Their new ‘SELF-DISCOVER’ prompting framework – published this week on arXiV and Hugging Face – represents a significant leap beyond existing techniques, potentially revolutionising the performance of leading models such as OpenAI’s GPT-4 and Google’s PaLM 2.

The framework promises substantial enhancements in tackling challenging reasoning tasks. It demonstrates remarkable improvements, boasting up to a 32% performance increase compared to traditional methods like Chain of Thought (CoT). This novel approach revolves around LLMs autonomously uncovering task-intrinsic reasoning structures to navigate complex problems.

At its core, the framework empowers LLMs to self-discover and utilise various atomic reasoning modules – such as critical thinking and step-by-step analysis – to construct explicit reasoning structures.

By mimicking human problem-solving strategies, the framework operates in two stages:

• Stage one involves composing a coherent reasoning structure intrinsic to the task, leveraging a set of atomic reasoning modules and task examples.
• During decoding, LLMs then follow this self-discovered structure to arrive at the final solution.

In extensive testing across various reasoning tasks – including Big-Bench Hard, Thinking for Doing, and Math – the self-discover approach consistently outperformed traditional methods. Notably, it achieved an accuracy of 81%, 85%, and 73% across the three tasks with GPT-4, surpassing chain-of-thought and plan-and-solve techniques.

However, the implications of this research extend far beyond mere performance gains.

By equipping LLMs with enhanced reasoning capabilities, the framework paves the way for tackling more challenging problems and brings AI closer to achieving general intelligence. Transferability studies conducted by the researchers further highlight the universal applicability of the composed reasoning structures, aligning with human reasoning patterns.

As the landscape evolves, breakthroughs like the SELF-DISCOVER prompting framework represent crucial milestones in advancing the capabilities of language models and offering a glimpse into the future of AI.

(Photo by Victor on Unsplash)

See also: The UK is outpacing the US for AI hiring

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with Digital Transformation Week and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post DeepMind framework offers breakthrough in LLMs’ reasoning appeared first on AI News.

The new Science and Technology Framework is the first major piece of work from the newly created Department for Science, Innovation and Technology and will challenge every part of government to better put the UK at the forefront of global science and technology this decade through 10 key actions – creating a coordinated cross-government approach.

In doing so, the Government will foster the right conditions for industry innovation and world leading scientific research to deliver high-paid jobs of the future, grow the economy in cutting-edge industries, and improve people’s lives from better healthcare to security.

The 10 points of the new Science and Technology Framework centre on:

• identifying, pursuing and achieving strategic advantage in the technologies that are most critical to achieving UK objectives
• showcasing the UK’s S&T strengths and ambitions at home and abroad to attract talent, investment and boost our global influence
• boosting private and public investment in research and development for economic growth and better productivity
• building on the UK’s already enviable talent and skills base
• financing innovative science and technology start-ups and companies
• capitalising on the UK government’s buying power to boost innovation and growth through public sector procurement
• shaping the global science and tech landscape through strategic international engagement, diplomacy and partnerships
• ensuring researchers have access to the best physical and digital infrastructure for R&D that attracts talent, investment and discoveries
• leveraging post-Brexit freedoms to create world-leading pro-innovation regulation and influence global technical standards
• creating a pro-innovation culture throughout the UK’s public sector to improve the way our public services run

The delivery of this new Framework will begin immediately with an initial raft of projects, worth around £500 million in new and existing funding, which will help ensure the UK has the skills and infrastructure to take a global lead in game-changing technologies.

Prime Minister Rishi Sunak said: “Trailblazing science and innovation have been in our DNA for decades. But in an increasingly competitive world, we can only stay ahead with focus, dynamism and leadership.

“That’s why we’re setting out 10 key actions under a bold new plan to cement our place as a global science and technology superpower by 2030 – from pursuing transformational technologies like AI and supercomputing to attracting top talent and ensuring they have the tools they need to succeed.

“The more we innovate, the more we can grow our economy, create the high-paid jobs of the future, protect our security, and improve lives across the country.”

Science, Innovation and Technology Secretary Michelle Donelan said: “Innovation and technology are our future. They hold the keys to everything from raising productivity and wages, to transforming healthcare, reducing energy prices and ultimately creating jobs and economic growth in the UK, providing the financial firepower allowing us to spend more on public services.

