Tech Talk: The 10 digital trends shaping our world

Why we need more women in AI

What do women think about AI? In August 2022, I wrote about the findings of a survey by the Pew Research Centre that revealed a significant fact that most of us would have had a gut feeling about—women think differently about technologies such as artificial intelligence. They are more concerned about the negative effects of AI; are not exactly happy with driverless cars; are not sure whether AI applications like facial recognition are good or bad for society; and are more likely to support the inclusion of a wider variety of groups in AI design.

Given that AI is a male-dominated field and that most AI research and development is predominantly in the hands of men, it’s safe to conclude that this is one of the reasons why women view AI with suspicion.

A 2019 UNESCO report pointed out that women represented only 29% of science R&D positions globally and were already 25% less likely than men to know how to leverage digital technology for basic uses. A year later, a World Economic Forum report revealed that women made up only 26% of data and AI positions in the workforce. The Stanford Institute for Human-Centered AI’s 2021 AI Index Report found that women made up just 16% of tenure-track faculty focused on AI globally.

India is no different. According to a 2020 United Nations report, nearly 40% of STEM graduates from India were women, which is among the highest in the world. However, only 14% of the country’s 280,000 scientists, engineers and technologists were female, exposing the yawning gap in the female talent we see in the workforce.

Researchers have repeatedly pointed out that AI and machine learning models would always produce biased results as long as AI continues to be a male-dominated field. Many women, according to a 2021 Deloitte survey, feel they are not treated equally as men in AI, which prompts them to leave the field.

Read more about this here.

Is decentralised Web3 a Utopian idea? Is Fediverse here to stay?

In September 2022, I wrote that Web3 as a concept had many positives—at least in theory. DeFi, or decentralised finance, to begin with, is likely to be the first killer app of Web3, which could dramatically increase the level of control borrowers have on their data. The Data Empowerment and Protection Architecture aims to give citizens and businesses control over their financial data, which currently still sits in bank platforms with a complicated account aggregator structure to release this to users of this information.

The national blockchain roadmap proposed by the ministry of electronics and information technology in December 2021 recommends several use cases related to non-fungible tokens, or NFTs, including for land records, student certificates, and patient health records.

NFTs, which represent ownership of digital assets in Web3, could also change the way we think about virtual property. Already, digital art and music are successful use cases of NFTs. Unfortunately, though, Web3 itself seems to be getting very ‘centralised’ as a large part of the funding of successful Web3 companies is by large corporate-controlled VC firms. Even a blockchain-based Web3 metaverse will be an “open metaverse” only if built on open standards.

Further, it is not easy to get information and insights out of blockchain systems. Experts also rightly note that centralised identity databases such as Aadhaar and KYC regulations do not co-exist naturally with pseudonymous accounts on a blockchain.

Moreover, Web3 cannot prevent cybercrimes. About $1.7 billion worth of cryptocurrency was stolen between January and April, with 97% of those crimes taking place on DeFi protocols, according to Chainanalysis. Last, but not the least, valuations of Web3 startups are rising without any clear business models for monetisation. Read more about it here.

In this context, you may also want to read about the rise of the Fediverse, or the federated universe, in ‘Meta’s Threads may boost the rise of a new social internet’. The Fediverse is a network of interconnected social media platforms including Mastodon, Pleroma, Pixelfed, GNU Social and PeerTube.

In this Mint Explainer, I go in how the decision of Meta’s microblogging site, Threads, to join the Fediverse has underscored the need for, and significance of, a truly decentralised social internet that enables users from one platform to interact with those on other platforms, even if they do not have an account on those networks.

India’s digital currency and rise of Beckn protocol

In October 2022, I wrote about how the world has come a long way from the days of bartering goods to issuing official currencies while also moving from just coins to paper. Mobile payments and virtual currencies like bitcoin have further transformed the market.

It appears, though, that money will keep changing form, given that while governments are wary of cryptocurrencies, they seem to like digital currencies backed by their respective central banks. These central bank digital currencies, or CBDCs, are essentially tokens issued on a blockchain to represent a country’s national or fiat currency. The use of CBDCs can also help to reduce costs.

