The PhiSat-1 satellite is about the size of a cereal box and was ejected from a rocket’s dispenser alongside 45 other satellites. The rocket launched from Guiana Space Centre on September 2nd.

Intel has integrated its Movidius Myriad 2 Vision Processing Unit (VPU) into PhiSat-1 – enabling large amounts of data to be processed on the device. This helps to prevent useless data being sent back to Earth and consuming precious bandwidth.

“The capability that sensors have to produce data increases by a factor of 100 every generation, while our capabilities to download data are increasing, but only by a factor of three, four, five per generation,” says Gianluca Furano, data systems and onboard computing lead at the ESA.

Around 30 percent data savings are expected by using AI at the edge on the PhiSat-1.

“Space is the ultimate edge,” says Aubrey Dunne, chief technology officer of Ubotica. “The Myriad was absolutely designed from the ground up to have an impressive compute capability but in a very low power envelope, and that really suits space applications.”

PhiSat-1 is currently in a sun-synchronous orbit around 329 miles (530 km) above Earth and travelling at over 17,000mph (27,500kmh).

The satellite’s mission is to assess things like polar ice for monitoring climate change, and soil moisture for the growth of crops. One day it could help to spot wildfires in minutes rather than hours or detect environmental accidents at sea.

A successor, PhiSat-2, is currently planned to test more of these possibilities. PhiSat-2 will also carry another Myriad 2.

Myriad 2 was not originally designed for use in orbit. Specialist chips which are protected against radiation are typically used for space missions and can be “up to two decades behind state-of-the-art commercial technology,” explains Dunne.

Incredibly, the Myriad 2 survived 36 straight hours of being blasted with radiation at CERN in late-2018 without any modifications.

ESA announced the joint team was “happy to reveal the first-ever hardware-accelerated AI inference of Earth observation images on an in-orbit satellite.”

PhiSat-1 and PhiSat-2 will be part of a future network with intersatellite communication systems.

(Image Credit: CERN/M. Brice)

The post Intel, Ubotica, and the ESA launch the first AI satellite appeared first on AI News.

The Perelman School of Medicine at UPenn is working with Intel Labs to co-develop technology based on federated learning, a machine learning technique which trains an algorithm across various devices without exchanging data samples.

The goal is therefore to preserve privacy. Penn Medicine and Intel Labs have claimed they were first to publish a paper on federated learning in medical imaging, offering accuracy with a trained model to more than 99% of a model trained in a non-private method. Work which will build on this, according to the two companies, will ‘leverage Intel software and hardware to implement federated learning in a manner that provides additional privacy protection to both the model and the data.’

The two companies will be joined by 29 healthcare and research institutions from seven countries.

“AI shows great promise for the early detection of brain tumours, but it will require more data than any single medical centre holds to reach its full potential,” said Jason Martin, principal engineer at Intel Labs in a statement.

Artificial intelligence initiatives in healthcare continue apace. Microsoft recently announced details of a $40 million ‘AI for Health’ project, while last month startup Babylon Health stated its belief that it can appropriately triage patients in 85% of cases.

Read the full Intel announcement here.

Photo by jesse orrico on Unsplash

The post Intel and UPenn utilising federated learning to identify brain tumours appeared first on AI News.

With concerns around AI spanning from dangerous indifference to innovation-halting scaremongering; it’s clear there’s a need for a system to achieve a healthy balance.

“AI systems have been developed in ways that are inconsistent with the stated values of those developing them,” the researchers wrote. “This has led to a rise in concern, research, and activism relating to the impacts of AI systems.”

The researchers note that significant work has gone into articulating ethical principles by many players involved with AI development, but the claims are meaningless without some way to verify them.

“People who get on airplanes don’t trust an airline manufacturer because of its PR campaigns about the importance of safety – they trust it because of the accompanying infrastructure of technologies, norms, laws, and institutions for ensuring airline safety.”

Among the core ideas put forward is to pay developers for discovering bias in algorithms. Such a practice is already widespread in cybersecurity; with many companies offering up bounties to find bugs in their software.

“Bias and safety bounties would extend the bug bounty concept to AI and could complement existing efforts to better document data sets and models for their performance limitations and other properties,” the authors wrote.

“We focus here on bounties for discovering bias and safety issues in AI systems as a starting point for analysis and experimentation but note that bounties for other properties (such as security, privacy protection, or interpretability) could also be explored.”

Another potential avenue is so-called “red teaming,” the creation of a dedicated team which adopts the mindset of a possible attacker to find flaws and vulnerabilities in a plan, organisation, or technical system.