“That is why we are putting the full might of the British government and our private sector partners behind our push to become a scientific and technological superpower, because only through being world-leaders in future industries like AI and quantum will we be able to improve the lives of every Briton.”

The initial package of projects to drive forward the actions of the Science and Technology Framework includes:

• £250 million investment in three transformational technologies to build on the UK’s global leadership in AI, quantum technologies and engineering biology, so they can help a range of industries tackle the biggest global challenges like climate change and health care. This forms part of our commitment to the five technologies within the science and technology framework, which also includes semiconductors and future telecoms
• publication of Sir Paul Nurse’s Independent Review of the Research, Development and Innovation Organisational Landscape with recommendations to make the most of the UK’s research organisations, ensuring they are effective, sustainable and responsive to global challenges
• testing different models of funding science, to support a range of innovative institutional models, such as Focused Research Organisations (known as FROs), working with industry and philanthropic partners to open up new funding for UK research. For example, this could include working with a range of partners to increase investment in the world leading UK Biobank, to support the continued revolution in genetic science
• up to £50 million to spur co-investment in science from the private sector and philanthropists to drive the discoveries of the future, subject to business cases. The government is already talking to Schmidt Futures, a philanthropic initiative of Eric and Wendy Schmidt, about additional support of up to $20 million as part of this work
• £117 million of existing funding to create hundreds of new PhDs for AI researchers and £8 million to find the next generation of AI leaders around the world to do their research in the UK
• a £50 million uplift to World Class Labs funding to help research institutes and universities to improve facilities so UK researchers have access to the best labs and equipment they need to keep producing world-class science, opening up entirely new avenues for economic growth and job creation
• a £10 million uplift to the UK Innovation and Science Seed Fund, totalling £50 million, to boost the UK’s next tech and science start-ups who could be the next Apple, Google or Tesla
• plans to set up an Exascale supercomputer facility – the most powerful compute capability which could solve problems as complex as nuclear fusion – as well as a programme to provide dedicated compute capacity for important AI research, as part of the response to the Future of Compute Review
• £9 million in government funding to support the establishment of a quantum computing research centre by PsiQuantum in Daresbury in the North-West

The Framework has been designed in consultation with industry experts and academics, to help deliver stronger growth, better jobs, and bold discoveries to tackle the challenges of today and tomorrow.

The plan will be a cross-government endeavour led by the Department for Science, Innovation and Technology (DSIT) to bring together responsibility for the UK’s world class research and innovation system with the 5 technologies of tomorrow – quantum, AI, engineering biology, semiconductors, future telecoms plus life sciences and green technologies, into one single department for the first time.

Director of the Francis Crick Institute and lead reviewer of the Landscape Review, Paul Nurse, said: “It is absolutely right, as the Prime Minister has said, that the future of the UK depends upon research, science and technology. Only by being a leading science nation can the UK drive a sustainable economy, increased productivity and generate societal benefits such as improved healthcare and protecting the environment.

“The Government’s endorsement of this approach is to be fully supported. My Review of the research, development and innovation landscape makes a range of recommendations across the whole RDI endeavour, which if adopted together, provides a blueprint for government to make the UK a genuine science superpower.”

Today the government is also announcing a further extension until 30 June 2023 of the financial guarantee provided to the UK’s Horizon Europe applicants so that eligible, successful bids for calls closing on or before this date continue to be guaranteed funding, supporting them to continue their important work in research and innovation.

Science, innovation and technology are the drivers of economic growth and productivity. More than half of the UK’s future labour productivity growth will come from adopting the best available technologies and the rest from ‘pushing the frontier’ of technology even further. Each £1 of public R&D investment leverages £2 of private R&D investment in the long run.

The announcements build on existing Government efforts to support science and technology. This includes setting up the Advanced Research and Invention Agency (ARIA) to fund high-risk, high-reward R&D; investing £100 million in a pilot bringing together national and local partners in Glasgow, Greater Manchester and the West Midlands to accelerate their growth into major, globally competitive centres for research and innovation; and publishing the UK Digital Strategy committed to rolling out world-class digital infrastructure, unlocking the value of data to create growth, innovation and societal benefits across the UK and harnessing digital transformation to build a more inclusive, competitive and innovative digital economy.

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post £370m plan launched to turn UK into ‘tech superpower’ appeared first on AI News.

