is very inspirational.
Yam said that a designer’s relationship with AiDA is like a romantic one where the former gradually gets to know her [AiDA], and she [AiDA] gets to know how the designer thinks. He added, “In accordance with my lines, styles, and databases, the system will propose something for me that I may not have ever considered, but she (AiDA) thinks is suitable for me.”
The fashion maven AiDA was launched at the Fashion X AI show and is still a work in progress—the developers will continue working on it and making it more beneficial to the fashion industry.
Over the past few years, many language models, like GPT-3, LaMDA, etc., have made it into the news because of their capabilities across multiple tasks, like text generation, problem-solving, sentiment analysis, and much more. Many companies are actively exploring the domain to harness this potential. Meta is one of the most successful companies working on natural language processing (NLP) technology, specifically for systems with over 100 parameters. These systems, called large language models, can potentially transform artificial intelligence research in the language and conversation domain. In the most recent development, Meta has introduced a new large language model (LLM) called OPT-IML.
As per Meta’s latest research, fine-tuning language models via pre-determined instructions can potentially enhance their functionality for newer tasks. However, there is a lack of understanding among researchers about instruction tuning. Following the same line of research, Meta created OPT-IML as a benchmark for fine-tuning large language models while scaling both the model and benchmark sizes. OPT-IML, or OPT for Instruction Meta-Learning, is a model of 2000 NLP tasks as a benchmark framework for evaluating all model generalizations. Using this benchmark, Meta presents how instruction-tuning works for OPT-30B and then for OPT-175B.
Core Technology: Instruction Fine-Tuning
Instruction fine-tuning entails optimizing LLMs using input formats designed for instruction on various NLP tasks. Researchers consolidated eight meta-datasets in a collection of 1,991 NLP tasks to put the technology into use. These tasks and instructions were stored in more than 100 collections and later transformed into the evaluating framework, as seen in the image below:
OPT-IML was instruction-tuned across three levels of generalizations:
Model performance on fully-held-out task categories not used for tuning.
Model performance on unseen tasks seen during instruction tuning (partially supervised).
Model performance on held-out instances of tasks seen during tuning (fully supervised).
The last condition assesses the generalization of supervised multi-task learning, while the first two assess cross-task generalization of instruction tuning. The diversity and distribution of tuning tasks, the manner of their prompts, and the targets used for fine-tuning all impact how effective instruction-tuning is for LLMs. The researchers found that instruction tuning is preferred only when it improves the performance of fully-held out and partially supervised tasks without compromising on fully-supervised task performance.
From thereon, researchers finetuned OPT similarly to a next-word prediction objective that is pre-trained based on previously inputted tokens. The training sequence was bifurcated into a source and a target sequence, out of which the researchers minimized the loss terms in the target sequence. Formally, fine-tuning a dataset D, containing instances sᵢ and their corresponding target tokens tᵢ = {tᵢⱼ}. Say that the model is pre-trained with parameters 𝛩, the researchers minimize the following loss function:
In the following steps, researchers fine-tuned the hyperparameters for all 30B models on 64 A100 Tensor Core GPUs and 175B models on 128 A100 Tensor Core GPUs. Ultimately, each 30B model is tuned for 4000 steps, and 175B models for double the number of steps with half the batch sizes.
The evaluation datasets comprised tasks with answer options and those without options. For the former category, the researchers used score-based classification of tasks based on the likelihood of an output. For the latter category, researchers decode a token until a maximum of 256 tokens are predicted. Lastly, researchers examined the results and concluded that OPT-IML outperforms standard OPT models (not fine-tuned) on all benchmarks. It is a specific outcome as standard OPT prompts take on from GPT-3 undergoing prompt engineering. At the same time, OPT-IML (with fine-tuning) enhances robustness and reduces the need for prompt engineering.
While the model is a significant improvement for instruction-tuning technology, it is yet subjected to some limitations. The evaluation framework used to tradeoff instruction-tuning variables may interact, resulting in different tuning settings. Additionally, all variable tradeoffs studied on 30B models may waver off from the usual trends at larger scales.