“Knowledge that a lab has a red team can potentially improve the trustworthiness of an organization with respect to their safety and security claims.”

A red team alone is unlikely to give too much confidence; but combined with other measures can go a long way. Verification from parties outside an organisation itself will be key to instil trust in that company’s AI developments.

“Third party auditing is a form of auditing conducted by an external and independent auditor, rather than the organization being audited, and can help address concerns about the incentives for accuracy in self-reporting.”

“Provided that they have sufficient information about the activities of an AI system, independent auditors with strong reputational and professional incentives for truthfulness can help verify claims about AI development.”

The researchers highlight that a current roadblock with third-party auditing is that there’s yet to be any techniques or best practices established specifically for AI. Frameworks, such as Claims-Arguments-Evidence (CAE) and Goal Structuring Notation (GSN), may provide a starting place as they’re already widely-used for safety-critical auditing.

Audit trails, covering all steps of the AI development process, are also recommended to become the norm. The researchers again point to commercial aircraft, as a safety-critical system, and their use of flight data recorders to capture multiple types of data every second and provide a full log.

“Standards setting bodies should work with academia and industry to develop audit trail requirements for safety-critical applications of AI systems.”

The final suggestion for software-oriented methods of verifying AI ethics claims is the use of privacy-preserving machine learning (PPML).

Privacy-preserving machine learning aims to protect the privacy of data or models used in machine learning, at training or evaluation time, and during deployment.

Three established types of PPML are covered in the paper: Federated learning, differential privacy, and encrypted computation.

“Where possible, AI developers should contribute to, use, and otherwise support the work of open-source communities working on PPML, such as OpenMined, Microsoft SEAL, tf-encrypted, tf-federated, and nGraph-HE.”

The researchers, representing some of the most renowned institutions in the world, have come up with a comprehensive package of ways any organisation involved with AI development can provide assurance to governance and the wider public to ensure the industry can reach its full potential responsibly.

You can find the full preprint paper on arXiv here (PDF)

(Photo by Alexander Sinn on Unsplash)

The post Leading AI researchers propose ‘toolbox’ for verifying ethics claims appeared first on AI News.

Thermal imaging is often used to protect privacy because it obscures personally identifying details such as eye colour. In some places, like medical facilities, it’s often compulsory to use images which obscure such details.

AI is opening up many new possibilities so Intel’s researchers set out to determine whether thermal imaging still offers a high degree of privacy.

Intel’s team used two sets of data sets:

• The first set, known as SC3000-DB, was created using a Flir ThermaCam SC3000 infrared camera. The data set features 766 images of 40 volunteers (21 women and 19 men) who each sat in front of a camera for two minutes.

• The second set, known as IRIS, was created by the Visual Computing and Image Processing Lab at Oklahoma State University. It features 4,190 images collected by 30 people and differs from the first set in that it contains various head angles and expressions. 

Each image from the data sets were first cropped to only contain each person’s face. 

A machine learning model then sought to numerically label facial features from the images as vectors. Another model, trained on VGGFace2 – a model trained on visible light images – was used to validate whether it could be applied to thermal images.

Here’s the full results for each data set:

The model trained on visible image data performed well in distinguishing among volunteers by extracting their facial features. 99.5 percent accuracy was observed for the SC3000-DB data set and 82.14 percent for IRIS.

Intel’s research shows that thermal imaging may not offer the privacy that many currently believe it to and it’s already possible to distinguish people using it.

“Many promising visual-processing applications, such as non-contact vital signs estimation and smart home monitoring, can involve private and or sensitive data, such as biometric information about a person’s health,” wrote the researchers.

“Thermal imaging, which can provide useful data while also concealing individual identities, is therefore used for many applications.”

You can find Intel’s full research here.

Interested in hearing industry leaders discuss subjects like this? , , , AI &

The post Intel examines whether AI can recognise faces using thermal imaging appeared first on AI News.

The new Nervana Neural Network Processor for Inference (NNP-I) processor has a more approachable codename of Spring Hill.

Spring Hill is a modified 10nm Ice Lake processor which sits on a PCB and slots into an M.2 port typically used for storage.

According to Intel, the use of a modified Ice Lake processor allows Spring Hill to handle large workloads and consume minimal power. Two compute cores and the graphics engine have been removed from the standard Ice Lake design to accommodate 12 Inference Compute Engines (ICE).