Breakthroughs in medical AIs are helping to reduce patient suffering, the likelihood and/or severity of long-term complications, and even save lives across a number of ailments.

Some of the benefits from medical AI breakthroughs are achieved through improved understanding leading to better treatment, while others are due to reducing the amount of time healthcare professionals have to spend on repetitive tasks.

Over 100,000 people in the UK suffer from a stroke per year; with over 32,000 deaths as a result. NHS SBS sought out how AI can help tackle one of the UK’s leading causes of death and disability.

Adam Nickerson, NHS SBS Senior Category Manager – Digital & IT, said:

“This use of AI is a prime example of how new technologies have the potential to transform NHS patient care, speeding up diagnosis and treatment times by ensuring that expert clinical resource is targeted where it has the greatest impact for the patient. 

By identifying areas in which technology can be used to help speed up patient pathways, clinicians have more time for providing personalised care and patient waiting lists – exacerbated by the pandemic, are reduced.

We have been pleased to work alongside some of the country’s leading tech minds, expert stroke clinicians, and policy leaders to develop this unique framework, which will go a long way to enabling more rapid uptake of Stroke AI software across the NHS.”

While AI can be a powerful tool in medicine, it can be difficult to ensure solutions are evidence-based and cost-effective. That’s where the new ‘Provision of AI Software in Neuroscience for Stroke Decision Making Support’ procurement framework comes in.

The framework was developed with contributions from across NHS England and NHS Improvement (NHSEI), clinical leads from the 20 Integrated Stroke Delivery Networks across England, the Academic Health Science Network, and with further input from NHSX and the Care Quality Commission.

Darrien Bold, National Digital and AI Lead for Stroke at NHSEI, commented:

“We are already seeing the impact AI decision-support software is having on stroke pathways across the country, and the introduction of this framework will drive forward further progress in delivering best-practice care where rapid assessment and treatment are of the essence.

Over the past 18 months, the heath and care system has been compelled to look to new technologies to continue providing frontline care, and the stroke community has embraced new ways of working in times of unprecedented pressure.

This framework agreement will be of great benefit as we implement the NOSIP – driving better outcomes, better patient experience and better patient safety, using new technology quickly, safely and innovatively.”

Time is very much of the essence when it comes to strokes. The framework will enable the procurement of AI solutions that analyse images to detect ischaemic or haemorrhagic strokes and provide real-time interpretations to augment the review, diagnosis, and delivery of time-dependent treatments.

While manual review of imagery can take up to 30 minutes to interpret, AI is able to do so within seconds.

“Rapid brain imaging and its interpretation is arguably one of the most important steps in the care of patients with stroke-like symptoms,” commented Dr David Hargroves, Getting It Right First Time (GIRFT) Clinical Lead for Stroke and National Specialty Advisor for Stroke Medicine at NHSEI.

“Incorporating AI decision support software is likely to improve access to disability-saving interventions to thousands of patients. This framework agreement supplies a valuable platform to support providers of hyperacute stroke care in the purchase of AI software.”

As part of the NHS Long Term Plan, the health service aims to achieve a tenfold increase in the proportion of stroke victims who receive a thrombectomy by 2022—estimated to enable around 1,600 more patients per year to live independently.

AI will be key to achieving the NHS’ long-term goals across care for stroke patients and more. We look forward to seeing all the ways health services around the world put AI to good use over the coming years to improve patient outcomes.

(Photo by Ian Taylor on Unsplash)

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo. The next events in the series will be held in Santa Clara on 11-12 May 2022, Amsterdam on 20-21 September 2022, and London on 1-2 December 2022.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post The NHS can now access ‘pioneering’ AI stroke diagnosis software appeared first on AI News.

PyTorch Live was unveiled during PyTorch Developer Day and enables anyone to build mobile ML demo apps using JavaScript, the world’s most popular programming language.

While on-device AI demos cannot currently be shared, Meta says that functionality is on the way. Developers can start building custom machine learning models to later share with the broader PyTorch community.

PyTorch was publicly launched by Meta back in January 2017, when the company was still known as Facebook. The open-source machine learning library quickly became a firm favourite among the developer and data science communities.

As the PyTorch name suggests, the main library’s interface is designed around Python but it also has a C++ interface. 

The once-dominant machine learning library, TensorFlow, had a two-year headstart on PyTorch but has been falling behind in usage in recent years.