However, the future scope of instruction-tuning can benefit many areas, including multi-task learning, meta-training, and prompting. In multi-task learning (MTL), instruction tuning enhances generalization performance for fully-held-out tasks. Instruction tuning also betters prompting techniques that have become a dominant paradigm in recent years. From hereon, the researchers will continue to work on instruction tuning LLMs to improve their generalization abilities and find more applications for the technology.
The Wadhwani Institute for Artificial Intelligence announced a US$1m grant from Google to develop AI-based solutions for agriculture. The institute will use the grant to develop AI-powered tools that would give farmers rapid, accurate, and localized weather, crop, and other farm-related information while also enhancing the effectiveness and capabilities of India’s agriculture knowledge systems.
Google India’s Research Director, Manish Gupta, highlighted that over half the Indian population relies on the agricultural sector for their livelihood. However, the sector is further away from technological development that can lead to “pervasive benefits.”
He added, “We’re therefore happy to announce a new 1 million dollar grant to Wadhwani AI via Google.org to support their deployment of AI-based solutions to improve agricultural outcomes.” Creating more tech-driven solutions can help harness India’s agriculture dividend due to different terrains.
Small farmers rely largely on government and nonprofit organizations for guidance on what to do at each stage of the crop cycle. Shekar Sivasubramanian, CEO of Wadhwani AI, said that the company’s solutions ensure that all communities and farmers can meaningfully work with them through the government’s support.
He added, “It [the grant] will help us to identify and address multiple challenges affecting agriculture in India through the use of timely and effective AI-based interventions.”
In India, Delhi has the highest crypto adoption in terms of value invested, says the annual investor report by CoinSwitch. It was followed by Bangalore and Hyderabad. While in the tier-2 cities, Jaipur saw the highest adoption of cryptocurrency, followed by Lucknow and Pune.
The report titled ‘India’s Portfolio 2022’ provided insights into how the country has responded to a year of landmark events such as the Ethereum Merge, rising inflation, and market downturns.
Among the metros, the city with the highest crypto adoption was Delhi, with 7.87 percent, while the tech hubs, Bengaluru and Hyderabad, were at 4.87 percent and 3.27 percent, respectively.
Among the tier-2 cities, the pink city, Jaipur’s crypto adoption was recorded at 3.04 percent, followed by Lucknow at 2.02 percent, and Pune stood at 1.75 percent. It was also revealed that the country is most active in the crypto market during post office hours, i.e., between 6 pm and 9 pm. Further, Bitcoin will continue to be the most popular coin for retail investors in 2022.
Dogecoin and Ethereum were other popular crypto coins held by the Indians. Interestingly, Jaipur allocated the highest share of its investment in meme coins compared to the rest of India.
Also, eight percent of the crypto investors in India are female, and the investing patterns followed by Indian men and women were broadly similar. Further, it was revealed that more than 60 percent of crypto investors in Gurugram are under 25 years of age.
The Ministry of Electronics and Information Technology (MeitY) organized a session to discuss the vitality of the quality of datasets used to train artificial intelligence (AI) models. Shri Abhishek Singh, the President and CEO of NeGD (National eGovernance Division), chaired the session and hosted many industry professionals–AI enthusiasts, AI practitioners, and government officials.
The session started with discussions on data for AI, key initiatives of the Indian government to work on dataset quality, and the National Data Governance Framework policy. Others, like Shri Srikanth Velamakanni, CEO of Fractal Analytics, spoke on the state of existing data ecosystems, challenges, and innovation.
The AI Pe Charcha session also discussed the “Unlocking Potential of India’s Open Data,” a recently released report by INDIAai. NASSCOM jointly authors the report to drive innovation and facilitate the adoption of open government data platforms.
The Director & Co-founder of CivicDataLabs, Shri Gaurav Godhwani, spoke about his experience with open data platforms and the challenges organizations face in sourcing, developing, and ensuring open, high-quality data access.
AI Pe Charcha has been going on since India’s first global AI Summit, organized by MeitY, in 2022, as a part of Responsible AI for Social Empowerment (RAISE). The previous AI Pe Charcha session was held on October 28, 2021, themed “AI for Data-Driven Governance.”