The 2021 Bank for International Settlements (BIS) survey on CBDCs conducted on 81 central banks revealed that 90% of central banks were engaged in some form of CBDC work, and that more than half were developing them or running concrete experiments. New research from Atlantic Council’s CBDC tracker reveals that 130 countries are now exploring a CBDC, representing 98% of global GDP. About 64 countries are now in the advanced phase of exploration (launch, pilot, or development).

Notably, the European Central Bank (ECB) announced a preparation phase to lay the foundations for a digital euro, and PetroChina, a Chinese oil and gas company, completed the first international crude oil trade using the digital yuan (e-CNY).

India has joined other nations such as Australia, China, and the US in announcing pilots for its very own CBDC, also known as the ‘Digital Rupee’. As of 30 June, more than a million users as well as 262,000 merchants had registered for the pilot on retail CBDC transactions.

The Open Network Digital Commerce (ONDC) is built using an open-source, decentralised protocol called Beckn, which is used for location-aware, local commerce. It’s an application layer protocol that is compatible with transport layer protocols such as TCP/IP (transfer control protocol/internet protocol) and UDP (user datagram protocol), both of which are used for communication on the internet.

You may liken Beckn for digital commerce to what HTTP (hypertext transfer protocol) means for the World Wide Web or SMTP (simple mail transfer protocol) for emails.

Put simply, the Beckn protocol allows computers online to understand what phrases like ‘select an item from menu’ or ‘place an order’ mean with the help of open, standardised, machine-readable information. It enables services and sales offers to be discovered across industries and fetched by any consumer application. Beckn leverages 5G and high-speed network technologies to implement e-commerce capability in the network and/or the transport layer.

Other than ONDC, Beckn also powers Bengaluru’s Namma Yatri autorickshaw-hailing mobile app. This will eventually be extended to taxis and public transport. And there will be many more applications. Beckn’s open digital infrastructure allows for the creation of an unbundled and decentralised digital market, akin to an open playground that allows everyone to have a fair playing ground. I argued that while the Beckn protocol may not share the hype of Worldcoin, it can certainly show the world how to democratize digital commerce.

No! The metaverse is not dying… and AI is certainly not killing it

Meta’s (formerly, Facebook) grand experiment in virtual reality will have racked up more than $50 billion in operating losses, according to a recent Financial Times article, which indicates that interest is virtual reality and the metaverse is on the wane. Facebook changed its company name to Meta in October 2021 to reflect its new focus on the “metaverse”, even as many pointed out that the move was more to negate the negative press around Facebook.

The following month, Meta decided to let go more than 11,000 of its employees, or 13% of its workforce, as it “shifted more of our resources onto a smaller number of high priority growth areas—like our AI discovery engine, our ads and business platforms, and our long-term vision for the metaverse”.

While everyone had a fair idea of the losses that Meta was reporting because of the billions of dollars it had poured into its metaverse project, the figures disclosed during Meta’s earnings call in February 2022 told the complete story when Meta said it had lost more than $13.7 billion in its “Reality Labs” unit, which houses its metaverse efforts.

Many speculated then that it was the end of the metaverse for Meta, given the rise of generative AI, or GenAI, tools such as OpenAI’s large language model (GPT-4) and its smart chatbot ChatGPT.

The metaverse is essentially a 3D representation of physical assets in the real world—an ecosystem with gaming, payments as NFTs, buying and selling, and many more immersive experiences—developed with the help of technologies such as augmented reality (AR), virtual reality (VR), mixed reality (MR), software, content, and, of course, hardware, including servers, and headsets. Many people still believe that a metaverse is a place just for gamers or Web 3.0 investors, but nothing can be further from the truth.

