This year’s AI Forum, the fourth of its kind, is being held over two days this November 2 and 3. The first day of the event, hosted by the Samsung Advanced Institute of Technology (SAIT), Samsung’s R&D hub dedicated to cutting-edge future technologies, is enabling participants to facilitate discussions around how to make the best use of AI technologies in a way that can benefit our daily lives in a rapidly changing world, particularly within the context of the unprecedented situations that have arisen recently due to the global pandemic.

AI Forum Day 1: The Past, Present and Future of AI

On November 2, Dr. Kinam Kim, Vice Chairman & CEO of Device Solutions at Samsung Electronics, commemorated the start of the first day of the AI Forum 2020 by delivering an opening speech that highlighted how AI technologies have shown remarkable progress over the years. He went on to note that, given these changes, many are expecting AI to address the issues brought on by the recent pandemic, but highlighted that since AI bases its models on massive amounts of real-life data and simulations, the task of modeling the current pandemic and other natural disasters with AI was a daunting one.

Dr. Kim went on to provide his own views on the ways in which AI technologies can move forward and be harnessed to have meaningful impact on real world problems, and also highlighted that Samsung Electronics, as a major provider of core technologies in the AI ecosystem, is proactively co-operating with global researchers to seek solutions to such real world problems. Dr. Kim ended his opening speech with the expectation that meaningful discussions on the present and future of AI technologies and their benefit for humanity were set to take place during this year’s Forum.

Recognizing Leading Talent in the Field

At this year’s AI Forum, Samsung introduced their inaugural Samsung AI Researcher of the Year awards with the view to identify prominent emerging researchers in the field from around the world and to support their research activities.

This year’s Samsung AI Research of the Year awards went to Professor Kyunghyun Cho of New York University, Professor Chelsea Finn of Stanford University, Professor Seth Flaxman of Imperial College London, Professor Jiajun Wu of Stanford University and Professor Cho-Jui Hsieh of UCLA.

Professor Kyunghyun Cho, a globally recognized researcher in natural language processing, has been publishing a consistent stream of acclaimed papers across the medicine, biology and optimization disciplines. “I am honored to have received a Samsung AI Researcher of the Year award and am committed to developing AI-focused research further down the road,” said Professor Cho of the recognition.

Expert Highlights: Keynote Speeches

Professor Yoshua Bengio, who served as this year’s co-chair and was selected as Samsung AI Professor of the Year, gave a presentation titled Towards Discovering Casual Representations. In his lecture, Professor Bengio explained that, up until now, conventional deep learning technologies have been relying on inference to recognize sensual information and learn from it, but AI technologies that are instead capable of learning the causality between hidden variables before drawing conclusions could be capable of making inferences just as humans do, and hence would be able to respond to unprogrammed situations. With visions of such a type of AI in mind, Professor Bengio shared the initial outcomes of his research and suggested how, based on this, AI technologies can make steps forward.

Professor Yann LeCun of New York University, a researcher who pioneered the Convolutional Neural Network widely applied to video recognition technologies, presented his latest model related to Self-Supervised Learning. Unlike supervised learning which returns a given answer to each given data set, self-supervised learning adopts a learning model consisting of autonomously creating questions within data and subsequently finding answers. Such a method has been applied to a massive linguistic model capable of generating sentences just as people do. Professor LeCun highlighted how self-supervised learning is similar to the way children experience and learn the world, and presented an energy-based model based on such a comparison.

Professor Chelsea Finn of Stanford University, a young researcher in the spotlight within the field of meta learning, gave a lecture titled From Few-Shot Adaptation to Uncovering Symmetries. In her lecture, Professor Finn introduced meta learning technologies in which AI, in spite of changes in data, can adapt swiftly to untrained data, and proceeded to share success stories of the application of these technologies in the areas of robotics and new drug candidate material design.

