Many asylum seekers face long waits in Mexico due to a shortage of translators for Indigenous language speakers. Photo/iStock.

Imagine fleeing persecution at home, surviving a difficult journey, arriving in a new country to seek asylum, only to be turned away at the border because nobody speaks your language. This is the reality for hundreds of migrants coming into the United States from remote areas of Central America who do not speak common languages, such as Spanish or Portuguese.

A shortage of translators for Indigenous asylum seekers speaking traditional languages means many must wait for months or even years in Mexico to apply for asylum, creating a long backlog in an already overwhelmed immigration system.

Katy Felkner is developing a machine translation system for Mexican and Central American Indigenous languages to help asylum seekers at the border. Photo/Katy Felkner.

The U.S. immigration system is set up to handle English and Spanish, said Katy Felkner, a Ph.D. student in computer science at the USC Viterbi School of Engineering, but there are several hundreds of people a year who are minority language speakers, in particular, speaking Indigenous languages from Mexico and Central America, who are not able to access any of the resources and legal aid that exists for Spanish-speaking migrants.

In other cases, people are unable to explain the threats to their lives in their hometowns, which could be the basis for asylum. When migrants cannot understand or be understood, there is no way to establish the threat to their safety during a credible fear interview conducted by the U.S. Department of Homeland Security.

The statistics are staggering: asylum-seeking immigrants without a lawyer prevailed in only 13 percent of their cases, while those with a lawyer prevailed in 74 percent of their cases, according to a study in the Fordham Law Review.

Felkner, who conducts her research at the USC Information Sciences Institute (ISI) under Jonathan May, a research associate professor, is working on developing a solution: a machine translation system for Mexican and Central American Indigenous languages that can be used by organizations providing legal aid to refugees and asylum-seekers.

People are being directly adversely impacted because there arent interpreters available for their languages in legal aid organizations, said Felkner. This is a concrete and immediate way that we can use natural language processing for social good.

People are being directly adversely impacted because there arent interpreters available for their languages in legal aid organizations. Katy Felkner.

Felkner is currently working on a system for Kiche, a Guatemalan language, which is one of the 25 most common languages spoken in immigration court in recent years, according to The New York Times.

Were trying to provide a rough translation system to allow nonprofits and NGOs that dont have the resources to hire interpreters to provide some level of legal assistance and give asylum seekers a fair chance to get through that credible fear interview, said Felkner.

Felkners interest in languages began during her undergraduate degree at the University of Oklahoma, where she earned a dual degree in computer science and letters, with a focus on Latin. During her first year of college, she worked on a project called the Digital Latin Library, writing Python code to create digital versions of ancient texts.

Thats what got me thinking about language technology, said Felkner. I taught myself some basics of natural language processing and ended up focusing on machine translation because I think its one of the areas with the most immediate human impact, and also one of the most difficult problems in this area.

While Felkner and May are currently focused on developing a text-to-text translator, the end goal, years from now, is a multilingual speech-to-speech translation system: the lawyer would speak English or Spanish, and the system would automatically translate into the asylum seekers Indigenous language, and vice-versa.

Translation systems are trained using parallel data: in other words, they learn from seeing translation pairs, or the same text in both languages, at the sentence level. But there is very little parallel data in Indigenous languages, including Kiche, despite it being spoken by around one million people.

Thats because parallel data only exists when there is a compelling reason to translate into or out of that language. Essentially, said Felkner, if its commercially viableDisney dubbing films from English to Spanish, for instanceor stemming from a religious motivation.

In many cases, due to the influence of missionaries throughout Latin America, the only parallel data sourcethe same text in both languagesis the Bible, which doesnt give researchers much to work with.

Were really trying to push the lower bound on how little data you can have to successfully train a machine translation system. Katy Felkner.

Imagine youre an English speaker trying to learn Spanish, but the only Spanish youre ever allowed to see is the New Testament, said Felkner. It would be quite difficult.

Thats bad news for the data-hungry deep learning models used by language translation systems that take a quantity over quality approach.

The models have to see a word, phrase, grammatical construction a bunch of times to see where its likely to occur and what it corresponds to in the other language, said Felkner. But we dont have this for Kiche and other extremely low resource Indigenous languages.

The numbers speak for themselves. From English to Kiche, Felkner has roughly 15,000 sentences of parallel data, and 8,000 sentences for Spanish to Kiche. By contrast, the Spanish to English model she trained for some baseline work had 13 million sentences of training data.

