MIT has unveiled a new method for training robots that scales up data in a way similar to large language models (LLMs), marking a shift from the narrow, task-focused data sets traditionally used in robotics. Imitation learning, where robots learn by observing humans, often struggles with new variables like lighting changes or unexpected obstacles. By adopting a vast data approach similar to that used in models like GPT-4, MIT’s researchers aim to help robots adapt more flexibly in varied environments.
The team developed a new architecture called Heterogeneous Pretrained Transformers (HPT), which combines information from multiple sensors and diverse settings to build robust training models. Larger transformers yielded improved outcomes, aligning with trends seen in LLMs, as HPT integrates data from multiple sources for more adaptable robotic responses.
Ultimately, researchers aspire to create a universal ‘robot brain’ that can be downloaded and used immediately without extra training. While still in early stages, the project has support from Toyota Research Institute, which recently partnered with Boston Dynamics to integrate learning research with advanced robotic hardware.
At TechCrunch Disrupt 2024, data management leaders advised AI-driven businesses to focus on incremental, practical applications rather than expansive, large-scale projects. Chet Kapoor, CEO of DataStax, stressed that AI’s effectiveness relies heavily on having robust, unstructured data at scale, but warned companies against rushing into overly ambitious initiatives. The discussion featured insights from Kapoor, Vanessa Larco of NEA, and Fivetran’s CEO George Fraser, all of whom advocated a targeted approach to data application in generative AI.
Rather than applying AI across all company functions immediately, Larco suggested that firms begin with well-defined objectives. Identifying relevant data is key, she said, and applying it selectively can avoid the pitfalls of costly errors. Companies looking to capitalise on AI should ‘work backwards’, focusing first on the issue to be solved and gathering the specific data required, Larco added.
Fraser underscored the importance of addressing current needs before planning for broader scaling. Many innovation costs, he pointed out, stem from projects that fail rather than those that succeed. His advice: ‘Only solve the problems you have today’.
Kapoor likened today’s generative AI era to the early days of mobile apps, emphasising that most AI projects are currently in exploratory stages. He believes next year will see transformational AI applications begin to shift company trajectories.
The Federal Energy Regulatory Commission (FERC) is examining the rapid growth of energy-intensive data centers being built next to US power plants. Known as co-location, this trend is driven by the tech sector’s need for large amounts of power for AI and other data-heavy operations. Co-locating data centers near power plants offers companies quicker access to electricity, bypassing the longer process of connecting to the broader grid.
However, regulators and industry experts are concerned about the impact on costs and reliability for other electricity consumers. If data centers use power plants that typically supply the public grid, there are questions about how such facilities will handle power disruptions and whether they will lean on the grid as backup. This could mean higher electricity bills for consumers who fund grid infrastructure, a point raised by FERC Commissioner Mark Christie.
The regulatory scrutiny comes as companies like Amazon and Google look to establish co-located data centers to meet growing energy needs. A recent arrangement in Pennsylvania, where Amazon bought a data center linked to a nuclear plant, has stirred debate among electric utilities over infrastructure costs and reliability. FERC’s review could lead to new guidelines clarifying financial responsibilities and operational rules for these partnerships.
e& and ZTE have forged a significant strategic collaboration by signing a Memorandum of Understanding (MOU) to foster innovation in the telecommunications sector amid current geopolitical challenges affecting technology adoption. The partnership is centred on developing energy-efficient and sustainable solutions for network infrastructure, addressing the growing global emphasis on environmental responsibility.
Additionally, they will design and develop advanced 5G and 5G-Advanced (5G-A) solutions while exploring innovative use cases for private networks to enhance connectivity and operational efficiency. Through shared insights and knowledge exchange, both organisations will promote leadership and collaborative learning, positioning themselves as responsible leaders in the industry while striving for net-zero emissions and inspiring further innovation.
Why does it matter?
The partnership emphasises the importance of creating opportunities for joint initiatives that enhance capabilities. By fostering a shared learning environment, both organisations can learn from each other’s successes and challenges, allowing for effective navigation of the rapidly evolving telecommunications landscape. Their joint efforts aim to bolster technological development and pave the way for a more connected and sustainable future.
The Indonesia Investment Authority (INA) and Singapore-based Granite Asia announced a joint plan to invest up to $1.2 billion in Indonesia’s technology sector and related businesses. According to a joint statement, the initiative will involve both equity and hybrid capital, though specific financial contributions from each firm and the timeline for investments remain undisclosed. Targeted companies were also not named.
INA, Indonesia‘s sovereign wealth fund, was established in 2020 to attract both domestic and international investors to the country. Last week, the fund announced that its toll road platform, with backing from Dutch pension fund APG and the Abu Dhabi Investment Authority, is funding parts of Indonesia’s Trans Sumatra Toll Road.