Professor Donhee Ham, Fellow at the Samsung Advanced Institute of Technology and Professor at Harvard University, delivered a presentation titled Reconstruction of the Brain. In his presentation, he highlighted that the current level of AI is based on the human brain but in fact works in a way different from how the brain functions, causing limitations to its capability. Professor Ham introduced cutting-edge neural science technologies that could mimic the structure and functionalities of the human brain circuit and create computer integrated circuits on their own.

Industry experts also took part in giving presentations. Dr. Tara Sainath of Google Research released the latest research outcomes of end-to-end models developed for speech recognition capable of enhancing the accuracy, efficiency and multi-lingual capability of voice assistant services widely available across smart devices.

Dr. Jennifer Wortman Vaughan of Microsoft Research gave a lecture titled Intelligibility Throughout the Machine Learning Life Cycle. She shared a human-centric machine learning concept, highlighting that, in order to develop a fair machine learning system capable of garnering the trust of people, people’s clear understanding of the system is required. Dr. Wortman Vaughan then introduced research outcomes that can objectively verify such a mechanism.

Since the Samsung AI Forum 2020 was held virtually this year, students and researchers alike in the AI research field from all over the world were able to engage in online discussions and exchanges. When tuning in to the Forum’s lectures on Samsung Electronics’ YouTube channel, attendees could ask questions to and receive answers from the distinguished speakers thanks to a real-time chat functionality.

Stay tuned to Samsung Newsroom for more information on the Samsung AI Forum 2020.

Since blue is known to be the most difficult color to implement out of the three primary QLED colors (red, blue and green), this achievement – coming in the wake of Samsung’s development of red QLED technology last November – once again proves Samsung’s excellence in the quantum dot technology sphere.

Blue Proves the Most Difficult of the Three Primary QLED Colors

Quantum dots (QDs) are semiconductor particles that measure a few nanometers in diameter (tens of thousands of times narrower than a single human hair). When illuminated, they re-emit light of a certain color depending on their size.

The blue QD, which has the largest band gap¹ among the three primary colors, rapidly oxidizes upon exposure to external light, resulting in a short lifespan and low luminous efficiency.² For this reason, up to now the industry had failed to develop even the technology required for blue quantum dot light-emitting diodes.

Overcoming Another Challenge by Developing Blue QLED Technology

But now, SAIT has successfully developed blue QLED technology, achieving industry-leading results such as 20.2% improved luminous efficiency, 88,900 nits of maximum luminance and 16,000 hours of QLED lifetime (measured at half-brightness for 100-nit luminance). These results were recorded in a study titled “Efficient and stable blue quantum dot light-emitting diode”, which was published by the journal Nature on October 14, 2020.

Eunjoo Jang, Samsung Fellow

“Samsung’s distinctive quantum dot technology has once again overcome the limitations of existing technology in the industry,” noted Dr. Eunjoo Jang, Samsung Fellow and corresponding author for the study. “I hope that this study goes on to help accelerate the commercialization of Quantum Dot light-emitting diodes (QLEDs).”

Quantum dots are made up of a basic structure of a core, a shell, and multiple ligands.³ In order to better stabilize the QD materials and secure durable photoresponse functionality, the researchers applied a structure with quantum dot double emitting layers and shorter ligands on the surface of the blue-light-emitting QDs while also improving current injection rates.

Taehyung Kim, Principal Researcher

Dr. Taehyung Kim, Principal Researcher and the first author of the study, said, “This research is meaningful in that we have not only established Quantum Dot light-emitting diode performance, but have also proven that the technology can deliver top-notch performance at the element level.”

(From left) Kwang-Hee Kim, Taehyung Kim, Eunjoo Jang, Sungwoo Kim, Seon-Myeong Choi from SAIT

¹ The difference of energy between the valence band of electrons and the conduction band.

² The ratio of the emitting luminous flux to the total input flux of source.

³ The core absorbs and re-emits light, while the shell layer surrounding the QD core improves lifespan and photoluminescence efficiency by preventing temperature/humidity-related damage. The branch-shaped ligands are located on the surface of the QD’s shell and help maintain inter-particle distance.