Were trying to work with essentially no data, said Felkner. And this is the case for pretty much all low-resource languages, even more so in the Americas.

One tactic in existing low-resource work uses closely related, higher resource languages as a starting point: for instance, to translate from English into Romanian, you would start training the model in Spanish.

But since Indigenous languages of the Americas developed separately from Europe and Asia, the majority are low resource, and most of them are extremely low resource, a term Felkner coined to describe a language with less than around 30,000 sentences of parallel data.

Were really trying to push the lower bound on how little data you can have to successfully train a machine translation system, said Felkner.

But Felkner, with her background in linguistics, was undeterred. Over the past two years, she has worked on creating language data for the models using some tricks of the trade in natural language processing.

One tactic involves teaching the model to complete the abstract task of translation and then setting it to work on the specific language in question. Its the same principle as learning to drive a bus by learning to drive a car first, said Felkner.

To do this, Felkner took an English to Spanish model, and then fine-tuned it for Kiche to Spanish. It turned out, this approach, called transfer learning, showed promise even in an extremely low resource case. That was very exciting, said Felkner. The transfer learning approach and pre-training from a not-closely-related language had never really been tested in this extremely low resource environment, and I found that it worked.

She also tapped into another resource: using grammar books published by field linguists in the mid-to-late 70s to generate plausible synthetic data that can be used to help the models learn. Felkner is using the grammar books to write rules that will help her construct syntactically correct sentences from the dictionaries. The technical term for this is bootstrapping or data augmentation or colloquially, fake it til you make it.

We use this as pre-training data, to essentially teach the models the basics of grammar, said Felkner. Then, we can save our real data, such as the Bible parallel data, for the fine-tuning period when it will learn whats semantically meaningful, or what actually makes sense.

Finally, shes testing a technique that involves parsing nouns in the English and Kiche sides of the Bible, replacing them with other nouns, and then using a set of rules to correctly inflect the sentences for grammar.

For example, if the training data has the sentence: the boy kicked the ball, the researchers could use this approach to generate sentences like the girl kicked the ball, the doctor kicked the ball, the teacher kicked the ball, which can all become training data.

The idea is to use these synthetically-generated examples to essentially build a rough version of the system, so that we can get a lot of use out of the small amount of real data that we do have, and finetune it to exactly where we want it to be, said Felkner.

Working in extremely low-resource language translation is not easy, and it can be frustrating at times, admits Felkner. But the challenge, and the potential to change lives, drive her to succeed. Her work is being noticed, too: she was recently awarded a National Science Foundation Graduate Research Fellowship to continue working on the border translation project.

Within the next year, she plans to undertake a field trip to observe how legal aid organizations are working at the border, and where her system could fit into their workflow. She is also working on a demo website for the system, which she hopes to unveil in 2023, and once developed, she hopes the system could one day be applied to other Indigenous languages.

Hill climbing on high resource languages can make your Alexa, Google Home or Siri understand you better, but its not transformative in the same way, said Felkner. Im doing this work because it has an immediate humanitarian impact. As JFK once said, we choose to go to the moon not because it is easy, but because it is hard. I often think the things that are worth doing are difficult.

Published on August 24th, 2022

Last updated on August 24th, 2022

Continue reading here:

Lost in Translation at the Border - USC Viterbi | School of Engineering - USC Viterbi School of Engineering

SACRAMENTO Governor Gavin Newsom today announced that the stateis awarding$54 million in grants to Los Angeles County, the Inland Empire, and the border region to strengthen the K-16 education-to-career pipeline. The grants are part of theK-16 Education Collaboratives Grant Program, which is a $250 million investment providing new pathways to career opportunities and addressing long-standing equity gaps for students in their local communities.

California is preparing the next generation for the jobs of the future, said Governor Gavin Newsom. Were closing equity gaps, providing more resources to help our students achieve their career goals right in their own communities, and streamlining the pipeline from K-12 to higher education to careers.

The program, administered by the Department of General Services (DGS), Office of Public School Construction, and Foundation for California Community Colleges, provides funding to enhance or create collaborative efforts between the University of California system, the California State University system, Community Colleges, K-12 School Districts, and workforce partners. This is a key component of a statewide strategy for cultivating regional economies and ensuring that education, vocational, and workforce programs work together to strengthen education and employment opportunities.