Image Credit: Paulo Nunes dos Santos for The New York Times
On Friday, Google filed an appeal with the National Company Law Appellate Tribunal (NCLAT) against a Competition Commission of India (CCI) judgment regarding unfair conduct in the Android mobile device ecosystem. In October of this year, the Indian antitrust body fined the leading technology giant INR ₹1,337.6 (US$162 million) crores for allegedly abusing its influence with respect to the Android smartphone ecosystem.
Since a year ago, CCI has been investigating whether Google has established a formidable presence in five key markets: the market for licensable OS for smart mobile devices in India, the app store for Android smart mobile OS in India, the market for general web search services, the market for non-OS specific mobile web browsers and the market for the online video-hosting platform (OVHP). In its decision in October, the tribunal determined that Google’s demand that device makers pre-install the full Google Mobile Suite and mandate prominent placement of those applications imposed an unfair condition on the manufacturers and violated Section 4(2)(a)(i) of the Act.
The antitrust watchdog said that Google shouldn’t compel device manufacturers to include its suite of applications and that the search engine giant shouldn’t bar vendors from accessing its Play Services APIs and monetary and other incentives to vendors. The regulator also concluded that Google was indeed the market leader in those areas.
Google would have had to make major modifications to Android and pay a fine of INR ₹1338 crores if the decision had been carried out as written. Some major improvements include not preinstalling Google apps on Android, allowing the listing of alternative app stores on Play Store, and enabling users to choose the default search engine directly from the initial setup page.
After approaching NCLAT on Friday, Google announced that it decided to challenge the CCI’s verdict on Android because it feels the judgment is a huge setback for our Indian consumers and companies that rely on Android’s security features, as well as potentially increasing the cost of mobile devices. The company also stated that it thinks CCI was wrong to disregard the solid documentation provided by OEMs, developers, and consumers proving how the open Android business model encourages competition for the benefit of all parties, including those in India especially. Google is confident that the NCLAT will fully consider the evidence that has been submitted as well as the enormous contribution that Android has made to the rapid expansion and success of the Indian mobile ecosystem.
As per Financial Express’s notable sources, Google stated that it had not violated any rules set by the watchdog or the government. However, a token fine could have been imposed in accordance with the competition statutes if the CCI had determined that its actions were anti-competitive. Google believes that the CCI’s fines are penal in nature which are only enforced where there is mala-fide intent. Google also considers the patent rules, which it asserts have been maintained in this instance. According to sources cited by the Financial Express, the company is permitted under patent rules to impose restrictions on the use of open-source versions in order to maintain the ecosystem’s overall security.
The CCI fined Google another INR ₹937 crores (US$113 million) in October for allegedly exploiting the dominance of its Google Play Store and ordered the company to permit app developers to utilize other payment processing systems for in-app purchases or app purchases.
As governments worldwide have started to raise questions or probe about the influence of technology giants, Google is facing increasing scrutiny. In September, Google saw another setback in less than a year when Europe’s top court agreed with EU antitrust investigators that it had misused its dominant position and was slapped with €4.12 billion fine.
After discovering violations in Google’s negotiations with publishers to pay them to reuse their content, France fined the company 500 million euros in July of last year. Because of this, Google provided the first set of commitments in December 2021, which were examined, and in May of this year, the internet giant filed further commitments in response to the issues brought up by the French Competition Authority.
An appeal by Google against a decision by Italy’s antitrust authority to fine the company €10 million (US$10.36 million) was rejected by an Italian administrative court in November. The court claimed that Google had not provided clear and immediate information on collecting and using the data of those who accessed their services.
The International Chess Federation (FIDE) and Ava Labs, a Web3 company in charge of building and promoting the Avalanche (AVAX) blockchain, have forged a long-term agreement to bring blockchain technology to the world of international chess. By publishing tournament information and player rankings on-chain and hosting tournament prize pools through AVAX, the partnership would enhance game integrity and operational efficiency for players and federations.
