‘On-device AI is a better bet for India’


It’s barely two years since ChatGPT appeared on the scene and dazzled the world with glimpses of never-before-seen capabilities of artificial intelligence (AI). Yet, already, AI applications running on one’s phones or laptops no longer seem a novelty. However, much of this AI workload runs on the cloud, which offers vast storage but poses challenges such as slower connectivity, energy consumption and network congestion. This means user experience in cloud-based AI apps remains subpar, spurring manufacturers of smartphones, computers and even automobiles to support development of local AI inferencing. Chip majors like Qualcomm, too, have thrown their weight behind on-device AI.

Qualcomm’s AI-fuelled innovations for mobile phones

Businessline caught up with Dr Vinesh Sukumar, senior director, product management, Qualcomm Technologies, at the company’s recent Snapdragon Summit in Hawaii. With over 20 years of industry experience, Sukumar leads the generative AI and machine learning (GenAI/ML) product management team across multiple business units. Edited excerpts from the interview:

What does running AI apps on devices mean for a market like India?

When it comes to AI deployment, we at Qualcomm have always thought about which user pain point we can solve, and what value we can add to the device to justify customer refresh or upgrade. Improvements in photography and videography quality without much user intervention remain important use cases for OEM (original equipment manufacturer) partners. In the PC (personal computer) market, instantaneous productivity increase is a key request. In India, there is a lot of innovation in mobile devices, but other form factors such as automotives are still challenging. For instance, the success of ADAS (advanced driver assistance systems) is still difficult, but the use case of car infotainment is in demand. When you introduce any of these new features, it’s not going to be a splash right at the first try. It’s about more people getting used to it and making it better, depending on the maturity of the market.

How do you balance pricing with user experience, especially in the price-conscious Indian market?

There will always be a grading of experiences at different price points. You have to decide how much [AI processing] will be on-device, how much device-to-cloud, or which features you don’t enable at all. For instance, in video streaming, if something is at 30 frames per second, you may drop it at 15 or 24 frames, which the compute power can handle. Similarly, in on-device AI-based voice translation, we may go with fewer languages at lower price points… we continue to work with OEMs to have more even in lower price points.

Is India better equipped for AI workloads on-device rather than on cloud?

On-device AI is slowly growing in India, starting with smartphones, and will catch up on computers, wearables, etc. Given the population, utilisation of devices will be higher here than in other markets. Also, users in India interact with their phones a lot more for daily tasks like digital payments. That’s where data privacy and security concerns also come in, with preference to keep data within the device and not send to the cloud. The cloud computing infrastructure is still coming up in India, and the latency — or response time of going back and forth to the cloud for every task — could be more. So, the preference for on-device AI inferencing sits well for India. Moreover, on-device AI is less energy-intensive compared to the cloud.

Can one get these experiences on mid-premium or budget smartphones too?

The entire Generative AI boom started with a 70-billion parameter model. But with time, we have moved to smaller AI models. If these models can get down to a billion or less parameters while maintaining accuracy, you can deploy it in phones with less memory/storage or compute power. Qualcomm is pushing for smaller AI models along with our partners like Meta, Mistral and others. 

Your competitors are eyeing AI software for data centres. What makes Qualcomm optimistic about on-device AI?

When Generative AI first came in, there was only talk of ChatGPT and Dall-E, and only from the lens of text or images. Qualcomm has shown performance improvements with on-device AI, whether it’s image generation on devices or smaller models such as Google Gemini Nano. We are aiming for more personalisation in the Generative AI experience — mapping the user’s style into the AI, which is where on-device helps. Next is going ‘app less’. Can I use my phone’s voice command to buy a ticket for a trip, and the AI knows my preference of airline, seats and other things? It is a difficult problem to solve, but we are marching towards it.

What about data privacy and sovereignty regulations for on-device AI?

Many governments are putting in standards to ensure AI engines are responsible, don’t have bias, and the generated synthetic content meets the standards. These regulations are still being defined. We then have to test it in the real world. Qualcomm is working closely with governments on this and I expect more developments by 2025 or 2026. 

(The writer’s visit to Snapdragon Summit, Hawaii, was sponsored by Qualcomm)





Source link

Why India wants 1 GeV particle accelerator for thorium


The Department of Atomic Energy plans to build a 1 giga-electron volt (GeV) particle accelerator — a machine that will help convert thorium, abundantly available in India, into nuclear fuel. The idea is to build a continuous wave, high-intensity proton accelerator, Quantum has learnt. 