The Department of General Services is pleased to continue funding efforts for the first phase of this innovative program and is excited to see the work that will be accomplished in these newly awarded regions, said DGS Director Ana Lasso. There is a tremendous need statewide for the role the collaboratives will fill in working to transform the public education system and meeting the needs of regional workforces while ensuring equitable pathways to meaningful careers for all learners of California. We are looking forward to rolling out the second phase of funding for the program soon, which will provide additional opportunities to expand this program throughout all regions of the state.

Approximately $18 million will be awarded to each of the following collaboratives:

Los Angeles County: The Los Angeles Region K-16 Collaborative aims to improve enrollment, persistence and degree completion of underrepresented students in fields that lead to increased economic mobility, starting with streamlined pathways to health care, engineering, and computer science careers. Five sub-regional partnerships anchored around five CSU campuses (Dominguez Hills, Los Angeles, Long Beach, Northridge, and Pomona) and their feeder K-12 and community colleges will work collectively to enhance dual enrollment, transfer pathways, and work-based learning in the region.

Border Region: The Border Region Inclusive Talent Pipeline Collaborative brings together partners California State University San Marcos, San Diego State University, UC San Diego, Imperial County and San Diego County Offices of Education committed to improving the K-16 education system in the Border Region. The collaborative aims to strengthen pathways to build a representative talent pipeline in business, tech, health, and education sectors.

Inland Empire: Leveraging the experience and success of two major collaboratives in the region, the Inland Empire Collaborative (IEC) will deploy a regional framework that incorporates educational, community, workforce, and employer engagement in San Bernardino and Riverside Counties to support equitable educational and economic opportunities for students and address equity barriers in career and technical education pathways in the areas of Healthcare, Engineering, and Computing, Business, and Education.

Summaries of each collaborative can be found on the Regional K-16 Education Collaboratives Grant Programwebsite.

Earlier this year, the Central San Joaquin Valley, North State, Kern County, Redwood Coast, Orange County, and Sacramento regionswere awarded grants by the program. Summaries of each grant can be foundhere.

The second phase of funding available to regions that require additional time and planning to establish collaborative partners or to determine their program goals will be launching soon. Information about the second phase of funding can be foundhere.

###

Read more:

California Distributes $54 Million in Grants to Bolster the Education-to-Career Pipeline California Governor - Office of Governor Gavin Newsom

Published 08-24-22

Submitted by Bristol-Myers Squibb Company

Originally published on Bristol Myers Squibb News & Perspectives

Scientists and experts at Bristol Myers Squibb are working to revolutionize the process of drug development with sophisticated scientific strategies driven by technology. Using computer science, advanced clinical trial solutions and robust digital platforms, the company is accelerating and evolving its pipeline to get therapies to patients faster. But merely having these data and capabilities is not what propels progress more important are the curiosity and questions that spur new discoveries, drive predictive analysis and yield meaningful conclusions from experts who ultimately transform data into innovation.

A strong foundation built over decades

Bristol Myers Squibb has deep biological knowledge of key pathways implicated in diseases and of pioneering treatment platforms and modalities. Harnessing this legacy, the company is leveraging past learnings in combination with robust novel datasets from both internal and external sources to amplify predictive power throughout all stages of the research and development (R&D) pipeline.

This library is built from vast amounts of data that has not only been collected over the years, but also interrogated in innovative ways. We have molecular, translational, imaging, clinical, wearable and real-world data that, like pieces of a puzzle, create a picture that advances our own understanding of the biology of the disease. Coupled with predictive machine learning methods, data helps us ask and answer questions, see the reality of the disease, the drug, and the types of patients we seek to help. They say every revolution in art is a return to realism. I think the greatest gift data science can bring us is a closer look at reality through the data.

Just as important as the actual data is the integrity of that data. This is key as the data serve as the foundation upon which future efforts will be undertaken. We must ensure that data are accurate, reliable and free of bias. We must also ensure that the data are collected and used in an ethical, compliant way that respects the rights of others and maintains patient privacy.

There are a lot of people involved in generating reliable data, and reliable data take more than one form, said Matthew Trotter, PhD, vice president, Predictive Sciences. There is literal reliability: have we captured it correctly? Has it become corrupted at any point? There is also reliability from an application perspective: does it match the research objective we are pursuing? We can generate a lot of data, but if we don't do so while asking the right questions, then both the data and our work with it may be less impactful than they could be.

