[Samsung AI Forum 2021] Advancing AI Technologies That Can Help Humankind – Samsung Global Newsroom

From November 1–2, Samsung Electronics held its fifth Samsung AI Forum (SAIF) completely on-line. The occasion introduced world-renowned lecturers and AI specialists collectively to debate and set up analysis instructions for growing AI that may be scaled to learn humanity.
 
Speakers representing varied fields launched newly developed AI algorithms, in addition to progressive AI options that may profit our lives sooner or later. Samsung Electronics livestreamed this yr’s discussion board on its YouTube channel and provided members – which included engineers, researchers and college students within the area of AI – the chance to work together with specialists throughout a Q&A session.
 
Read on for Samsung Newsroom’s recap of the shows and key matters that took heart stage throughout the two-day occasion.
 
 
Samsung AI Forum Day One
Developing AI That Addresses Common Problems

 
Hosted by the Samsung Advanced Institute of Technology (SAIT), Samsung’s R&D hub devoted to cutting-edge future applied sciences, day one of many Samsung AI Forum started with opening remarks from Dr. Kinam Kim, Vice Chairman and CEO of Samsung Electronics. “Digital transformation has been accelerated in each business, to which information science and machine studying are important,” stated Dr. Kim. “We at Samsung are open to discussing methods to sort out essential, widespread issues with researchers from all around the world, and we hope that the Samsung AI Forum may help facilitate that purpose.”
 

 
This was adopted by a keynote speech from Professor Yoshua Bengio of the University of Montreal, the co-chair of the Samsung AI Forum and a Samsung AI Professor. During his speech, Professor Bengio launched a brand new machine studying software referred to as GFlowNets.
 
After explaining how the algorithms might be utilized to the event of recent medicine, he emphasised how “We discover that [the model] converges to good options sooner than different strategies, and as well as, it finds a extra various set of options. So that is very encouraging, and we’re very excited in regards to the potential functions in discovery normally.” After ending his speech, the professor mentioned methods to use the algorithms throughout a Q&A session that featured scientists from all over the world.
 
The keynote was adopted by three know-how periods entitled “Scalable & Sustainable AI Computing”, “AI for Scientific Discovery” and “Trustworthy Computer Vision”. During these periods, main lecturers and startups spoke alongside a few of Samsung’s high researchers.
 
Professors Kunle Olukotun of Stanford University, Gerbrand Ceder of the University of California – Berkeley and Antonio Torralba of the Massachusetts Institute of Technology shared key findings of their respective areas of AI analysis. Founders of startups primarily based in Silicon Valley, together with Andrew Feldman, CEO of Cerebras Systems, Bryce Meredig, CSO of Citrine Informatics and Daniel Bibireata, Vice President of Landing AI, introduced insights on enterprise fashions for varied areas of AI analysis, in addition to future enterprise methods. Representing Samsung have been a number of main researchers, together with Changkyu Choi, Senior Vice President and Head of SAIT’s AI & SW Research Center, who launched the corporate’s imaginative and prescient for AI and summarized the progress it has made by way of its analysis within the area.
 

 
The occasion additionally showcased rising abilities and researchers within the area of AI. Samsung revealed this yr’s 5 winners of the Samsung AI Researcher of the Year award, which was launched final yr to acknowledge promising world AI researchers.
 
“I’m particularly grateful to my college students, whose work is actually what’s being rewarded right here,” stated Professor Phillip Isola of the Massachusetts Institute of Technology, who acquired the award. “We’re making an attempt to make progress to make AI methods which can be nearer to [reaching] human-like [and] animal-like skills,” he added, describing pure intelligence.
 
“My analysis lies on the intersection of laptop imaginative and prescient and machine studying, and my total purpose is to create imaginative and prescient methods which can be dependable and accessible for everybody,” added Professor Judy Hoffman of the Georgia Institute of Technology.
 
 
For the Coexistence of Humans and AI
Day one of many discussion board closed with a panel dialogue by which lecturers engaged in vigorous conversations and shared their insights. The panel’s moderator, Youngsang Choi, Vice President of SAIT, launched matters associated to every panelist’s space of experience. After the dialogue, members got free rein to ask the panelists questions.
 

 

 
One participant requested the panelists in the event that they believed that it could be potential for AI algorithms to realize human-level information effectivity in coaching, to which Professor Antonio Torralba stated sure.
 
“When we take into consideration the information that people have, it’s not simply visible information. They actually sense the world by way of plenty of completely different mechanisms,” Professor Torralba defined. “Also, people really aren’t passive observers of the world. They are literally interacting with the world and performing every kind of experiments. I believe, with the intention to obtain [a human-like] degree of effectivity, we have to incorporate all of these items and make them actually like the primary characters of the film that AI is enjoying now.”
 
The panel dialogue additionally provided a chance for college students majoring in AI-related fields to share their issues with the specialists and obtain recommendation. In the sphere of AI pure language processing (NLP), for instance, the variety of parameters is repeatedly rising, which implies that the prices required to coach a mannequin are too. Considering these circumstances, members mentioned which method tutorial analysis needs to be heading.
 
