Page 22«..10..21222324..3040..»

SVP Shlomit Weiss leaving Intel amid engineering reorganization – CTech

Just three years after her return to the company, Shlomit Weiss, the most senior Israeli figure at Intel, is departing the chip giant. Concurrently, Intel is undergoing an organizational change, integrating the company's Design Engineering Group, where Weiss served as co-general manager, into the company's business groups. Intel confirmed these details but asserted that Weiss' retirement was unrelated to the reorganization process.

Weiss boasts over 30 years of experience in the chip industry, primarily acquired at Intel, supplemented by approximately four years in a senior role at Mellanox. In her current capacity, Weiss oversees the development and design of chips in the realm of edge computing, alongside co-general manager Sunil Shenoy. Her retirement signifies a significant loss of knowledge and expertise for Intel, potentially impacting the status of Intel's R&D center in Israel within the company.

"Ms. Weiss has chosen to retire from Intel after an illustrious career," the company stated. "She will continue to serve as a consultant until the end of the summer. She is an exceptional leader who has made substantial contributions to both Intel and the industry."

Weiss has held many executive roles at Intel during her 28-year stint at the company. Over the years, she was the laureate of Intel's top award the Intel Achievement Award - for developing Intel's Dual-Core processor architecture in Israel, and led the development of the Sky Lake processor, ultimately named as "Intel's best-ever processor" by its then-CEO. She worked at Intel for 28 years before joining Mellanox in 2017. It was then acquired by Nvidia where she served as its Senior VP for Silicon Engineering before stepping down in July 2021.n

Link:

SVP Shlomit Weiss leaving Intel amid engineering reorganization - CTech

Read More..

DeepMind CEO Says Google Will Spend More Than $100 Billion on AI – Bloomberg

The chief of Googles AI business said that over time the company will spend more than $100 billion developing artificial intelligence technology another sign of the investing arms race that has gripped Silicon Valley.

Google DeepMind Chief Executive Officer Demis Hassabis was asked at a TED conference in Vancouver on Monday about a potential $100 billion supercomputer dubbed Stargate, being planned by Microsoft Corp. and OpenAI, according to a report in the Information last month.

Original post:
DeepMind CEO Says Google Will Spend More Than $100 Billion on AI - Bloomberg

Read More..

DeepMind CEO says Google to spend more than $100B on AGI despite hype – Cointelegraph

Googles not backing down from the challenge posed by Microsoft when it comes to the artificial intelligence sector. At least not according to the CEO of Google DeepMind, Demis Hassabis.

Speaking at a TED conference in Canada, Hassabis recently went on the record saying that he expected Google to spend more than $100 billion on the development of artificial general intelligence (AGI) over time. His comments reportedly came in response to a question concerning Microsofts recent Stargate announcement.

Microsoft and OpenAI are reportedly in discussions to build a $100 billion supercomputer project for the purpose of training AI systems. According to the Intercept, a person wishing to remain anonymous, who has had direct conversations with OpenAI CEO Sam Altman and seen the initial cost estimates on the project, says its currently being discussed under the codename Stargate.

To put the proposed costs into perspective, the worlds most powerful supercomputer, the U.S.-based Frontier system, cost approximately $600 million to build.

According to the report, Stargate wouldnt be a single system similar to Frontier. It will instead spread out a series of computers across the U.S. in five phases with the last phase being the penultimate Stargate system.

Hassabis comments dont hint at exactly how Google might respond, but seemingly confirm the notion that the company is aware of Microsoft's endeavors and plans on investing just as much, if not more.

Ultimately, the stakes are simple. Both companies are vying to become the first organization to develop artificial general intelligence (AGI). Todays AI systems are constrained by their training methods and data and, as such, fall well short of human-level intelligence across myriad benchmarks.

AGI is a nebulous term for an AI system theoretically capable of doing anything an average adult human could do, given the right resources. An AGI system with access to a line of credit or a cryptocurrency wallet and the internet, for example, should be able to start and run its own business.

Related: DeepMind co-founder says AI will be able to invent, market, run businesses by 2029

The main challenge to being the first company to develop AGI is that theres no scientific consensus on exactly what an AGI is or how one could be created.

Even among the worlds most famous AI scientists Metas Yann LeCun, Googles Demis Hassabis, etc. there exists no small amount of disagreement as to whether AGI can even be achieved using the current brute force method of increasing datasets and training parameters, or if it can be achieved at all.