While our key datasets serve as the foundation for generating insights, it is the activities that surround the data and how scientists use them that drive decisions and enable new discoveries and advancements.

The key to unlocking true innovation

Much of the complex data generated by experiments, drug development or patient experiences do not, to the casual or even trained observer, yield immediate insights. Patterns and answers do not simply jump out for the taking but rather are elucidated via further analysis, interpretation and critical thinking. What are the questions that must be answered? What are the challenges that must be solved? Knowing what to ask and how to leverage answers and make predictions derived from identifying patterns in the data is the key that unlocks the true potential of data and turns them into knowledge that can inform decisions that advance the development of novel therapies. For example, if new, validated and precise clinical trial endpoints can be developed using data, then there is an opportunity to accelerate the path to approval for new therapeutics, thereby giving patients quicker access to medication.

At Bristol Myers Squibb, teams of interdisciplinary scientists contribute to the generation, understanding and impactful use of data. One common misconception is that as capabilities in analyzing complex datasets grow, the need for experts shrinks; this is far from the truth. Advancements in informatics do not remove the need for human expertise. In fact, they broaden the disciplines that have traditionally been involved in drug discovery and development.

All of our work is truly cross-functional, with everybody bringing their subject matter expertise together, said Micsinai-Balan. Biostatisticians, computer scientists, development teams, machine learning experts, safety colleagues, clinical operations, information technology (IT), biologists, chemists and clinical scientists, to name a few, all work together to not only generate usable data but, more importantly, to test hypotheses and figure out how we can do better for the patients we serve. We not only complement each other but learn from each other and help each other get the full picture.

Within the company, these experts employ sophisticated technologies, such as artificial intelligence (AI) or machine learning, to produce a 360-degree view of specific diseases and specific patient types and to answer key questions from early research through treatment availability. This includes clinical trial recruitment, where both data and technology can be leveraged to help develop ways to recruit a more diverse and representative patient group compared to traditional approaches.

In addition to in-house experts, strategic collaborations help Bristol Myers Squibb achieve its goal of building industry-leading data science capabilities to drive advancements across the drug development continuum. We have a lot of great, innovative collaborations in this space, which allow us to leverage our combined capabilities and share knowledge towards scientific progress that goes beyond incremental improvements to deliver potentially transformational advances, said Trotter.

Key examples from the companys early pipeline research include collaborations with Exscientia to accelerate the discovery of small molecule drug candidates using AI in various therapeutic areas including oncology and immunology, and Insitro to create disease models and drive therapeutic discovery and development for neurodegenerative disorders using machine learning, human genetics and functional genomics.

In the clinical stage, the company is working in collaboration with Owkin to enhance clinical trial design and execution with AI-powered approaches that optimize endpoint definitions, patient subgroups and treatment effect estimation with covariate adjustment and external control arms. Bristol Myers Squibb will apply Owkins AI capabilities to design potentially more precise and efficient clinical trials, initially for the companys cardiovascular development program.

Illuminating new possibilities

These advances and experts come together to inform the ecosystem of drug development and help Bristol Myers Squibb achieve a deeper knowledge and understanding of the challenges that need to be solved. Together, data and powerful computers and algorithms, applied by teams of scientific and technological experts, help lead the way for the R&D organization from discovery to approval.

Most importantly, these advancements come together in the hope of accelerating the preclinical development timeline, clinical trials and overall time to approval.

At the end of the day, the goal is to bring a successful and safe therapeutic to patients faster. Part of achieving this relies on the ability to conduct clinical trials in a timely, smart and efficient manner, and we rely on data, our experts and strategic collaborations to do so, said Micsinai-Balan. It is not only that science generates data, but data gives feedback to science and helps us search in novel directions.

Never finished: How continual learning helps drive patient outcomes

Efforts to glean knowledge from data do not stop once a therapeutic is approved but rather shift in focus and priority. Data and the insights they provide can help improve how therapeutics are brought to patients. Bristol Myers Squibb works with real-world anonymized patient data to understand the patient journey, identify unmet needs and optimize actions both within the context of the approved medicine and within the pipeline.