Professor Bengio concluded the panel dialogue by providing some insightful recommendation to younger AI researchers and college students. “Don’t be afraid to go in instructions which can be very completely different from what has been established as cutting-edge,” stated the professor. “Brain energy is the factor that’s actually behind innovation and [the] superb progress that science brings us. So don’t be afraid to attempt issues [and] don’t be afraid to query what has been apparently established for years or many years. That’s how we’re all going to make progress.”
 
 
Samsung AI Forum Day Two
The Latest AI Research, All in One Place

 
Day two of the discussion board was hosted by Samsung Research, Samsung Electronics’ superior R&D hub, which leads the event of future applied sciences for its Consumer Electronics and IT & Mobile Communications divisions. Dr. Sebastian Seung, President and Head of Samsung Research, emphasised that “AI is a know-how that makes individuals’s lives higher,” and provided an summary of the varied AI-related tasks that Samsung Research was engaged in, together with these associated to smartphone cameras, on-device AI, Open Source AI System Software, Machine Translation, and AI applied sciences for robots. “I’m actually wanting ahead to at present’s lectures by main researchers in AI,” stated Dr. Seung, heightening viewers’ expectations.
 

 
The day started with a keynote from Professor Leslie Valiant of Harvard University, who provided particulars on methods to increase supervised studying with reasoning. “To make AI work, it takes a number of parts,” Professor Valiant defined. “The first element is figuring out which phenomenon or performance you wish to notice.”
 
Next got here lectures delivered by lecturers who’ve been actively main AI analysis. These embrace Professor Felix Heide of Princeton University, Research Scientist Been Kim of Google Brain and Professor Max Welling, a analysis chair in machine studying on the University of Amsterdam and a Distinguished Scientist at Microsoft Research.
 
 
AI’s Evolution Into a Tool for Gaining Insights

 
Day two’s panel dialogue noticed specialists share their opinions on how AI know-how will influence individuals’s lives sooner or later. The panel’s moderator, Dr. Daniel D. Lee, Executive Vice President and Head of Samsung Research’s Global AI Center, kicked off the dialogue with a query.
 
“The first time AI got here into presence, there was plenty of emphasis at the moment on logical reasoning,” Dr. Lee defined. “But now, data-driven approaches comparable to deep neural networks are rising. And what we simply heard from Leslie’s speak was [about] how we will really use the logic [now] together with these extra superior neural community methods. What could be the large benefit of doing that sort of return, in some sense, to logic with neural networks?”
 
“The concept that each studying and logic are essential has been understood for a very long time, [albeit separately],” Professor Valiant defined. “We are in a great place as a result of I believe the place of studying is now very a lot superior. So, we now have cause to be assured that there’s plenty of competence that we now have so far as studying, and it’s a great foundation on which to construct logic.”
 
Researcher Efi Tsamoura of the Samsung AI Center in Cambridge added that “An rising variety of functions for a lot of completely different areas, from laptop imaginative and prescient to pure language processing, are profiting from background data with the intention to construct extra sturdy and less complicated fashions. Why is that? It’s as a result of logic gives us with the flexibility to [complement] lacking labels and to make use of the lacking labels with the intention to practice the mannequin.” Tsamoura additionally identified that “An rising variety of researchers from completely different fields, principally utilized fields, are realizing the potential of logic.”
 

 
The dialogue additionally touched on scientific discoveries which were made with machine studying. “I believe it’s value noting that with the appearance of 5G and with 6G coming down the pipe, communication networks have gone from being extraordinarily sophisticated to super-extremely sophisticated… and I believe the alternatives to optimize and handle the methods to make them much more environment friendly are huge. So I believe there’s a nice probability to deliver machine studying and AI instruments to bear on the construction and operation of those communication networks to make them extra environment friendly,” stated Gregory Dudek, Head of the Samsung AI Center in Montreal. “We’ve had some very good success in Montreal in adapting the instruments that exist to those comparatively new issues for that area, and [have] really considerably moved the needle to extend the efficiency of those methods.”
 
In order to commercialize machine studying to be used in varied areas, steady simulations should be carried out. How then can the hole between simulation outcomes and real-world phenomena be narrowed? Professor Welling shared his ideas: “[Since simulations do not actually reflect all the complexities of the world,] I believe most likely the answer is a few hybrid resolution the place you’d simulate as a lot as you possibly can, however you additionally establish the place your system is unsure about its predictions. And at that time, kind of in an energetic sense, you’re then going to amass information for that individual drawback. So, energetic sensing may be an attention-grabbing resolution.”
 
At the Lightning Talks session, workers from Samsung Research’s Global AI Centers introduced a few of their newest analysis together with Adaptive Sharpness-Aware Minimization(ASAM), which is a deep studying optimizer developed by Samsung Research, and Named Entity Correction for Automatic Speech Recognition (ASR).
 
The ideas and findings that have been shared on the Samsung AI Forum point out {that a} world by which AI is merged seamlessly with our each day lives is probably not that far off. Full replays of each days of the Samsung AI Forum 2021, by way of which viewers can be taught extra in regards to the present standing of AI know-how, its functions, and what the long run could maintain, could be considered on the occasion’s official web site and on Samsung Electronics’ YouTube channel.

Recommended For You