In a Financial Times article published in March, Hassabis made a negative comparison to the current AI/AGI hype cycle and the scams its attracted to the cryptocurrency market. Despite the hype, both AI and crypto have exploded their respective financial spaces in the first four months of 2024.

Where Bitcoin, the worlds most popular cryptocurrency sat at about $30,395 per coin in April of 2023, its now over $60,000 as of the time of this articles publishing, having only recently retreated from an all-time-high about $73K.

Meanwhile, the current AI industry leader, Microsoft, has seen its stock go from $286 a share to around $416 in the same time period.

Visit link:
DeepMind CEO says Google to spend more than $100B on AGI despite hype - Cointelegraph

Read More..

VueReal Strengthens Global Partner Ecosystem with Toray Engineering Alliance, Accelerating Global Adoption of … – AccessWire

Partnerships with Industry Leaders Create a Blueprint to Empower Customers to Manufacture MicroLED Displays and Micro Semi Products Efficiently

WATERLOO, ON / ACCESSWIRE / April 17, 2024 / VueReal, a pioneering force in MicroSolid Printing, today announced a strategic partnership with Toray Engineering Co., Ltd., a renowned semiconductor tool manufacturer. This collaboration reflects another significant step in VueReal's mission to bolster its global partner ecosystem, which is dedicated to propelling the widespread adoption of cutting-edge microLED displays and micro semiconductor products.

VueReal's collaboration with Toray Engineering extends across multiple customized and optimized tools and steps. The integration of Toray affiliate TASMIT's cutting-edge PL inspection tool 3000SR-300 with VueReal's turnkey MicroSolid Printing platform exemplifies VueReal's core objective to empower businesses of all sizes to design and deliver cost-effective, premium-quality, and environmentally sustainable microLED displays and micro semiconductor devices on a large scale.

Toray affiliate TASMIT's cutting-edge PL inspection tool 3000SR-300

"Partnering with leaders such as Toray Engineering is a key part of VueReal's strategy to empower companies of all sizes to innovate efficiently and sustainably with micro technologies," said Dr. Reza Chaji, Founder and CEO of VueReal. "VueReal's turnkey platform meets these needs, offering a comprehensive solution for high-quality, mass-producible, and intricate displays and micro-device structures. Integrating Toray Engineering's cutting-edge tools further solidifies our commitment to providing the essential blueprint to establish MicroSolid Printing as the gold standard."

Establishing a $30 Billion MicroLED Display Market by 2030

Leading the charge to establish a $30 billion microLED display market by 2030, VueReal's MicroSolid Printing platform tackles the primary challenge in the mass adoption of microLED displays-the efficient transfer of LEDs from wafer to backplane. VueReal's patented process transforms millions of micrometer-sized LEDs with unmatched efficiency, scalability, and high yield, paving the way for a revolution in the microLED display industry.

VueReal continuously assesses semiconductor tools worldwide, identifying key components to facilitate customers' cost-effective design and development of next-generation microLED displays, lighting solutions, and micro-semiconductor products across various industries. The company leverages its Canadian manufacturing facility to support innovative product development for customers, from feasibility studies to proof-of-concept development, product design, pre-production, and early-stage production. OEMs, display and semiconductor fabs, hardware companies, and other businesses can also license VueReal's turnkey platform to independently develop microdevices that far exceed the capabilities of today's display, lighting, and sensory solutions in a cost-competitive manner.

"Toray Engineering is excited about our strategic collaboration with VueReal," said Toyoharu Terada, Managing Director at Toray Engineering. "Our joint commitment to advancing the semiconductor industry through streamlined production of microLED displays is evident in this partnership. Integrating the PL inspection tool 3000SR-300 into VueReal's platform underscores our dedication to delivering sophisticated tools aligned with our shared vision for innovation and progress."

The PL inspection tool 3000SR-300 seamlessly integrates with VueReal's MicroSolid Printing platform to further reduce surface inspection device review time and contribute to greater inspection efficiency. This system acquires images of defects with a high level of magnification and categorizes defect types.

About Toray Engineering Co., Ltd.