In 2018, GitHub’s Octoverse report highlighted the growth of PyTorch as an open-source project outpacing that of TensorFlow. PyTorch grew by 2.8x that year compared to TensorFlow’s still not insubstantial 1.8x.

That edge for PyTorch appears to be eating into TensorFlow’s early mover advantage.

TensorFlow appeared in three times more job listings in Indeed, Monster, SimplyHired, and LinkedIn as PyTorch in April 2019. However, TensorFlow’s edge in job-listing mentions dropped to 2x in 2020.

Over the past year, PyTorch has also overtaken TensorFlow in worldwide Google searches:

PyTorch Live looks set to accelerate the success of the machine learning library. The tools use React Native for building cross-platform visual user interfaces and PyTorch Mobile powers on-device inference.

Anyone wanting to get started with PyTorch Live can do so through its command-line interface setup and/or its data processing API.

(Image Credit: Meta)

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo North America on 11-12 May 2022.

The post Meta releases PyTorch Live for creating mobile ML demos ‘in minutes’ appeared first on AI News.

The tech giant is often considered as “China’s Google” and, just like its Western counterpart, is one of the largest AI companies in the world.

“AI technology is growing increasingly complex, and integrated innovation has made AI more powerful,” said Haifeng Wang, CTO of Baidu.

“As AI technology plays an expanding role in a wider range of industries and drives a new era of technological revolution and industrial transformation, it is increasingly important to lower the threshold for different real-world applications and to increase accessibility to AI development platforms.”

At Baidu World 2021, the company made two significant AI announcements.

The first is Baidu Brain 7.0 which promises deeper integration of knowledge sources and deep learning. The open platform now features language comprehension and reasoning.

Baidu’s latest AI platform version works in tandem with the company’s new Kunlun II chip. The chip uses a 7nm process to match the current leaders including Graphcore and Huawei.

Kunlun II is equipped with Baidu’s second-gen XPU architecture. Over the previous generation, Baidu claims the latest Kunlun chip offers 2-3x more processing power.

The latest chip from Baidu works with the company’s open-source deep learning framework PaddlePaddle that has been used by more than 3.6 million developers around the world to build 400,000 AI models.

Baidu claims the models built using PaddlePaddle have led to applications to help water management systems run more efficiently, improve quality control in manufacturing, and even help athletes improve their training.

(Image Credit: Baidu)

Find out more about Digital Transformation Week North America, taking place on 9-10 November 2021, a virtual event and conference exploring advanced DTX strategies for a ‘digital everything’ world.

The post Baidu debuts Brain 7.0 alongside mass production of Kunlun II chip appeared first on AI News.

While still technically in pre-release, the Mac-optimised TensorFlow fork supports native hardware acceleration on Mac devices with M1 or Intel chips through Apple’s ML Compute framework.

The new TensorFlow release boasts of an over 10x speed improvement for common training tasks. While impressive, it has to be taken in the context that the GPU was not previously used for training tasks. 

A look at the benchmarks still indicates a substantial gap between the Intel and M1-based Macs across various machine learning models:

In a blog post, Pankaj Kanwar, Tensor Processing Units Technical Program Manager at Google, and Fred Alcober, TensorFlow Product Marketing Lead at Google, wrote:

“These improvements, combined with the ability of Apple developers being able to execute TensorFlow on iOS through TensorFlow Lite, continue to showcase TensorFlow’s breadth and depth in supporting high-performance ML execution on Apple hardware.”

We can only hope that running these workloads doesn’t turn MacBooks into expensive frying pans—but the remarkable efficiency they’ve displayed so far gives little cause for concern.

Interested in hearing industry leaders discuss subjects like this? Attend the co-located 5G Expo, IoT Tech Expo, Blockchain Expo, AI & Big Data Expo, and Cyber Security & Cloud Expo World Series with upcoming events in Silicon Valley, London, and Amsterdam.

The post TensorFlow is now available for those shiny new ARM-based Macs appeared first on AI News.

The Huawei Ascend 910 is the “world’s most powerful AI processor,” according to a press release on Friday. The chip’s specs were first announced during last year’s Huawei Connect event in Shanghai.

Eric Xu, Rotating Chairman of Huawei, said:

“We have been making steady progress since we announced our AI strategy in October last year. Everything is moving forward according to plan, from R&D to product launch.