A particle accelerator is like a gun — only it has subatomic particles (electrons, protons and neutrons) instead of bullets. You shoot these particles on to something to get an effect — like converting thorium into uranium-233, a fissile material that can be used in a nuclear reactor to produce electricity. 

India has many particle accelerators (cyclotrons and synchrotrons), but at 30 mega electron-volt (MeV); none are in the GeV range. 

Dr Anil Kakodkar, renowned nuclear scientist and former chairman of the Department of Atomic Energy, tells Quantum that the use of particle accelerators is one means of tapping India’s immense thorium resources. This, in turn, can guarantee India’s energy security. 

Leveraging thorium

“There are a number of pathways to leverage thorium for nuclear energy production,” Kakodkar says, mentioning three. 

The first involves breeding uranium-233 by irradiating thorium in a nuclear reactor. In a fast-breeder reactor, one can produce more fissile material than what is consumed. The surplus fissile material can be used to set up additional nuclear power capacity. This pathway is part of India’s three-stage nuclear power strategy. 

The second method involves using thorium along with uranium in a high burn-up configuration in a reactor to derive surplus energy from thorium through in situ fission of generated uranium-233. This is an upcoming method, thanks to a company called Clean Core Thorium Energy, which is developing a nuclear fuel consisting of uranium-238 enriched to 14-15 per cent and thorium. Its ANEEL fuel is currently undergoing irradiation tests at Idaho Laboratories, USA. 

The third pathway is the use of high-energy (more than 1 GeV), high-current proton accelerators to create neutrons to breed fissile uranium-233 from thorium, either directly or in an ‘accelerator-driven subcritical reactor system’ (ADSS). This is why the DAE wants to build a 1 GeV accelerator. 

Select grouping

“India is pursuing all these strategies,” Kakodkar said, adding that they complement each other. 

The DAE plans to have two more medium high-energy accelerators. A second 1 GeV high-current, pulsed proton accelerator for generating neutrons (spallation neutron source) is planned for scientific experiments such as probing the atomic structure of materials. There are technical differences between the two 1 GeV accelerators.

The third is a ‘synchrotron radiation source’ to produce tunable X-rays or UV light, extremely useful in scientific experiments. 

These particle accelerators, when they come into being, will put India in an elite group of countries that have 1 GeV accelerators.





Source link

Electrochemical means of producing ‘green’ soda ash


Sodium carbonate (aka soda ash) is a key industrial chemical, but its production emits copious amounts of carbon dioxide — 800 kg per tonne. The conventional method (Solvay process) involves burning limestone with coke to produce carbon dioxide, which reacts in the subsequent steps with brine, ammonia and lime to produce soda ash. It also produces calcium chloride and wastewater, which need to be disposed of. 

The Fraunhofer Institute for Ceramic Technologies and Systems, Germany, has developed an alternative means to produce soda ash. At the heart of the electrochemical route to produce ‘green soda’ is bipolar electrodialysis, a process that uses ultra-thin membranes. The pores in these membranes are so tiny that only individual ions can pass through them. The membranes function as an exchange medium by allowing only the negatively charged anions or positively charged cations to pass through. 

This means that a mixture of salt and water — brine — is split into sodium hydroxide and hydrochloric acid. Hydroxides are salt-like compounds that form sodium hydroxide on contact with water. Then, when carbon dioxide is added to the sodium hydroxide, the final product is soda. 

“This lets us produce sodium carbonate without spewing greenhouse gases into the atmosphere and without harmful industrial wastewater increasing the salinity of rivers or other bodies of water,” says Hans-Jürgen Friedrich, group manager for technical electrolysis.

Sustainable chemical generation of amides

Amides are essential in chemistry, serving as key components in a wide range of organic compounds, including proteins, pharmaceuticals, and synthetic materials. Traditional amide synthesis often requires high temperatures and harsh conditions, leading to significant environmental impact and inefficiency. These conventional approaches typically involve transition metal catalysts and generate substantial waste, prompting the need for more sustainable alternatives. 

Scientists at the SN Bose National Centre for Basic Sciences have found a green and efficient chemical process for preparing amides that can revolutionise industrial manufacturing of pharmaceuticals and synthetic materials. 

The scientists have developed a novel method for synthesising amides from alcohols using a covalent organic framework (COF) as a photocatalyst under red light irradiation. 

This catalytic method can be helpful in chemical processes across industries — including pharmaceutical manufacturing, materials science, and green chemistry — offering a more sustainable, efficient, and recyclable approach to creating vital chemical structures, says a press release. 