Toray Engineering is a global leader in innovative engineering technologies. Established in 1960, Toray Engineering has been designing and providing plant construction and FA equipment, as well as state-of-the-art manufacturing facilities and equipment such as FPD/semiconductor production equipment. Under its business brand "TRENG", Toray Engineering contributes to society by creating new value and realizing solutions to create a sustainable society. https://www.toray-eng.com/

About VueReal

VueReal, a pioneer in MicroSolid Printing, is revolutionizing the micro semiconductor device industry with its eco-friendly micro-pixel manufacturing process. The platform enables the efficient transfer of MicroLEDs and other micro semiconductor devices, offering unmatched efficiency, reliability, and scalability. VueReal's patented method ensures high yield, throughput, and industry-leading defect rates, driving the microLED display market to a projected value of $30 billion by 2029. In addition to manufacturing microLED displays and lighting products for auto in Canada, VueReal licenses its MicroSolid Printing Blueprint to global OEMs, display fabs, and hardware manufacturers. For more information, please visit http://www.vuereal.com.

For media inquiries, contact:

Katherine Waite Trier & Company for VueReal [emailprotected] Phone: 617-599-9798

SOURCE: VueReal

See the rest here:

VueReal Strengthens Global Partner Ecosystem with Toray Engineering Alliance, Accelerating Global Adoption of ... - AccessWire

Read More..

Boston Dynamics’ new robot Atlas is more agile and all-electric – Interesting Engineering

Just a day after Boston Dynamics announced the retirement of their hydraulic Atlas robot, the robotics company revealed its replacement: A fully electric Atlas robot designed for real-world applications.

This newest robot builds upon decades of their research. According to the company, it will be more than just an impressive R&D project.

A decade ago, we were one of the only companies putting real R&D effort into humanoid robots, said Boston Dynamics in a statement. Now the landscape in the robotics industry is very different. Our customers have seen success with Spot and Stretch, and they are eager to tackle the next challenge with Atlas.

More here:

Boston Dynamics' new robot Atlas is more agile and all-electric - Interesting Engineering

Read More..

Google will spend more than $100 billion on AI, exec says – Quartz

After comparing the billions of dollars going into AI development to crypto hype, Googles AI chief executive said Monday the company will spend over $100 billion over time to develop AI technology.

ChatGPT requires 15 times more energy than a traditional web search, says Arm exec

Demis Hassabis, chief executive of Google DeepMind, talked about the tech giants investment into AI during a TED conference in Vancouver on Monday, where he was asked about OpenAIs and Microsofts reported plans for a U.S.-based data center referred to as Stargate, Bloomberg reported. The data center would house a supercomputer made up of millions of AI chips, and could cost up to $100 billion, The Information reported, citing unnamed sources.

We dont talk about our specific numbers, but I think were investing more than that over time, Hassabis said in response to the question about Stargate. He didnt offer further details on Googles spending plans, Bloomberg reported. Hassabis, who co-founded AI startup DeepMind in 2010 before it was acquired by Google in 2014, reportedly added that Google parent Alphabet has better computing power than its rivals, including Microsoft.

Thats one of the reasons we teamed up with Google back in 2014, is we knew that in order to get to AGI we would need a lot of compute, Hassabis said. Artificial general intelligence (AGI) is the point at which AI reaches human-level knowledge across a range of tasks. Google had and still has the most computers, Hassabis said.

In March, Hassabis told the Financial Times that the billions of dollars being poured into AI is reminiscent of crypto hype, and is taking attention away from the phenomenal science and research behind its development.

The investment into AI brings with it a whole attendant bunch of hype and maybe some grifting, he said, comparing it to crypto and similar areas, adding that the sentiment has now spilled over into AI, which I think is a bit unfortunate.

However, Hassabis said he thinks the industry is only scratching the surface of what is possible. Were at the beginning, maybe, of a new golden era of scientific discovery, a new Renaissance, he said.

More here:
Google will spend more than $100 billion on AI, exec says - Quartz

Read More..

Michigan manufacturers work to expand engineering training to high schools – Gladwin County Record & Beaverton Clarion

LANSINGThe Michigan Manufacturers Association plans to expand its Partnership Response in Manufacturing Education program, or PRIME, into more high schools throughout the state.This is an initiative that was created to bring engineering to schools. (Manufacturers) were focusing a lot on the colleges and realizing that there wasnt a lot of preparation in the K-12 arena, said Bill Rayl, the

Continue reading here:

Michigan manufacturers work to expand engineering training to high schools - Gladwin County Record & Beaverton Clarion

Read More..