In a summary, Intel detailed six main benefits it expects from Spring Hill:

1. Best in class perf/power efficiency for major data inference workloads.
2. Scalable performance at wide power range.
3. High degree of programmability w/o compromising perf/power efficiency.
4. Data centre at scale.
5. Spring Hill solution – Silicon and SW stack – sampling with definitional partners/customers on multiple real-life topologies.
6. Next two generations in planning/design.

Intel’s first chip for AI comes after the company invested in several Isreali artificial intelligence startups including Habana Labs and NeuroBlade. The investments formed part of Intel’s strategy called ‘AI Everywhere’ which aims to increase the firm’s presence in the market.

Naveen Rao, Intel vice president and general manager, Artificial Intelligence Products Group, said:

“To get to a future state of ‘AI everywhere,’ we’ll need to address the crush of data being generated and ensure enterprises are empowered to make efficient use of their data, processing it where it’s collected when it makes sense and making smarter use of their upstream resources.

Data centers and the cloud need to have access to performant and scalable general purpose computing and specialized acceleration for complex AI applications. In this future vision of AI everywhere, a holistic approach is needed—from hardware to software to applications.”

Facebook has said it will be using Intel’s new Spring Hill processor. Intel already has two more generations of the NNP-I in development.

Interested in hearing industry leaders discuss subjects like this? , , , AI &

The post Intel unwraps its first chip for AI and calls it Spring Hill appeared first on AI News.

As part of the collaboration, Lenovo will be enhancing Intel’s complete portfolio of HPC and AI hardware and software solutions. Both will help accelerate the convergence of the technologies to unravel new levels of customer insight. The merging of the Lenovo Neptune liquid cooling technology together with the second Generation Intel Xeon Scalable platform has already produced significant results from joint engineering and using a unique combination of HPC IP from the two companies.

The Lenovo-Intel alliance would focus mainly on three areas:

• Systems and solutions
• Software optimisation for HPC and AI convergence
• Ecosystem enablement

Currently, 173 of the TOP500 fastest supercomputers, across 19 markets, in the world run on Lenovo servers.

In addition, Microsoft recently announced £819.6m investment in OpenAI to make artificial general intelligence (AGI). Sam Altman, CEO of OpenAI, said: “The creation of AGI will be the most important technological development in human history, with the potential to shape the trajectory of humanity. Our mission is to ensure that AGI technology benefits all of humanity, and we’re working with Microsoft to build the supercomputing foundation on which we’ll build AGI. We believe it’s crucial that AGI is deployed safely and securely and that its economic benefits are widely distributed. We are excited about how deeply Microsoft shares this vision.”

 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, , and Cyber Security & .

The post Intel and Lenovo enter into multi-year partnership over AI and HPC appeared first on AI News.

Speaking on stage at the conference in Beijing, Intel corporate vice president Naveen Rao made the announcement.

“The next few years will see an explosion in the complexity of AI models and the need for massive deep learning compute at scale. Intel and Baidu are focusing their decade-long collaboration on building radical new hardware, codesigned with enabling software, that will evolve with this new reality – something we call ‘AI 2.0.’

Intel’s so-called Neural Network Processor for Training is codenamed NNP-T 1000 and designed for training deep learning models at lightning speed. A large amount (32GB) of HBM memory and local SRAM is put closer to where computation happens to enable more storage of model parameters on-die, saving significant power for an increase in performance.

The NNP-T 1000 is set to ship alongside the Neural Network Processor for Inference (NNP-I 1000) chip later this year. As the name suggests, the NNP-I 1000 is designed for AI inferencing and features general-purpose processor cores based on Intel’s Ice Lake architecture.

Baidu and Intel have a history of collaborating in AI. Intel has helped to optimise Baidu’s PaddlePaddle deep learning framework for its Xeon Scalable processors since 2016. More recently, Baidu and Intel developed the BIE-AI-Box – a hardware kit for analysing the frames of footage captured by cockpit cameras.

Intel sees a great deal of its future growth in AI. The company’s AI chips generated $1 billion in revenue last year and Intel expects a growth rate of 30 percent annually up to $10 billion by 2022.

The post Smells Like AI Spirit: Baidu will help develop Intel’s Nervana neural processor appeared first on AI News.

Extracting value from sentiment analysis

Determining the value of cryptocurrencies is tricky. Unlike the conventional stock market, worth does not tightly correspond with factors such as cash flow or available assets. Instead, investors must rely on sentiment. But how do they make sense of all that is being said in a timely manner? Developer Teju Tadi believes he may have the answer.