Over 500 million people play chess worldwide, and FIDE competitions like the World Chess Championship and Chess Olympiad draw hundreds of millions of viewers. Avalanche and Core (a self-custody crypto wallet) will be promoted as sponsors in FIDE competitions, such as the ongoing 2022 World Rapid and Blitz Championships, and they will contribute to FIDE’s product development to bring chess to Web3.
The chess community’s growing adoption of digital platforms and FIDE’s Web3 integrations herald the optimistic scope of the crypto industry. According to the federation, over 100 million players actively play online chess and engage in over 25 million virtual chess matches daily.
Although interest in Web3 and non-fungible tokens (NFT) has waned in the general public in recent months, federations like FIDE still consider that the market has the ability to advance their sport. In an attempt to expand the game, World Table Tennis (WTT), the organization that oversees tournaments for the sport, stated in August that it would be pursuing Web3 projects in collaboration with non-fungible token startup NFT Tech.
The blockchain was selected by FIDE as a technical platform for its planned on-chain apps for chess players, according to an official announcement by the Avalanche team. Solutions that are avalanche-centric will be employed for tamper-proof experience tracking, transparent data processing, and other applications.
1/ FIDE is excited to formally announce a partnership with @avalancheavax to bring chess into Web 3. Since the birth of FIDE in 1924, the organization has been the proud steward of chess and has been a part of championing innovation in the sport.https://t.co/bymmy08dBVpic.twitter.com/inoh1ZrrgZ
The major objectives of Avalanche’s technical solutions will be to leverage the subnet’s technology to demonstrate individualized experiences through public services, share tournament data on-chain, and compute official player rating on-chain. Avalanche intends to develop a FIDE game explorer as the next stage in the new partnership, which will be entirely run on on-chain data collected by Avalanche’s smart contracts.
After Argentina’s World Cup triumph, the professional football league is all set to foray into Upland’s metaverse via an official licensing agreement between the Silicon Valley-based firm and the Argentine Football Association (AFA).
The deal allows the AFA to take the Liga Profesional de Fútbol (LPF) on Upland’s metaverse platform, enabling fan engagement between Argentine football teams, players, and fans.
Gamified opportunities are expected to include several Argentine football digital collectibles, including tickets, highlights from games, historical moments, teams, clubs, players, and other exclusive offers.
The partnership also offers exclusive ownership of highlight videos of moments of the LPF matches. The sale of various nonfungible tokens (NFT) will provide the Argentine top-flight league with a source of income as users buy and trade the licensed digital assets.
Upland will also allow users to start virtual businesses to resell LPF digital collectibles in user-operated and owned shops located on virtual properties. The metaverse platform is creating a digital version of the world to offer digital land ownership, asset creation, and a virtual economy.
Metaverse is not a specific type of technology but an umbrella term for how people interact with multiple latest technologies. Broadly speaking, “the metaverse” may include virtual reality, consisting of virtual worlds, and augmented reality, a combination of physical and digital worlds. Another way to closely define the term is to consider it as a graphical and hypothetical iteration of virtual spaces where people can interact with each other, work, play, and socialize. Many tech companies have ventured into the metaverse space in tandem with their vision of a new digital economy and interoperable platforms. These companies aim to allow people to exchange virtual items from one platform to another. Many top Indian companies working on metaverse also aim to build and develop technologies to interact with virtual worlds.
Metaverse offers endless possibilities, but the most prominent is the potential for a full-fledged and decentralized economy with both physical and virtual elements. The concept goes hand-in-hand with two others, NFTs and cryptocurrencies. NFTs, or non-fungible tokens, will be the currency units to make transactions and work as proof of ownership. NFTs can also have non-economic value because they are worth whatever someone is willing to pay for them, based on some “intangible” criterion, like desirability. On the other hand, cryptocurrencies have a similar implication of being used in transactions, but they have a purely economic value.
The takeaway is that these technologies are a core element of the idea of a metaverse. This article enlists some of the metaverse companies and briefly discusses their approach and achievements in metaverse, cryptocurrencies, and NFTs.
Top Indian Metaverse Companies
HTC
HTC is a tech innovation company focused on developing powerful products and solutions in mobile and immersive technologies. It began with a vision to put a personal electronic device with the latest technologies in the hands of every customer. The company is now applying those technologies in evolving smartphones and connecting them virtually by venturing into the era of the metaverse.