DeepMind Head: Google AI Spending Could Exceed $100 Billion – PYMNTS.com

Googles top AI executive says the companys spending on the technology will surpass $100 billion.

While speaking Monday (April 15) at a TED Conference in Vancouver, DeepMind CEO Demis Hassabis was asked about recent reports of Microsoft and OpenAIs planned artificial intelligence (AI) supercomputer known as Stargate,said to cost $100 billion.

We dont talk about our specific numbers, but I think were investing more than that over time, said Hassabis, whose comments were reported by Bloomberg News.

Hassabis, who co-founded DeepMind in 2010 before it was bought by Google four years later, did not offer further details on the potential AI investment, the report said. He also told the audience Googles computer power surpasses that of competitors like Microsoft.

Thats one of the reasons we teamed up with Google back in 2014, is we knew that in order to get to AGI we would need a lot of compute, he said, referring to artificial general intelligence, or AI that surpasses the intelligence of humans.

Thats whats transpired, he said. And Google had and still has the most computers.

Hassabis added that the massive interest kicked off by OpenAIs ChatGPT AI model demonstrated the public was ready for the technology, even if AI systems are still prone to errors.

As PYMNTS wrote earlier this month, the Stargate project spotlights the increasing role of AI in fueling innovation and determining the future of commerce. Experts believe that as tech giants invest heavily in AI research and infrastructure, the creation of sophisticated AI systems could revolutionize areas like personalized marketing and supply chain optimization.

It is important to consider the potential impact on jobs and the workforce, Jiahao Sun, founder and CEO at FLock.io, a platform for decentralized AI models, said in an interview with PYMNTS.

As AI becomes more capable in multimodal and integrated into commerce, it may automate industries that currently cannot easily be transferred into a chatbot interface, such as manufacturing, healthcare, sports coaching, etc.

Microsoft and OpenAIs $100 billion project could make AI chips more scarce, leading to more price spikes and leaving more businesses and governments behind due to limited access to hardware, CEO and co-founder of AI company NeuReality Moshe Tanachtold PYMNTS, while adding that projects like Stargate will drive commerce forward in the short term.

The installed hardware will fuel more AI projects, features and use cases, leading Microsoft to offer it at consumable prices, driving innovation on the consumer side with secondary use cases built on this accessible AI technology, Tanach said.

Continue reading here:
DeepMind Head: Google AI Spending Could Exceed $100 Billion - PYMNTS.com

Read More..

NetZero Metals Awards Engineering Contracts and Advances Engineering Studies on Two Processing Facilities – PR Newswire

Highlights

TORONTO, April 17, 2024 /PRNewswire/ - NetZero Metals Inc. ("NetZero Metals" or the "Company"), a wholly-owned subsidiary of Canada Nickel Company Inc. ("Canada Nickel") (TSXV: CNC) (OTCQX: CNIKF) today provided an update on the progress of engineering studies for two processing facilities a nickel processing facility and a stainless steel and alloy production facility in the Timmins Nickel District.

Key engineering contracts for the facilities have been awarded to SMS group ("SMS"), Metso and Ausenco Engineering Canada ULC ("Ausenco").

NetZero Metals is advancing design work to develop two processing facilities to fill a critical link in the development of North American critical minerals supply chains and the province's electric vehicle strategy. Each production facility is expected to utilize Canada Nickel's carbon storage capacity at its Crawford Nickel Project to deliver net zero carbon nickel and stainless steel and alloy production through capture of the CO2emissions from the downstream operation. The studies are progressing well and completion of feasibility studies remains targeted for year-end.

Mark Selby, CEO of Canada Nickel, said, "We are very pleased to work with global engineering leaders like SMS, Metso, and Ausenco to advance our NetZero Metals processing plants. These processing facilities will position the Timmins Nickel District and Canada at the forefront of the global transition to greener energy and materials. We are proud to lead the way in the development of new capacity to meet the growing demand for the local supply of critical minerals, and most importantly, zero carbon, environmentally responsible production in North America."

SMS group, an industry leader in the supply of steel making equipment and plant design, has been awarded the initial scope for the NetZero Metals stainless steel and alloy plant, which will treat the Crawford iron-chrome-nickel concentrate and other feeds to produce a blend of stainless steel and alloys. The target start up date of the steel plant is year-end 2027, aligned with expected first production from Crawford.