The majority of cryptocurrency price movements can be determined by the herd instinct – when people think and act in the same way as the majority around them. Based on this, Teju says the sentiment analysis of news headlines, Reddit posts, and tweets is a good indicator of the direction of cryptocurrency price movements. Using recursive neural tensor networks (RNTNs), Teju is already analysing the sentiment of these short texts to develop an AI cryptocurrency trading bot.

Recursive neural tensor networks

RNTNs assess the semantic compositionality of text, which is vital to being able to accurately determine sentiment from a sparse set of information like a tweet.

RNTNs parse data into a binary tree. Specific vector representations are formed of all the words and are represented as leaves. From the bottom up, these vectors become the parameters to optimize and serve as feature inputs to a softmax classifier*. Vectors are classified into five classes and assigned a score.

“When similarities are encoded between two words, the two vectors move across to the next root. A score and class are outputted. A score represents the positivity or negativity of a parse while the class encodes the structure in current parses. The first leaf group receives the parse and then the second leaf receives the next word. The score of the parse with all three words are outputted and it moves on to the next root group,” says Teju.

‘The recursion process continues until all inputs are used up, with every single word included. In practical applications RNTN’s end up being more complex than this. Rather than using the immediate next word in a sentence for the next leaf group, an RNTN would try all the next words and eventually checks vectors that represent entire sub-parses. Performing this at every step of the recursive process, the RNTN can analyse every possible score of the syntactic parse.”

The figure below shows how a sentence is parsed and analysed using an RNTN approach. Teju also explains the process in this video.

Figure 1: Example of scoring from the Stanford Treebank.

Supporting technologies

As a member of the Intel® AI Academy, Teju was able to use the Intel® AI DevCloud to run the recurrent neural networks and experiment with Twitter data to see how the models were working. Running on Intel® Xeon® Scalable processors, the Intel AI DevCloud is preloaded with frameworks and tools to quickly launch machine learning and deep learning projects. These include neon™ framework, Intel® Optimization for Theano*, Intel® Optimization for TensorFlow*, Intel® Optimization for Caffe*, Intel® Distribution for Python* (including NumPy, SciPy, and scikit-learn*) and the Keras* library.

If you’re considering starting a deep learning project, reading To Get Started with Intel AI DevCloud will give you a good understanding of available models and how to initiate training using the AI DevCloud.

Opportunities for developers

Teju has set up a business Mycointrac*, focused on providing cryptocurrency market intelligence. “Once the product is fully developed,” he said, “I plan to utilise the data provided by it as one of the factors to make key investment decisions for my new cryptocurrency hedge fund, Sentience Investments L.P., which has been operational since January first. The plan is to develop trading strategies based on a number of high-frequency, machine-learning techniques, as well as deep learning and sentiment analysis.”

Teju hopes RNTNs will also help him take advantage of other opportunities, such as arbitrage – the simultaneous buying and selling of an asset in different markets. The profit being the difference between the two market prices.

The financial sector offers many great opportunities for developers who can utilise AI to address some of its most pressing challenges. For example, consistently gaining high returns on stock market investments.

The post How AI is helping to predict cryptocurrency value appeared first on AI News.

Speaking during the company’s Innovation Summit in Santa Clara, Intel Executive VP Navin Shenoy revealed a new focus on AI development.

The company’s AI-focused Xeon processors generated $1 billion in revenues during 2017. By 2022, it expects to be generating around $10 billion per year.

AI is set to be implemented in many areas of our lives in the coming years, across a variety of devices.

Shenoy claims recent breakthroughs have increased the company’s AI performance by 200x since 2014. He teases further improvements are on their way in upcoming releases.

The company will be launching its ‘Cascade Lake’ Xeon processor later this year with 11 times better performance for AI image recognition.

Arriving in 2019 will be ‘Cooper Lake’ which uses 14-nanometer manufacturing and will feature even better performance. In 2020, however, the company is targeting ‘Ice Lake’ with 10-nanometer manufacturing technology.

“After 50 years, this is the biggest opportunity for the company,” says Shenoy. “We have 20 percent of this market today.”

The admission it currently has a small share of the market today is bold and shows the company is confident about significantly upping that percentage in the coming years. It faces significant competition from Nvidia in particular.

Intel’s revenues were around a third data-centric five years ago. Now, it’s around half of Intel’s business.

Shenoy’s comments today show how seriously Intel is taking its AI business and the firm’s confidence it will be a major player.

What are your thoughts on Intel’s AI business?

The post Intel’s AI chip business is now worth $1bn per year, $10bn by 2022 appeared first on AI News.