At this stage, feedback from real-world experience with a medicine can help inform actions taken by the teams working directly with doctors and patients by providing useful information to the appropriate patients at the right point in their journey. Data collected at this stage can also help inform health economics and outcomes research (HEOR) efforts, which complement the information gained from the clinical development stage to guide decisions on patient access to medications.

These efforts include close collaboration with many departments at Bristol Myers Squibb, including R&D, as well as with academic and scientific partners throughout the industry, all working together toward a collective goal improving patient care.

Innovation for a powerful purpose

As rapid technological advancements continue, innovations in computing, machine learning and AI are part of our science and, applied to our data, will be foundational to discovering, developing and delivering innovative medicines that help patients prevail over serious diseases. Bristol Myers Squibbs commitment to being at the forefront of revolutionizing drug development through such innovation is evidenced by providing talented researchers and innovators the flexibility to drive R&D forward wherever the science takes them.

Learn more about Bristol Myers Squibb here

Bristol-Myers Squibb is a global biopharmaceutical company whose mission is to discover, develop and deliver innovative medicines that help patients prevail over serious diseases. For more information, please visitwww.bms.comor follow us on Twitter athttp://twitter.com/bmsnews.

More from Bristol-Myers Squibb Company

See original here:

CSRWire - From Numbers to Knowledge: Unlocking the Predictive Power of Data To Accelerate Transformative Science - CSRwire.com

A new semester means a new opportunity to join Pitts more than 400 student organizations anything from club sports to Knittsburgh, a knitting and crochet group.

With so many clubs on campus, students might find it hard to learn about them all and decide which ones to join. Pitts Student Organization Resource Center facilitates registration, manages business transactions and provides resources such as event planning and training. SORC is also hosting its annual activities fair on Aug. 28 outside the William Pitt Union from 12 p.m. to 2 p.m. and from 4 p.m. and 6 p.m. so students can find clubs that interest them.

For students interested in dancing, the Ballet Club at Pitt performs twice a year and offers a weekly masterclass for different dance styles and levels. Members performing in upcoming shows attend weekly rehearsals with a student choreographer.

Leah Mrozek, a senior marketing and supply chain management major and the clubs artistic vice president, said the club is for dancers, or those interested in dancing, who want to continue performing in college.

To join the club, you must fill out a form at the beginning of the semester to let us know about your dance experience, Mrozek said. The next step is to attend no-cut auditions so we can place each dancer in a piece. Join our email list to get the latest updates from the club!

Mrozek said the club will host a kickoff meeting at the beginning of the semester, along with a technical rehearsal and dress rehearsal later in the semester. An added perk, she said, is that club members can attend shows by the Pittsburgh Ballet Theater for a discounted price.

Pitt also has a variety of clubs for students looking to volunteer. Pitt Best Buddies, a chapter of a global nonprofit organization, pairs students with members of the Pittsburgh community with intellectual and developmental disabilities for one-on-one friendships. They also host events like trips to the Pittsburgh Zoo & PPG Aquarium and participate in fundraisers like the Best Buddies Friendship Walk.

Its main goal is to be inclusive. Pitt Best Buddies holds meetings at least once a month, and students can get involved by visiting our Instagram and emailing us to be put on the email list, said Maddie Kuzdzal, a junior biology major and vice president of the group.

Camp Kesem helps support Pittsburgh families with cancer patients through fundraising, outreach and volunteering. Pitts chapter also hosts a weeklong free summer camp for children.

Lauren Charlon, a senior neuroscience major and the clubs co-director, said general body members meet every two weeks. She said the coordinator board also is currently accepting applications for various committees.

Therapy Dogs, a new club at Pitt, raises money and resources for the Humane Animal Rescue of Pittsburgh and strives to be a place where all dog lovers can meet and connect. They also advocate for the continuation of Therapy Dog Tuesdays, when therapy dogs from the Western Pennsylvania Humane Society stop by the Cathedral of Learning at 7 p.m. to visit with students.

As a new club, we are testing out the frequency of meetings that work best for us. As of now, we are hoping to meet at least once a month. In these meetings, we plan to discuss our upcoming events that may include anything from service to fundraising opportunities, said Omkar Betsur, a junior biology major and the clubs president.

There are also groups that are more oriented to specific majors, such as the Computer Science club. The computer science club hosts socials and informational events to connect students with each other and industry professionals. They also host workshop events to assist students with assignments, resumes and applications.