The advantages of this method include mild reaction conditions, high efficiency, excellent recyclability, and the practicality of red-light activation, which is less harmful and penetrates more effectively, making it suitable for large-scale applications. 

The implications of this research are significant. In the pharmaceutical industry, this method could streamline drug production, reduce costs, and eliminate metal contamination. In materials science, it could enable the development of new polymers and materials with amide linkages, expanding the range of materials for various applications. 

“Further research may optimise the COF structure for even better performance and stability, and scaling up the process for industrial applications will be crucial to realising its full potential,” the release says.





Source link

Denizens of the deep: Mapping the ocean bed


The sun, we have been told, is the primary source of nutrition for every living thing on earth. But that is not entirely true. 

A recent study has shown that in the depths of the oceans, over 8,000 feet below, there are creatures living under the seabed, at the hydrothermal vents. These animals — they are not just microbes, but way bigger — get their nutrients from the mixture of magma and seawater. 

This discovery illustrates a point, hidden in plain sight: while we train our sights deep into the universe, with gizmos like the James Webb telescope, little do we know about what lies beneath, in the depths of our oceans. 

Fortunately, the scientific community has woken up to this miss. In 2018, the Nippon Foundation began a collaboration with an organisation called the General Bathymetric Chart of the Oceans (GEBCO), to do a complete mapping of the ocean bed by 2030. 

In June, the Seabed 2030 project reported that 26.1 per cent of the sea floor has been mapped. The exercise has revealed interesting underwater features, including tens of thousands of coral mounds with immense biodiversity. 

Studying the seabed between Costa Rica and Chile onboard a vessel called Falkor, owned by a company called Schmidt Ocean Institute, the crew made a groundbreaking discovery of four underwater mountains, the tallest of which is 1.5 miles high. More than a hundred new marine species were discovered. 

“Magnificent new surprises,” is how the find has been described by the Executive Director of Schmidt Ocean Institute, Jyotika Virmani. 





Source link

Catalysing Indian IP creation 


In a highly competitive world, it is perhaps time for India to rethink its strategy and protect its innovations instead of sharing them. 

Over the past century, the country has gradually transitioned from a largely agrarian to an industry- and service-dominated economy that is among the largest in the world. To sustain rapid development in sectors such as manufacturing, infrastructure, and defence, India largely relied on imported technologies in its early years. 

Today, its top universities and research labs have scientific and intellectual manpower that can rival any in the highly developed countries. 

The worries over ‘brain-drain’ are also a thing of the past, as more and more graduates from top institutions like the IITs choose to stay and work in the country. There is also a significant increase in the number of start-ups, many of which are in the deep-tech sector. 

Yet, despite these advances, India has a tiny share of the overall intellectual property (IP) creation globally. According to data from Questel Orbit Intelligence, China accounts for a lot more IPs in deep-tech domains than the US or Europe, and certainly 100 times more than India. Given India’s creditable share of global research publications, its achievement in IP creation lags greatly.

In fact, many academics believe that patents and publishing cannot go hand-in-hand. However, they are unaware that one can protect an idea by filing a provisional patent (granted within five working days), and then publishing the information in journals with an acknowledgement of the provisional patent. The complete specification can be filed in due course.

Nevertheless, patent filing from India is steadily increasing. The number of patents filed by, and granted to academic institutions has increased from 6,800 and 1,900, respectively, during 2018-19 to 8,500 and 3,000 during 2019-20; 10,000 and 3,400 in 2020-21; and 14,200 and 3,600 during 2021-22. These figures touched a high of 23,600 and 4,700, respectively, during 2022-23.

Speedy grants

The Indian Patent Office has introduced several measures to speed up the grant process, including reduced fee for start-ups, educational institutions, individuals, and small businesses; and permission for filing expedited publications and examinations. Particularly noteworthy is the expedited examination process, with patents being processed within 2-3 months for educational institutions and startups. 

Moreover, academic institutions and other organisations are fostering a culture of innovation by rewarding inventors (researchers) with IP filings and grants with promotions and other annual rewards. IP generation also contributes significantly to the national rankings by the National Institutional Ranking Framework (NIRF) and international rankings such as the QS World University Rankings.

India’s ranking in the Global Innovation Index (GII), published by the World Intellectual Property Organization, has risen steadily from 81 in 2015 to 40 in 2023. The Indian Patent Office has granted an impressive one lakh patents during fiscal year 2023-24. IIT-Madras received more than 400 patents during this period.