For instance, digital twins can be linked with real data sources and are able to update their twin in real-time. And enterprises are already building digital twins of their manufacturing plants and factories to identify missteps in the production process. We could have a product twin representing a product, or a production plant twin, which could represent an entire manufacturing facility. We could also have a procurement and supply chain twin, often called a network twin. One could also have an infrastructure twin.

Daimler, for example, allows customers to test-drive a vehicle without even getting inside a car. Given that every asset, process, or person within and related to an enterprise will be replicated virtually, and connected, enterprise metaverses will enable employees to gain real-world product design experience and training from their desks as they manipulate 3-D digital replicas of equipment, according to McKinsey. You may read this article for more information: Metaverse: A new lease of life with Apple’s “spatial computing”. I believe that it’s premature and naive to write-off the immersive power of the metaverse (does not apply to Meta’s metaverse).

Generative AI can spur economic growth but job losses, privacy remain grave concerns

When I wrote this piece in June, generative AI had been in the news almost daily for the previous six months or so because of its ability to transform businesses by generating or creating new content (text, images, audio, video, etc.) with the help of ‘prompts’ in natural languages like English and Hindi. Also because of its potential to put about 300 million full-time jobs at risk due to automation, according to a Goldman Sachs March report titled ‘The Potentially Large Effects of Artificial Intelligence on Economic Growth’.

McKinsey researchers, too, estimate that 50% of today’s work activities could be automated between 2030 and 2060, with a midpoint in 2045. This prediction is roughly a decade earlier than in their previous estimates. Given that the potential loss of jobs is a very sensitive issue around the world, and understandably so, these predictions get a lot of press.

What, however, gets ignored is the productivity boost that AI can provide. For instance, in the above-cited report, the Goldman Sachs researchers had also concluded that generative AI could also boost global labour productivity in many countries, including in India, and eventually increase annual global GDP by 7%, or almost a $7 trillion increase in annual global GDP over 10 years “if it delivers on its promise”.

That said, when writing about GenAI, a recurrent theme is the fear that these models will think and act like humans, plagiarize artists’ work, and replace thousands of routine jobs. The reason is that, unlike traditional machine learning models, which typically make predictions based on historical data, generative AI models can create entirely new content, including audio, code, images, text, simulations, and videos.

In a white paper released jointly with Accenture, the World Economic Forum talks specifically about the impact of large language models (LLMs) on jobs. It posits that with 62% of total work time involving language-based tasks, the widespread adoption of LLMs such as ChatGPT could significantly impact a broad spectrum of job roles. The study analysed over 19,000 individual tasks across 867 occupations, assessing the potential exposure of each task to LLM adoption. You can read the article here.

However, while individuals may fret over the loss of jobs, data privacy appears to be the primary concern of companies using generative AI, followed by transparency, data poisoning (datasets can be deliberately “poisoned” or “polluted” by hackers, resulting in inaccurate results), and intellectual property and copyright violation. You may read the article here.

Why scientists want to merge human brain cells with AI

In August, I wrote about how a group of researchers is using lab-grown brain cells embedded onto silicon chips, thus merging the fields of AI and synthetic biology “to create programmable biological computing platforms”. The fact, though, is that whenever we speak about AI using artificial neural networks, the ANNs do not really work like the biological human brain that keeps on learning new things continually.

Current AI, on the other hand, suffers from “catastrophic forgetting”, a major challenge that ANNs face. When they learn a new task, they tend to forget abruptly and entirely what they had previously learned. Essentially, ANNs tend to overwrite past data with new knowledge.

Researchers from the Institute of Computer Science of the Czech Academy of Sciences in a paper published last November wrote about the use of a spiking neural network model that could simulate sensory processing and reinforcement learning in an animal brain, to blend “new task training with periods of off-line reactivation, mimicking biological sleep”. You may read the article here.

Is Paytm building an AGI software stack? Or is it AGI-washing?