Samsung Electronics last month announced its goal to strengthen its leadership in the global system semiconductor industry by 2030 through expanding its proprietary NPU technology development. The company recently delivered an update to this goal at the conference on Computer Vision and Pattern Recognition (CVPR), one of the top academic conferences in computer vision fields.

This update is the company’s development of its On-Device AI lightweight algorithm, introduced at CVPR with a paper titled “Learning to Quantize Deep Networks by Optimizing Quantization Intervals With Task Loss”. On-Device AI technologies directly compute and process data from within the device itself. Over 4 times lighter and 8 times faster than existing algorithms, Samsung’s latest algorithm solution is dramatically improved from previous solutions and has been evaluated to be key to solving potential issues for low-power, high-speed computations.

Streamlining the Deep Learning Process

Samsung Advanced Institute of Technology (SAIT) has announced that they have successfully developed On-Device AI lightweight technology that performs computations 8 times faster than the existing 32-bit deep learning data for servers. By adjusting the data into groups of under 4 bits while maintaining accurate data recognition, this method of deep learning algorithm processing is simultaneously much faster and much more energy efficient than existing solutions.

Samsung’s new On-Device AI processing technology determines the intervals of the significant data that influence overall deep learning performance through ‘learning’. This ‘Quantization¹ Interval Learning (QIL)’ retains data accuracy by re-organizing the data to be presented in bits smaller than their existing size. SAIT ran experiments that successfully demonstrated how the quantization of an in-server deep learning algorithm in 32 bit intervals provided higher accuracy than other existing solutions when computed into levels of less than 4 bits.

When the data of a deep learning computation is presented in bit groups lower than 4 bits, computations of ‘and’ and ‘or’ are allowed, on top of the simpler arithmetic calculations of addition and multiplication. This means that the computation results using the QIL process can achieve the same results as existing processes can while using 1/40 to 1/120 fewer transistors².

As this system therefore requires less hardware and less electricity, it can be mounted directly in-device at the place where the data for an image or fingerprint sensor is being obtained, ahead of transmitting the processed data on to the necessary end points.

The Future of AI Processing and Deep Learning

This technology will help develop Samsung’s system semiconductor capacity as well as strengthening one of the core technologies of the AI era – On-Device AI processing. Differing from AI services that use cloud servers, On-Device AI technologies directly compute data all from within the device itself.

On-Device AI technology can reduce the cost of cloud construction for AI operations since it operates on its own and provides quick and stable performance for use cases such as virtual reality and autonomous driving. Furthermore, On-Device AI technology can save personal biometric information used for device authentication, such as fingerprint, iris and face scans, onto mobile devices safely.

“Ultimately, in the future we will live in a world where all devices and sensor-based technologies are powered by AI,” noted Chang-Kyu Choi, Vice President and head of Computer Vision Lab of SAIT. “Samsung’s On-Device AI technologies are lower-power, higher-speed solutions for deep learning that will pave the way to this future. They are set to expand the memory, processor and sensor market, as well as other next-generation system semiconductor markets.”

A core feature of On-Device AI technology is its ability to compute large amounts of data at a high speed without consuming excessive amounts of electricity. Samsung’s first solution to this end was the Exynos 9 (9820), introduced last year, which featured a proprietary Samsung NPU inside the mobile System on Chip (SoC). This product allows mobile devices to perform AI computations independent of any external cloud server.

Many companies are turning their attention to On-Device AI technology. Samsung Electronics plans to enhance and extend its AI technology leadership by applying this algorithm not only to mobile SoC, but also to memory and sensor solutions in the near future.

Four individuals who played key roles in developing Samsung’s On-Device AI Lightweight Algorithm. From Left to right; Jae-Joon Han, Chang-Young Son, Sang-Il Jung, Chang-Kyu Choi of Samsung Advanced Institute of Technology

1 Quantization is the process of decreasing the number of bits in data by binning the given data into sections of limited number levels, which can be represented in certain bit values and are regarded as having the same value per section

² Transistors are devices that control the flow of current or voltage in a semiconductor by acting as amplifiers or switches