We promised a full-stack, all-scenario AI portfolio and today we delivered, with the release of Ascend 910 and MindSpore. This also marks a new stage in Huawei’s AI strategy.”

Huawei claims the final version of the Ascend 910 not only performs as promised, but it does so with much lower power consumption.

For half-precision floating point (FP16) operations, Ascend 910 delivers 256 TeraFLOPS performance. For integer precision calculations (INT8), it delivers 512 TeraOPS.

Huawei initially expected the Ascend 910’s max power consumption to be 350W but the company has managed to deliver the promised performance with a max consumption of just 310W.

“Ascend 910 performs much better than we expected,” said Xu. “Without a doubt, it has more computing power than any other AI processor in the world.”

Alongside the Ascend 910, Huawei has launched an AI computing framework called MindSpore.

Last year, Huawei announced three goals for MindSpore:

• Easy development: Reduce training time and costs.
• Efficient execution: Use the least amount of resources with the highest possible OPS/W.
• Adaptable to all scenarios: Including device, edge, and cloud applications.

Huawei claims that MindSpore requires 20 percent fewer lines of code than other leading frameworks when used for a typical neural network for natural language processing.

“MindSpore will go open source in the first quarter of 2020,” said Xu. “We want to drive broader AI adoption and help developers do what they do best.”

The Chinese tech behemoth continues to expand its presence despite battling a US trade ban. The US has been pressuring its allies to ban Huawei over concerns it poses a national security threat.

While security must always be prioritised, few can dispute the innovation which Huawei brings across its business. Today’s announcements show the kind of innovations which US companies may miss out on if a deal cannot be reached, putting them at a disadvantage to Chinese rivals.

Interested in hearing industry leaders discuss subjects like this? Attend the co-located 5G Expo, IoT Tech Expo, Blockchain Expo, AI & Big Data Expo, and Cyber Security & Cloud Expo World Series with upcoming events in Silicon Valley, London, and Amsterdam.

The post Huawei unveils high-end AI chip for servers alongside MindSpore framework appeared first on AI News.

Reinforcement learning has been used for some of the most impressive AI demonstrations thus far, including those which beat human professional gamers at Alpha Go and Dota 2. Google subsidiary DeepMind uses it for its Deep Q-Network (DQN).

Building a reinforcement learning framework takes both time and significant resources. For AI to reach its full potential, it needs to become more accessible.

Starting today, Google is making an open source reinforcement framework based on TensorFlow – its machine learning library – available on GitHub.

Pablo Samuel Castro and Marc G. Bellemare, Google Brain researchers, wrote in a blog post:

“Inspired by one of the main components in reward-motivated behavior in the brain and reflecting the strong historical connection between neuroscience and reinforcement learning research, this platform aims to enable the kind of speculative research that can drive radical discoveries.

This release also includes a set of collabs that clarify how to use our framework.”

Google’s framework was designed with three focuses: flexibility, stability, and reproducibility.

The company is providing 15 code examples for the Arcade Learning Environment — a platform which uses video games to evaluate the performance of AI technology — along with four distinct machine learning models: C51, the aforementioned DQN, Implicit Quantile Network, and the Rainbow agent.

Reinforcement learning is among the most effective methods of training. If you’re training a dog, offering treats as a reward for the desired behaviour is a key example of positive reinforcement in practice.

Training a machine is a similar concept, only the rewards are delivered or withheld as ones and zeros instead of tasty goods or a paycheck.

“Our hope is that our framework’s flexibility and ease-of-use will empower researchers to try out new ideas, both incremental and radical,” wrote Bellemare and Castro. “We are already actively using it for our research and finding it is giving us the flexibility to iterate quickly over many ideas.”

“We’re excited to see what the larger community can make of it.”

What are your thoughts on Google’s open-sourcing of its reinforcement learning framework? Let us know in the comments.

 Interested in hearing industry leaders discuss subjects like this and sharing their use-cases? Attend the co-located AI & Big Data Expo events with upcoming shows in Silicon Valley, London and Amsterdam to learn more. Co-located with the  IoT Tech Expo, Blockchain Expo and Cyber Security & Cloud Expo so you can explore the future of enterprise technology in one place.

The post Google improves AI model training by open-sourcing framework appeared first on AI News.