For the past few years, HTC’s mobile division has been looking for the best approach to rekindle its mobile technologies by shifting to the metaverse. In 2018, HTC released Exodus; a smartphone focused on blockchain and cryptocurrencies that was ahead of its time. But it did not gain momentum then.
With the rising popularity of metaverse companies, many smartphone companies like HTC are aligning themselves with the new age of virtual reality. Now, HTC has re-entered the market with the HTC Desire 22 Pro, the “metaverse smartphone.” The device will be compatible with NFT and crypto functionality and was released as a companion device to the company’s VR headset HTC Vive Flow.
The highlight is the company’s Viverse, a platform leveraging AI, AR, and VR, enabling users to create their own virtual spaces and make NFT transactions.
Capgemini India Pvt Ltd
Capgemini is an IT consulting firm specializing in engineering and technology professional services. The company is striding towards the metaverse to provide its clients with a premium interface that will help them create advanced applications featuring blockchain-related attributes. Capgemini has a dedicated metaverse lab to work on newer technologies and enhance customer experience.
Capgemini won the HeroX/Airbus metaverse challenge, becoming one of the most potent metaverse companies. The challenge aimed to draw ideas for enhancing the passenger experience and bringing virtual reality into air travel. The winning entry by Capgemini proposed the virtual Airbus town square that would come alive after the plane takes off and serve as a hub of experiences. Passengers simply have to log in via tablets/smartphones and use VR headsets to experience the virtual township.
The company also recently entered into a partnership with Unity, a real-time 3D content-creating organization, to build on business opportunities and benefits of metaverse experiences across multiple industries.
TCS
The next on our list of metaverse companies is Tata Consultancy Services, a tech consulting firm that offers a cognitive-driven approach to business, technology, and engineering service provision. Around 2017, when TCS was working with an undisclosed European airline client, the company started its metaverse adventure. The airline company asked to teleport photorealistic avatars for official meetings. The company built a platform called Avapresence, a combination of the words “avatar” and “presence,” and leverages XR, blockchain, cloud, AI, and IoT to help clients.
Recently, TCS began working with tech professionals in Ireland to build a new global delivery center in Letterkenny to service companies in the country. The center will work with clients interested in being a part of the metaverse and transforming digitally.
TCS is also engaging with other Tata group companies like Tata Motors, Croma, and Tanishq to deploy metaverse solutions. TCS is also developing a plan for the Tamil Nadu government to introduce a metaverse museum.
Zensar Technologies
Zensar Technologies is an Indian software and services company that provides tech solutions to enhance the digital experiences of its customers. The company utilizes AR/VR integrated technologies to create digital items based on non-fungible tokens. Zensar is also involved in modernizing legacy technologies, migrating to the cloud, and turning intelligent data into actionable insights.
Zensar Tech is exploring MaaS (metaverse-as-a-solution) as one of the uniquely emerging Indian metaverse companies for providing integrated services and implementing virtual reality in enterprise applications. It aims to enhance the productivity of enterprise workers, build self-learning systems, leverage MLops for automation, and maximize business ROI.
Some Indian Metaverse Startups
Many Indian startups have been on the path to joining other metaverse companies over the last few years. Some of them are mentioned below.
LOKA Metaverse
LOKA is one of the first few metaverse companies to launch an application that allows users to interact with people worldwide. It surfaced in 2020 amidst the COVID-19 pandemic. Its platform provides users with a multiplayer gaming application (available for Android and iOS) designed for social engagement, gaming, E-commerce, content, and entertainment. The app features actual locations and provides real-world experiences, allowing users to have genuine gaming and social engagement experience. LOKA was also featured on Shark Tank India and raised angel funding of ₹1.33m from Anupam Mittal, founder of Shaadi.com.
Zippy
Zippy is one of the most famous metaverse startup building ‘the third format’ of running by developing a metaverse for runners. It is an innovative mobile application that generates an immersive indoor running experience. Founded in July 2021, Zippy’s metaverse idea preserves all indoor and outdoor formats required for running, bringing it naturally to the people like other physical exercises.
