Amazon on Tuesday unveiled the latest generation of connected products, featuring enhanced artificial intelligence capabilities designed to make interactions with AI more frequent and natural.

Nearly 20 years after the launch of the Kindle e-reader, the Seattle-based online retail giant now offers a family of connected devices, from the Echo smart speaker to the Ring doorbell and Fire TV.

Amazon now aims to multiply their capabilities through AI, but wants to use it “without getting in the way,” said Panos Panay, head of devices and services, during a New York presentation.

The company had already made a major move into AI enhancements with the February launch of Alexa+, an upgraded version of the Alexa voice assistant.

Amazon’s ambition, like that of competitors Google, LG and Samsung, is to become the connected home nerve center.

But the sector has struggled to deliver on the promise of a fully connected home, with consumers forced to choose from competing ecosystems or left struggling with technology that fails to deliver on expectations.

“Alexa, what happened around the house today?” a user asks in a demonstration video. The smart assistant explains that the children walked the dog, a package was delivered and raccoons rummaged through the trash—using images captured by Ring or Blink cameras.

Has your dog run away? After the escape is reported on the Ring app, other Amazon doorbells in the neighborhood can detect if the animal passes by and alert you.

With the Kindle Scribe, readers can ask generative AI for a book summary to refresh their memory or ask questions about a character.

As for connected television, viewers can verbally request to see a scene from their favorite movie or receive a summary of a football game they missed.

Amazon believes in “ambient” AI, in Panay’s words, which “lives naturally in the products themselves.”

The generative AI revolution is playing out on both software and physical interfaces, with major tech players seeking to determine which product will prevail—smartphone, smart glasses, earbuds or speakers.

OpenAI is working on a new kind of device, while Meta is betting on glasses and Apple on earbuds.

© 2025 AFP

When Prime Minister Narendra Modi declared India’s “late entry” into the global semiconductor race, he pinned hopes on pioneers such as Vellayan Subbiah to create a chip innovation hub.

The chairman of CG Power, who oversees a newly commissioned semiconductor facility in western India, is seen as one of the early domestic champions of this strategic sector in the world’s fastest-growing major economy.

“There has been more alignment between the government, policymakers, and business than I’ve ever seen in my working history,” Subbiah, 56, told AFP.

“There’s an understanding of where India needs to go, and the importance of having our own manufacturing.”

As US President Donald Trump shakes global trade with tariffs and hard-nosed transactionalism, Modi has doubled down on self-reliance in critical technologies.

New Delhi, which flagged its push in 2021, has this year approved 10 semiconductor projects worth about $18 billion in total, including two 3-nanometer design plants, among the most advanced.

Commercial production is slated to begin by the end of the year, with the market forecast to jump from $38 billion in 2023 to nearly $100 billion by 2030.

Subbiah, whose CG Power is one of India’s leading conglomerates, predicts “over $100 billion, if not more”, will flow into the industry across the value chain in the next five to seven years.

He said “symbiotic” public-private partnerships were “very exciting”.

‘Ability to accelerate’

Chips are viewed as key to growth and a source of geopolitical clout.

India says it wants to build a “complete ecosystem”, and break the global supply chain dominance by a few regions.

The government has courted homegrown giants such as Tata, alongside foreign players like Micron, to push design, manufacturing and packaging in joint ventures.

CG Semi, a joint venture with CG Power, plans to invest nearly $900 million in two assembly and test plants, as well as to push its design company.

“We are looking to design chips, so that we can own the (intellectual property) too—which is very important for India,” said Subbiah, a civil engineer by training with an MBA from the University of Michigan.

Still, critics say India is decades late starting, and remains far behind chip leaders in Taiwan, the Netherlands, Japan and China.

“First we have to recognize there is a gap,” Subbiah said, noting Taiwan’s TSMC has a 35-year head start.

But he insists India’s scale and talent pool—the world’s most populous nation with 1.4 billion people—gives it “a significant ability to accelerate” production.

‘More complicated’

Modi this month said that “20% of the global talent in semiconductor design comes from India”.

But wooing talent who sought opportunities abroad back to India remains a challenge, even after Trump’s restrictions on the H-1B skilled worker visa program, heavily used by Indians.

India, the world’s fifth-largest economy, still struggles with bureaucratic inertia and a lack of cutting-edge opportunities.

Subbiah acknowledged that his own venture employs about 75 expatriates.

“That’s not the way we want to grow. We want to grow with Indians,” he said, calling for policies to lure back overseas talent. “How do we bring these people back?”

But the path is tougher than in 2021, when New Delhi first pushed for chip self-sufficiency.