Andy Gribben, VP of Metallurgy Sales at SMS group, said "We are thrilled to have been chosen by NetZero Metals for such a pivotal role in advancing sustainable nickel and stainless steel production. At SMS group, we have made it our mission to create a carbon-neutral and sustainable metals industry. Our metallurgical know-how and engineering skills combined with our digital expertise and plant technology enables us to be the leading partner on this project. We look forward to contributing our expertise to this transformative, environmentally responsible project."

Metso, an industry leader for the supply of pyrometallurgical process equipment, has been awarded the key portions of the initial nickel plant scope, which will process the Crawford nickel concentrate as well as third party feeds to produce a blend of nickel and nickel-cobalt products suitable for the steel and battery markets. The Company is targeting an early start up of the nickel plant in the first part of 2027 by leveraging an existing brownfield site for which positive discussions are already underway.

Ausenco, a long term partner of Canada Nickel Company, is performing the overall facility engineering and packaging of staged engineering studies.

About SMS

SMS group is renowned worldwide for its future-oriented technologies and outstanding service for the metals industry. The company applies its 150 years of experience and its digital know-how to provide the industry continuously with innovative products and processes even beyond its core business and generates worldwide sales of around 3.1 billion euros. SMS is the right partner for challenging projects, and supports its customers throughout the lifecycle of their plants, enabling profitable and resource-efficient value creation chains. Paving the way for a carbon-neutral and sustainable metals industry is the company's stated goal. For more information on SMS group and their portfolio, visit their website atwww.sms-group.us.

About Metso

Metso is a frontrunner in providing sustainable technologies, end-to-end solutions and services for the aggregates, minerals processing and metals refining industries globally. By helping its customers increase their productivity, improve their energy and water efficiency and environmental performance with process and product expertise, Metso is the partner for positive change. Metso's Fluid Bed business has been recently engaged to carry out fluid bed testwork and preliminary flowsheet development at its research and development center in Frankfurt, Germany. Metso has been developing world class fluid bed roasting applications for the non-ferrous industry since the 1950s.

About Ausenco

Ausenco is a global diversified engineering, construction and project management company providing consulting, project delivery and asset management solutions to the resources, energy and infrastructure sectors. Ausenco's experience in mining projects ranges from conceptual, pre-feasibility and feasibility studies for new project developments to project execution. Ausenco has been supporting Canada Nickel since 2020 when Ausenco was first engaged to lead the Preliminary Economic Assessment ("PEA") for the Crawford Nickel Sulfide Project. After successfully completing the study, Ausenco continued to support Canada Nickel through the Feasibility Study, which was released in the fourth quarter of 2023, by supporting the metallurgical pilot plant program, the process plant and building design and the site water management design, all while coordinating project consultants. Ausenco is now leading Front End Engineering and Design activities at Crawford as it advances towards a construction decision in 2025.

About the Nickel Processing Facility

Over three phases, the nickel processing plant is expected to reach a capacity of more than 80,000 tonnes of nickel annually, which is expected to make it the largest nickel processing facility in North America. Subject to permitting approvals, the plant plans to begin production by the start of 2027 utilizing third party feeds. The facility is expected to be further expanded with the startup and expansion of Canada Nickel's Crawford project. This plant is expected to utilize proven, low environmental footprint technology to produce high-quality nickel products, catering to the needs of both the stainless-steel/superalloy and the electric vehicle markets.

About the Stainless-Steel and Alloy Processing Facility

In addition to the nickel processing plant, a stainless steel and alloy production facility is expected to be established to process the iron-chrome-nickel concentrate from the Crawford Nickel project and other feeds to be transformed into more than 1 million tonnes of alloy products, including more than 500,000 tonnes of 304-grade stainless-steel annually. The facility is expected to grow along with the expansion at Crawford Nickel project and would become the largest stainless-steel production facility in Canada. Production is planned to begin by year end 2027, aligning with the planned start-up of the Crawford mine, subject to permitting approvals. Discussions are currently underway with leading global ferroalloy and stainless producers to partner on this project.

Qualified Person and Data Verification

Arthur G. Stokreef, P.Eng (ON), Manager of Process Engineering & Geometallurgy and a "qualified person" as such term is defined by National Instrument 43-101, has reviewed and approved the technical information in this news release on behalf of Canada Nickel Company Inc.