In his 23 August letter to shareholders as part of Paytm’s 492-page 2023 annual report, Vijay Shekhar Sharma, founder and chief executive officer of Paytm, said that as Paytm focuses on “small mobile credit with high credit quality” that fully complies with regulatory guidelines, it will require “sophisticated capabilities in AI and other technologies”. He added that Paytm was building “an India scale AI system to help various financial institutes in capturing possible risks and frauds, while also protecting them from new kinds of risks due to advancement in AI”.

Further, Sharma said Paytm “is investing in AI with an eye on building an Artificial General Intelligence (AGI) software stack”. He added that by building this stack in India, Patym was not only making it “our country’s tech capability” but also “creating something that could be leveraged outside India”. In May, too, he spoke about how the “advent of early-stage AGI in 2023″ would help make Paytm’s business more efficient.

Developing an advanced AI system is logical for the financial sector, an early adopter of AI tools. However, while talking about building an AGI stack and exporting it to other countries may tick the ‘Made in India’ box, it is a completely different proposition. It raises many questions, especially since Paytm is now a listed company on the Indian bourses, with its shareholders and potential investors keenly listening to the management discussion to understand the company’s roadmap. You may read the article here.

How AI is helping synthetic biology come of age

Early this month, I wrote about how synthetic biology—as opposed to the natural study of life, or the biology that we learnt in our schools—can artificially redesign organisms with new abilities (human-made from chemical parts), and has applications in fields such as medicine, manufacturing, and agriculture. Synthetic biology combines multiple areas including molecular biology, biophysics, biotechnology, and genetic engineering, and typically uses two techniques: top-down and bottom-up.

The top-down technique involves redesigning and fabricating existing biology systems to produce synthetic products, while the bottom-up approach is commonly used to design and construct genetic circuits by piecing together functional modules that are capable of reprogramming cells with novel behavior. Scientists, of late, have begun exploring combining the two approaches for best results. You may read the full article here.

Need more local LLMs, regional language datasets

India is home to more than 400 languages, making it one of the most linguistically diverse countries in the world. A language like Hindi alone has 48 official dialects, and Bengali is said to have about 50 variants. But large language models are trained primarily using internet data, which is predominantly English. As per Statista, English was the most popular language for web content, representing nearly 59% of websites as of January 2023. Russian ranked second with 5.3% of web content, followed by Spanish language with 4.3%.

It’s logical, then, that AI models need to be trained using local languages (rather than English) to bridge the digital divide. Further, a July 2022 report by the Internet and Mobile Association of India and Kantar pegged India’s active internet users at 692 million, including 351 million from rural India and 341 from urban India. The report estimates there will be 900 million internet users in India by 2025. Voice and Indic languages, thus, will prove to be key drivers of internet growth in the country.

In this context, it’s heartening to see the progress that the government’s Bhashini project is making in building a national public digital platform for languages to develop services and products for citizens by leveraging the power of AI and other emerging technologies. You may read more about this here.

Who will win the AI race?

In May, I wrote that it’s silly to bet on any one company to win the AI race because there are too many factors that could rapidly change the equations in this field. The factors include partnerships, acquisitions, the emergence of new technologies, and global regulations. I also wrote about what the major tech companies were doing to stay relevant in the AI race by enhancing their products and platforms while partnering and investing in other AI startups. You may read the article here.

Since then, Google Gemini has already been launched in three versions, making everyone wonder if Gemini can help Google regain its AI mojo. Further, Elon Musk-led xAI has announced that its generative AI-based chatbot Grok AI will now be available in India and 46 other countries, including Pakistan, Australia, Canada, New Zealand, Singapore, and Sri Lanka. Musk-owned X (formerly Twitter) has provided access to Grok to Premium+ subscribers in the US.

And in November, OpenAI CEO Sam Altman told the Financial Times that GPT-5 was in the early stages of development. “Until we go train that model, it’s like a fun guessing game for us—we’re trying to get better at it, because I think it’s important from a safety perspective to predict the capabilities,” Atlman told FT. Read about it here: “But I can’t tell you, ‘Here’s exactly what it’s going to do that GPT-4 didn’t.”

This article is republished from Mint’s weekly ‘Tech Talk’ newsletter.