Granite Asia, formerly part of GGV Capital, currently manages assets totalling $5 billion. The collaboration is expected to support the growth of Indonesia’s tech landscape and strengthen economic ties within Southeast Asia.
Nvidia CEO Jensen Huang has urged South Korea’s SK Hynix to speed up the delivery of its next-generation HBM4 memory chips by six months, according to SK Group Chairman Chey Tae-won. Initially scheduled for the latter half of 2025, the HBM4 chips are in high demand as Nvidia’s GPUs require them for advancing AI technology. Nvidia, which holds a dominant share of the AI chip market, relies on SK Hynix’s high-bandwidth memory to support AI processing.
Facing growing competition from Samsung and Micron, SK Hynix is working to deliver its latest HBM3E chips this year, with plans to release improved 16-layer versions early next year. Samsung has also announced progress on a new supply deal and aims to roll out its HBM4 products by the second half of 2024.
Shares of SK Hynix surged 5.1% on the news, reflecting strong investor confidence in its strategic response to the booming demand for advanced memory technology.
United States energy regulators have rejected an amended plan for an Amazon data centre to be powered directly from Talen Energy’s Susquehanna nuclear plant in Pennsylvania. The Federal Energy Regulatory Commission (FERC) cited potential risks to both consumer costs and grid reliability, concluding that diverting power from the regional grid to Amazon’s facility could raise public energy bills and create supply challenges.
The proposal came as Big Tech companies like Amazon seek rapid ways to meet growing energy demands for data centres, particularly those needed to expand AI technologies. Co-locating data centres with power plants has emerged as an appealing solution, yet FERC Commissioner Mark Christie warned that this arrangement could bring complex repercussions, including significant impacts on reliability and costs.
FERC Chairman Willie Phillips, however, dissented, arguing that blocking the project could hinder US leadership in AI and harm national security. The decision leaves questions about funding and infrastructure upgrades necessary to ensure reliable supply to such high-demand centres.
Meta Platforms has signed an agreement to purchase the full output of a new solar power plant from French utility giant Engie. The Sypert solar plant, expected to generate 260 megawatts of clean energy, is scheduled to go live in late 2025. This partnership aligns with Meta’s ongoing commitment to meet the energy demands of its expanding data centre operations with sustainable power sources.
The Sypert plant will add to Engie’s growing renewable energy portfolio, which currently includes about 8 gigawatts of solar, wind, and battery storage projects across North America. Earlier this month, Engie also secured a solar power agreement with Google for its largest US solar project, reinforcing the company’s role as a major clean energy supplier for tech firms.
Driven by technologies like AI, the demand for data centre power in the US is predicted to triple by 2030, according to Goldman Sachs. The Biden administration has called on tech companies to invest in green energy to support this growth, and partnerships like Meta and Engie’s reflect this broader push toward a more sustainable digital economy.
Foxconn subsidiary Shunsin has submitted a request for an $80 million investment permit to establish an integrated circuit manufacturing plant in Bac Giang province, northern Vietnam. This development is detailed in a document from Vietnam’s environment ministry and represents a significant expansion of Foxconn’s operations in the region. The proposed facility aims to produce and process electronic components, particularly integrated circuit boards and is expected to commence full-scale operations by December 2026, with an annual production capacity of 4.5 million units.
This move aligns with Foxconn’s ongoing strategy to diversify its manufacturing base outside of China, especially in Southeast Asia, where it has already made substantial investments. The products manufactured at the new Shunsin plant will be designated for export, targeting major markets including the US, EU, and Japan. This export focus underscores Foxconn’s commitment to meeting the growing global demand for advanced electronic components.
Foxconn, officially known as Hon Hai Precision Industry, is recognised as the world’s largest contract electronics manufacturer. Since entering Vietnam in the early 2000s, the company has invested over $3.2 billion in various operations across the country. Its manufacturing footprint is primarily concentrated in northern provinces like Bac Ninh and Bac Giang, which have become key hubs for electronics production. In July, Foxconn also received a license to invest $383 million in a factory dedicated to printed circuit boards, further enhancing its capabilities in the region.
The UAE’s energy giant ADNOC is pioneering the use of highly autonomous agentic AI in the energy sector through a partnership with G42, Microsoft, and AIQ, as announced by CEO Sultan Al Jaber at an industry event in Abu Dhabi. This move is part of a broader UAE strategy to reduce reliance on oil, with support from G42, which secured a $1.5 billion investment from Microsoft to fuel the nation’s tech industry diversification.
Agentic AI, viewed as the future of artificial intelligence, allows systems to operate independently and make proactive decisions. According to Jaber, this advanced AI will significantly enhance operations by analysing vast amounts of data, reducing seismic survey times from months to days, and improving production forecasts by up to 90%.
The UAE’s government is investing billions in AI, including regional language-specific chatbots, positioning the Gulf state to remain economically influential as global demand for oil wanes.