About Canada Nickel Company

Canada Nickel Company Inc. is advancing the next generation of nickel-sulphide projects to deliver nickelrequired to feed the high growth electric vehicle and stainless steel markets. Canada Nickel Companyhas applied in multiple jurisdictions to trademark the terms NetZero NickelTM, NetZero CobaltTM, NetZero IronTMand is pursuing the development of processes to allow the production of net zero carbon nickel, cobalt, and ironproducts. Canada Nickel provides investors with leverage to nickel in low political risk jurisdictions.Canada Nickel is currently anchored by its 100% owned flagship Crawford Nickel-Cobalt Sulphide Project in theheartoftheprolific Timmins-Cochraneminingcamp.Formoreinformation,please visitwww.canadanickel.com.

For further information, please contact:Mark Selby, CEO Phone: 647-256-1954 Email: [emailprotected]

Cautionary Note and Statement Concerning Forward Looking Statements

This press release contains certain information that may constitute "forward-looking information" under applicable Canadian securities legislation. Forward looking information includes, but is not limited to, the timing, outcome and completion of study activities including preliminary economic assessments, feasibility studies and/or front end engineering and design, the outcome of procuring a brownfield site, the product blend and overall production capacity of NetZero Metals and steel and alloy plants, the timing of plant start-up and expansions, the viability of either or both of the downstream processing lines including the nickel and steel and alloy processing lines, the ability to achieve zero carbon production from the nickel and or steel and alloy plants, the outcomes of the construction decision, the timing of first production at the Crawford Project, the receipt of permits, the ability to sell marketable materials, strategic plans, including future exploration and development plans and results, and corporate and technical objectives. Forward-looking information is necessarily based upon several assumptions that, while considered reasonable, are subject to known and unknown risks, uncertainties, and other factors which may cause the actual results and future events to differ materially from those expressed or implied by such forward-looking information. Factors that could affect the outcome include, among others: future prices and the supply of metals, the future demand for metals, the results of drilling, inability to raise the money necessary to incur the expenditures required to retain and advance the property, environmental liabilities (known and unknown), general business, economic, competitive, political and social uncertainties, results of exploration programs, risks of the mining industry, delays in obtaining governmental approvals, and failure to obtain regulatory or shareholder approvals. There can be no assurance that such information will prove to be accurate, as actual results and future events could differ materially from those anticipated in such information. Accordingly, readers should not place undue reliance on forward-looking information. All forward-looking information contained in this press release is given as of the date hereof and is based upon the opinions and estimates of management and information available to management as at the date hereof. Canada Nickel disclaims any intention or obligation to update or revise any forward-looking information, whether because of new information, future events or otherwise, except as required by law. Neither TSX Venture Exchange nor its Regulation Services Provider accepts responsibility for the adequacy or accuracy of this release.

SOURCE Canada Nickel Company Inc.

View post:

NetZero Metals Awards Engineering Contracts and Advances Engineering Studies on Two Processing Facilities - PR Newswire

Read More..

Engineering Household Robots to Have a Little Common Sense – The Good Men Project

By Jennifer Chu | MIT News

From wiping up spills to serving up food, robots are being taught to carry out increasingly complicated household tasks. Many such home-bot trainees are learning through imitation; they are programmed to copy the motions that a human physically guides them through.

It turns out that robots are excellent mimics. But unless engineers also program them to adjust to every possible bump and nudge, robots dont necessarily know how to handle these situations, short of starting their task from the top.

Now MIT engineers are aiming to give robots a bit of common sense when faced with situations that push them off their trained path. Theyve developed a method that connects robot motion data with the common sense knowledge of large language models, or LLMs.

Their approach enables a robot to logically parse many given household task into subtasks, and to physically adjust to disruptions within a subtask so that the robot can move on without having to go back and start a task from scratch and without engineers having to explicitly program fixes for every possible failure along the way.

Imitation learning is a mainstream approach enabling household robots. But if a robot is blindly mimicking a humans motion trajectories, tiny errors can accumulate and eventually derail the rest of the execution, says Yanwei Wang, a graduate student in MITs Department of Electrical Engineering and Computer Science (EECS). With our method, a robot can self-correct execution errors and improve overall task success.