Monetisation

Indian IP in the tech sector is largely in the mechanical engineering and chemistry domains, alongside emerging domains such as computers and communications, biomedical and polymer technologies. On the other hand, there is a distinct need to improve the innovation quotient in sectors such as electronics, physics, biotech and civil engineering.

This, in turn, calls for policy support. Equally important is the need for IP commercialisation and monetisation. Ideally, IPs from academic institutions could be licensed through home-grown start-ups and small businesses, to help cut costly tech imports.

That said, Indian industry must protect its processes and products. Innovative solutions that have evolved over time in the agriculture, manufacturing and construction sectors are potentially valuable IPs, which could benefit other emerging economies too. 

By fostering the right ecosystem for IP creation and monetisation, India looks set to capture a leading position in the foreseeable future.

(The writer is a professor, department of civil engineering, and dean for industrial consultancy and sponsored research at IIT-Madras)





Source link

How radio waves can power smart cities 


Earlier this year, a company called Betavolt made news by announcing a nuclear battery that can go on for 50 years before needing a recharge. Technology is marching further ahead. In the future, at least for small electronic devices, you wouldn’t ever need to charge the batteries — they would self-charge from (man-made) energy in the atmosphere. 

There is plenty of energy around us — only we don’t quite know how to tap it well. There is sunlight, which we are now harnessing. Then we have heat and vibrations — scientists are trying to figure out how to make use of them. However, the problem is that these energy sources are not available everywhere or always. 

Yet another ambient energy source is electromagnetic (EM) radiation. Now, EM is a broad spectrum, extending from radio waves at one end to gamma rays at the other, and with infra-red, visible light, ultraviolet and X-rays in between. While all these are energetic, scientists have trained their sights on radio waves — perhaps because they are now practically omnipresent, thanks to human activity. 

Radio frequency (RF) signals are generated by the millions of devices we now use all the time — including Wi-Fi routers, radio and television broadcasting stations, and mobile networks. You can capture these ambient RF signals and convert them into alternating current. The rest is then routine stuff — a rectifier converts the AC into direct current, which can go into a battery for use by devices such as wireless IoT sensors (especially those deployed in hard-to-reach areas), wearable electronics (including medical implants), and other appliances such as smart light switches and security systems in buildings. 

Effortless wireless

Dr Sumit Som, Director of Variable Energy Cyclotron Centre (VECC), a unit of the government’s Atomic Energy Department, in Kolkata, describes ‘radio frequency energy harvesting’ (RFEH) as a “promising technology for dynamic recharging of wireless devices”.

Dr Som explains that RFEH has “numerous distinct advantages” — it can work in any location with a strong RF signal, which is practically everywhere; it is not affected by lack of sunlight or weather conditions; and it can work just as fine in indoor spaces, without needing specialised transmitters. 

Since the batteries are dynamically recharged by RF, they can be small and, in turn, the electronic device can be smaller too. 

Above all, it is green — no carbon footprint. 

That said, it is important to know that RFEH is still a technology in the making, though there is little doubt it will be within reach in future. The fundamental challenge to be cracked is the ‘power conversion efficiency’. Researchers are stretching themselves to improve antenna and rectification efficiency. RFEH technology is getting better by the day but, by all accounts, it needs to improve more. Prof Manash Sarma of Gauhati University says in a scientific paper that researchers at his institute have developed a ‘transmission gate-based system’ that is “capable of generating output power at a low level of input with good conversion efficiency”. However, despite the “strong performance”, there is still “an opportunity for improvement”.

Connected devices

A look at various research papers shows that RFEH will prove to be a good technology for building smart cities, where there is a growing demand for connected devices and sensors. Powering IoT sensors is one area where RFEH will help big time. You can have millions of sensors without worrying about how to power them. A smart city would need sensors for monitoring air quality, ambient temperature and humidity, monitoring infrastructure (such as integrity of bridges and buildings), smart meters for utilities, regulating transport and parking, surveillance, waste management, and so on. 

In India, RF technologies spin out of the Department of Atomic Energy’s labs such as Bhabha Atomic Research Centre (BARC) and VECC. The latter uses RF in particle accelerators — namely to accelerate and control the motion of charged particles (such as protons). Thanks to its expertise with RF, VECC has developed quite a few societal applications for RF technologies. 

For example, RF can be used in drying agricultural produce. Here, RF systems generate electromagnetic waves that cause ‘dipole rotation’ in the water molecules in agricultural produce — the water molecules spin, producing heat, and vaporise. This method is said to be faster and more uniform than conventional air or sun drying. 

“RF technology is used in a variety of important fields and will always remain in high demand,” says Dr Som.





Source link