While India has secured semiconductor and AI investment pledges from partners such as Japan—which pledged $68 billion in August—Trump is expected to be less willing than past US leaders to back ventures that build Indian capacity.

“The geopolitical situation overall has become more complicated,” Subbiah said.

Yet he remains upbeat for the long run.

“There are only going to be two really low-cost ecosystems in the world: one is China, and the other is going to be India,” he said.

“You’re going to see the center of gravity move towards these ecosystems, if you start thinking about a 25-30 year vision”.

© 2025 AFP

Covering half of North America, the U.S. electric grid functions somewhat like a vast, complex organism. Researchers at the Department of Energy’s Oak Ridge National Laboratory have developed a new simulation platform for understanding and predicting the behavior of this modern grid. Using a combination of mathematical tools, automation and analysis, the approach provides highly accurate results with less computing time at a lower cost, increasing the reliability of electricity.

Simulation uses mathematical approaches to reproduce the dynamics of a real-world system. This allows utilities and planners to analyze grid management methods without any risk to safety, equipment or electrical service. ORNL researchers refined a cutting-edge grid modeling approach called Electromagnetic Transient simulation (EMT), which is especially effective for analyzing the split-second reactions of modern power electronics. This capability helps operators prevent cascading blackouts and unsafe operating conditions in modern electric grids brimming with power electronics.

“We are trying to understand electronics and systems in a way that mimics their real behavior with higher fidelity,” said ORNL researcher Phani Marthi. “The challenge today is that high-fidelity EMT simulation is extremely time-consuming to simulate large-scale modern power grid systems.”

The ORNL simulation approach is tackling those challenges, as Marthi and his co-authors explained in a paper that was presented in the best paper session at the July general meeting of the IEEE Power and Energy Society.

Representing the next phase of ORNL’s national leadership in EMT simulation, the ORNL tool is called RE-INTEGRATE for its enhanced speed and accuracy at simulating large-scale power systems that integrate many power electronics.

In the past, the grid relied on the natural momentum of huge rotating mechanical machines and power flowing in a single direction along established paths, like a locomotive on a track. But today, power electronics make the grid respond more like a sports car, with rapid electronic adjustments instead of built-in momentum. Unfortunately, today’s grid is not fully ready for that speed. RE-INTEGRATE helps utilities map the best route for the grid of the future.

Power electronics accommodate generating and moving electricity in different ways. They can also enable both alternating and direct current in long-distance power transmission. This could expand the capacity of the U.S. grid to support a growing population and economy, including new industries such as data centers for AI and cryptocurrency.

Unlike existing EMT models, RE-INTEGRATE is intended as an open-source platform that incorporates features such as numerical simulation techniques, automation and intelligence based on neural networks that function more like the brain for faster computation. These features offer unique advantages over existing tools in analyzing modern grids.

Eventually, the tool will be able to replicate faults—disruptions in the power grid caused by equipment failure, short circuits, or other technical issues—like the one that wiped out power to much of Spain and Portugal in April. “Analysis with the RE-INTEGRATE tool can give us new insights into how to consistently prevent or stop cascading blackouts and brownouts,” Marthi said.

One of the fundamental building blocks of RE-INTEGRATE is differential algebraic equation solvers. These algorithms reduce the degree of manual processing required for an immense volume of data. As a proof of concept, ORNL researchers validated the effectiveness of these solvers on simple power electronics circuits.

The long-term goal is honing the software to simulate all possible circumstances that could arise from fast-acting power electronics systems interacting with grid components in a large-scale power grid, equivalent to the grid of the eastern United States.

This will broaden the accuracy benefits of EMT while enabling greater understanding of how the parts of the broader grid affect each other across service areas and regions.

“Beyond accelerating the EMT simulation, the next major challenge lies in managing and sifting through the huge volumes of data generated by EMT simulations,” Marthi said. ORNL researchers are already developing advanced analysis techniques, including the use of specialized neural networks, so that the RE-INTEGRATE tool can enhance power system operations and support informed decision-making. “We want to create an entire EMT ecosystem with RE-INTEGRATE as the backbone, including all these capabilities so utilities use it more often and with more confidence.”

Researchers who contributed to the development of automation and solvers for RE-INTEGRATE include ORNL researchers Jongchan Choi and Suman Debnath with support from student Soumyajit Gangopadhyay and intern Kuan-Chieh Hsu.

RE-INTEGRATE advances will be presented during an EMT simulation workshop at ORNL, co-hosted by the North American Electric Reliability Corporation Oct. 7-9 in Knoxville, Tennessee.