Wang and his colleagues detail their new approach in astudythey will present at the International Conference on Learning Representations (ICLR) in May. The studys co-authors include EECS graduate students Tsun-Hsuan Wang and Jiayuan Mao, Michael Hagenow, a postdoc in MITs Department of Aeronautics and Astronautics (AeroAstro), and Julie Shah, the H.N. Slater Professor in Aeronautics and Astronautics at MIT.

Language task

The researchers illustrate their new approach with a simple chore: scooping marbles from one bowl and pouring them into another. To accomplish this task, engineers would typically move a robot through the motions of scooping and pouring all in one fluid trajectory. They might do this multiple times, to give the robot a number of human demonstrations to mimic.

But the human demonstration is one long, continuous trajectory, Wang says.

The team realized that, while a human might demonstrate a single task in one go, that task depends on a sequence of subtasks, or trajectories. For instance, the robot has to first reach into a bowl before it can scoop, and it must scoop up marbles before moving to the empty bowl, and so forth. If a robot is pushed or nudged to make a mistake during any of these subtasks, its only recourse is to stop and start from the beginning, unless engineers were to explicitly label each subtask and program or collect new demonstrations for the robot to recover from the said failure, to enable a robot to self-correct in the moment.

That level of planning is very tedious, Wang says.

Instead, he and his colleagues found some of this work could be done automatically by LLMs. These deep learning models process immense libraries of text, which they use to establish connections between words, sentences, and paragraphs. Through these connections, an LLM can then generate new sentences based on what it has learned about the kind of word that is likely to follow the last.

For their part, the researchers found that in addition to sentences and paragraphs, an LLM can be prompted to produce a logical list of subtasks that would be involved in a given task. For instance, if queried to list the actions involved in scooping marbles from one bowl into another, an LLM might produce a sequence of verbs such as reach, scoop, transport, and pour.

LLMs have a way to tell you how to do each step of a task, in natural language. A humans continuous demonstration is the embodiment of those steps, in physical space, Wang says. And we wanted to connect the two, so that a robot would automatically know what stage it is in a task, and be able to replan and recover on its own.

Mapping marbles

For their new approach, the team developed an algorithm to automatically connect an LLMs natural language label for a particular subtask with a robots position in physical space or an image that encodes the robot state. Mapping a robots physical coordinates, or an image of the robot state, to a natural language label is known as grounding. The teams new algorithm is designed to learn a grounding classifier, meaning that it learns to automatically identify what semantic subtask a robot is in for example, reach versus scoop given its physical coordinates or an image view.

The grounding classifier facilitates this dialogue between what the robot is doing in the physical space and what the LLM knows about the subtasks, and the constraints you have to pay attention to within each subtask, Wang explains.

The team demonstrated the approach in experiments with a robotic arm that they trained on a marble-scooping task. Experimenters trained the robot by physically guiding it through the task of first reaching into a bowl, scooping up marbles, transporting them over an empty bowl, and pouring them in. After a few demonstrations, the team then used a pretrained LLM and asked the model to list the steps involved in scooping marbles from one bowl to another. The researchers then used their new algorithm to connect the LLMs defined subtasks with the robots motion trajectory data. The algorithm automatically learned to map the robots physical coordinates in the trajectories and the corresponding image view to a given subtask.

The team then let the robot carry out the scooping task on its own, using the newly learned grounding classifiers. As the robot moved through the steps of the task, the experimenters pushed and nudged the bot off its path, and knocked marbles off its spoon at various points. Rather than stop and start from the beginning again, or continue blindly with no marbles on its spoon, the bot was able to self-correct, and completed each subtask before moving on to the next. (For instance, it would make sure that it successfully scooped marbles before transporting them to the empty bowl.)

With our method, when the robot is making mistakes, we dont need to ask humans to program or give extra demonstrations of how to recover from failures, Wang says. Thats super exciting because theres a huge effort now toward training household robots with data collected on teleoperation systems. Our algorithm can now convert that training data into robust robot behavior that can do complex tasks, despite external perturbations.

Reprinted with permission of MIT News

***

All Premium Members get to view The Good Men Project with NO ADS. A $50 annual membership gives you an all access pass. You can be a part of every call, group, class and community. A $25 annual membership gives you access to one class, one Social Interest group and our online communities. A $12 annual membership gives you access to our Friday calls with the publisher, our online community. Need more info? A complete list of benefits is here.

Photo credit: unsplash

See more here:

Engineering Household Robots to Have a Little Common Sense - The Good Men Project

Read More..