Jamir Grier, a junior computer science major and club president, said prospective club members dont have to be a computer science major or pay dues to join. He said meetings are typically held on Wednesdays, while corporate events are held on Mondays. Hackathons are occasionally held on the weekends or over a span of time.

CSC aims to be the largest CS organization on campus and has seized that goal by working hard to guarantee that the merit of our events, programs and avenues of communication (i.e., discord, newsletter) are good enough to drive members to participate when and how they can, he said.

And if students cant find a club that fits their interests, they can form one if they have 10 students willing to join the group and can secure a full-time faculty or staff member as an advisor. The new organization registration period occurs from Oct. 1-15 and Feb. 1-15.

No matter what type of club a student joins, Barbara Kucinski, a lecturer and faculty advisor at Pitts psychology department, said they offer friendship and spaces to figure out your interests with a diverse group of people.

Student organizations are a good place for students to develop and hone communication and listening skills, engage with a diverse group of people and make friends, Kucinski said.

Ive known students who got internships, jobs and volunteer positions by attending group meetings and meeting people, she added. Those that go on to hold leadership positions in a club/organization learn how to work with a team (which employers value) and gain leadership and management skills. Furthermore, it is simply fun.

See the article here:

Simply fun: Students find interests, friends with Pitts more than 400 clubs - The Pitt News

Chris Ramming

Chris Ramming, Senior Director for Research and Innovation at VMWare, will be a speaker at theSymposium on the Future of Computing Researchheld by USCs Information Sciences Institute on September 12-13, 2022.

What topics should we be researching to reflect the changing role of computers in our society?

The last few years have shown how impactful computer science research can be on society, but the related innovations may result in dangers, risks and ethical considerations. I think the community is beginning to realize that those issues may need more attention. There may be ways to incorporate some of these broader considerations into our research agendas.

One of the issues that Im particularly interested in right now is privacy. Data is available on the internet in forms that cant be controlled by the people who originated it. Im personally interested in technologies that give us not just the ability to keep secrets which is the way I think about security but also give people control over who sees those secrets which is how I think about privacy engineering.

How do you think we should stimulate research in these areas?

Partnerships matter a lot. And partnerships between industry, academia and government can be very effective. People can do a lot more together than they can by themselves.

One of the things that people can do when theyre tackling an important area like this, is to just put down a position paper outlining the range of issues that need to be addressed. Which agencies or which communities need to address them? What might a nation-scale research agenda look like to comprehensively tackle one of these societal issues?

Should researchers without traditional computer science backgrounds, such as those with social science degrees, get involved in computer science research? What is the best way for them to get involved?

As technologists, we may not have the tools to understand human needs, desires, societal issues. Bringing in social scientists, ethnography researchers and user-experience researchers earlier, during the problem formulation and design stage of a research program, could be very valuable.

One approach could be to create advisory boards or ethics boards. You cant expect people to work on these problems full-time, and that may not even be necessary. There may be ways of having part-time roles, opening up opportunities to forge collaborations early in the life of a large program.

We also see a lot of places where computer science research can be beneficial in other disciplines such as biology, physics and energy. Collaboration across those kinds of fields can potentially be very valuable as we bring, for example, state of the art AI techniques to new problems in fusion energy. Thats a form of collaboration that is becoming increasingly commonplace, because computer science is so fundamental to many different industries today.

In the past, youve talked about how to foster a positive relationship between corporate and faculty researchers.How can we encourage this kind of research?

The key is to find people who are interested in tackling the same big goal and who value each others insights and experience, and believe that by working together, they can accomplish more. The best thing we can do is create opportunities to connect like-minded researchers.

What is the best way for companies and universities to plan an innovative portfolio of research?

At VMWare, we try to grow our company and enrich our portfolio of products and services in a number of different ways. There are times when we acquire companies, which is referred to as inorganic innovation. There are times when we build our own internal technology, which is organic innovation. We deliberately try to do both.

Academia, increasingly, is finding multiple ways to do research translation. In some cases, people are creating their own faculty-led startups. In other cases, they are letting companies take their inventions to the next level and innovate with them.

We have choices on the corporate side about how to balance organic and inorganic innovation. Researchers at universities have choices about how they would like to see their inventions, commercialized or brought into the real world. Understanding each others point of view is a critical first step in building a successful collaboration. All of those approaches work and make for a productive relationship. But its important to figure out early on which path youre planning to go down.