Users typically need a treadmill and a computer, tablet, or smartphone to download and run the Zippy application (marketed as “Zippy Fit” on both iOS and Android platforms). Once a user gets on the treadmill, they are “teleported” into a virtual world where they can run with their digital twins or avatars through iconic marathon locations like Boston, London, Mumbai, and Tokyo or beautiful landscapes like a jungle path or beach.
NextMeet
NextMeet is a VR-based platform that allows users to generate 3D avatars and interact with their friends in real-time. This is yet another startup story that gained recognition during the COVID-19 pandemic to provide people who have been displaced from their offices, schools, and colleges with a platform where they can have an immersive experience via virtual reality. The app was launched in October 2020. After three upgrades, it now features four virtual environments: classrooms for education, meetings for businesses, conferences, and expos for commercial purposes, making it one of the most purpose-driven metaverse companies in India.
The platform has recently been tested by Amity University, Reputation Today magazine, and many other institutions. In the VR paradigm, schools have held summer camps and storytelling events where kids can complete tests while being physically present with their teachers and peers.
Invact
Invact Metaversity is one of the ed-tech metaverse companies developing immersive learning experiences for students across the globe. Their objective is to provide a possible campus-like environment in the metaverse, making education both experiential and accessible. No matter where they are physically located, students can communicate with their teachers and peers via animated avatars thanks to the platform. With Invact Metaversity courses, students can experience peer-to-peer learning with added community layers that other online learning platforms do not offer. The platform currently offers two courses, the metaMBA in Marketing and the metaMBA in Product.
Wall App
Non-fungible tokens, or NFTs, have gained momentum over the last few years. Realizing the need for a reliable platform allowing access to NFT trends, data, and analytics, Anuj Kodam, and Amarnath JV felt the need to develop Wall.app, a Web3 project to build a social NFT discovery platform. With Wall.app, users may explore NFT statistics on Ethereum, including volumes, price action, sales, and other leading NFT collections. The app also stores information on particular whale wallets that influence markets, significant sweeps in the industry, etc.
The startup plans to become a Solana-driven metaverse company, a decentralized blockchain, as the next chain for its chain-agnostic backend.
Atirath Gaming Technology
Atirath Gaming Technology is one of the many metaverse gaming companies that create mobile games involving strategy, puzzles, and even casual games with elements of sub-continent cultures. Many games like Assassin’s Creed and God of War are based on mythology and perform exceptionally well worldwide. With the same belief, the company, launched in mid-2020, focuses on Indian mythology and creates applications with a global appeal of making people aware of each other’s roots and traditions.
It plans incorporate AR/VR technologies in its upcoming application Asva, developed after raising an undisclosed amount of seed funding from venture capitalist Kalaari Capital in February 2022. The investment will go into improving the company’s services, growing its workforce, and gearing up for the metaverse.
A TikTok video posted by foodiemunster shows a branch of McDonald’s in Texas, US, experimenting with robots to serve food. As per the 29-second video clip, the food is being served via a conveyer belt from an empty restaurant with no visible human presence.
The video demonstrates how clients can place an order using computerized screens and pick it up via a machine. The restaurant has an “order ahead lane” and is located outside Fort Worth, Texas. The meals are delivered to consumers via a conveyor belt, kiosks for placing takeout orders, and meal pick-up shelves.
Keith Vanecek, the franchisee operator of the Texas branch, said, “the technology in this restaurant not only allows us to serve our customers in new, innovative ways, it gives our restaurant team the ability to concentrate more on order speed and accuracy.”
As per McDonald’s, customers notice a significant difference when they enter the restaurant. Firstly, the restaurant is considerably smaller than the traditional restaurants. This is because the features used inside and outside are aimed at customers planning to dine at home or on the go. McDonald’s also wrote that there are dedicated parking, curbside order pick-up spaces, and parking for delivery drivers outside the branch.
While the advancement is exciting, many fear its consequences on the job market, as the fully-automated restaurant is a response to the struggle for a higher minimum wage in Texas.