What is one outdated practice that you see a lot in your field that you believe is limiting innovation and that researchers should move past?

I often hear of people get hung up on intellectual property negotiations before they begin a university-industry collaboration. I think being too concerned about fully triaging intellectual property issues at the beginning of a relationship can slow down the launch of that relationship in ways that may not be necessary. Im encouraged that some leading research groups are proactively adopting a public dedication approach based on permissive open source licenses and patent avoidance. Ive also been encouraged to learn that the public dedication philosophy doesnt seem to be in conflict with the requirements of faculty and student startups there are even a few instances of unicorns (startups with billion dollar valuations) coming out of academic centers with a public dedication philosophy. Perhaps this relatively relaxed approach is possible because our industry competes more on implementation than on intellectual property.

Who is someone in your field that is doing exciting, forward-thinking research?

One example is Scott Shenkers [Professor of Computer Science at UC Berkeley] Extensible Internet Project. Im a sucker for reinventing the internet projects. The Extensible Internet idea turns some conventional wisdoms upside down and recognizes certain practical realities about the nature of internetworking. Its doing it in a way that may address the roadblocks of putting a reinvented internet into practice. Im particularly excited about that kind of thing, where we know that there are some improvements to be made and were doing it in a way thats a little unconventional, so that it might become something that can be introduced in the real world.

A second example is another UC Berkeley effort around sky computing. In computer science we always look at things through the lens of what the next level of abstraction can be, and what it will look like. The sky computing effort is interesting, because it explicitly recognizes a phenomenon that we call multi-cloud. It focuses around the idea of service brokers and abstraction layers that sit above the cloudwhich is why its called sky computing.

What will computer science research look like in ten to twenty years?

I think what were going to see is a continuation of the current trend where computing becomes increasingly embedded in many different domains: medicine, biology, physics, astronomy. It has the potential to add tremendous capability and maybe even accelerate the development of new advances in certain fields. But, of course, accelerating research and doing it at a greater scale also has the increased potential for unintended consequences.

Who should regulate those potential consequences?

You may have heard about a ceremony Canadian engineers undertake at the beginning of their careers that involves an iron ring. The ring provides a reminder of a bridge collapse that was the result of some calculations that werent done correctly. The ring is worn in a way that rubs against papers when people are doing calculations, to remind them that they need to do those calculations carefully and consider the social consequences of their work.

What is the responsible code of social responsibility that computer scientists should be adhering to? Maybe we can develop that ourselves. The computing research community as a whole should be thinking about its own standards for social responsibility. Its not the province of any one party, and does not need to come from government or academia or industry specifically.

I recently attended the Computing Research Associations biennial meeting; it was the first time people have gotten together in four years after the COVID-19 pandemic. There was extensive discussion about how we as a field need to address social responsibility issues. Theres really a great deal of awareness of the need and the opportunity. The conversations are beginning. I think people are taking them seriously. Well see if there is actionable progress.

Do you have any advice for early career researchers?

One of the things that early career researchers should think about is that sometimes it takes 20 years to see an idea through. To see that idea through may be something you can accelerate by moving from academia to industry to government, by being flexible in how your career evolves. I would encourage people not to get too dialed in to one track, but to think about the opportunities that come from taking on a different point of view at different points in ones career.

A second piece of advice that I wish I had been given myself is not to over appreciate the state of the art at the beginning of your career. Sometimes, research results can seem so polished, and the expression of those research results can seem so carefully articulated that its hard to imagine replacing it with something else or improving on it. Its important to recognize that everything is a work in progress. Its also important not to be shy about rethinking things and recognize there are completely novel ways of looking at the problem and taking a different perspective on it. Recognizing that even things that look near perfect can be evolved is an important aspect of creating new inventions and advancing the state of the art.

What do you hope the next generation of researchers are thinking about?

My hope is that the next generation recognizes that their work can be tremendously impactful. I think that people in the field of computer science are in a position to make a meaningful contribution to the evolution of society and the health of the planet. Im hoping that the next generation recognizes that opportunity, and doesnt just pursue new ideas for the sake of new ideas. Instead, they should research ideas that are poised to do something constructive in the world.

Published on August 23rd, 2022

Last updated on August 23rd, 2022

Visit link:

The computing research community should think about its social responsibility - USC Viterbi | School of Engineering - USC Viterbi School of...