CLOSE
Tags Posts tagged with "artificial intelligence"

artificial intelligence

by -
0 2176
Ramana Davuluri

By Daniel Dunaief

Ramana Davuluri feels like he’s returning home.

Davuluri first arrived in the United States from his native India in 1999, when he worked at Cold Spring Harbor Laboratory. After numerous other jobs throughout the United States, including as Assistant Professor at Ohio State University and Associate Professor and Director of Computational Biology at The Wistar Institute in Philadelphia, Davuluri has come back to Long Island. 

As of the fall of 2020, he became a Professor in the Department of Biomedical Informatics and Director of Bioinformatics Shared Resource at Stony Brook Cancer Center.

“After coming from India, this is where we landed and where we established our life. This feels like our home town,” said Davuluri, who purchased a home in East Setauket with his wife Lakshmi and their six-year-old daughter Roopavi.

Although Davuluri’s formal training in biology ended in high school, he has applied his foundations in statistics, computer programming and, more recently, the application of machine learning and deep algorithms to the problems of cancer data science, particularly for analyses of genomic and other molecular data.

Davuluri likens the process of the work he does to interpreting language based on the context and order in which the words appear.

The word “fly,” for example, could be a noun, as in an insect at a picnic, or a verb, as in to hop on an airplane and visit family for the first time in several years.

Interpreting the meaning of genetic sentences requires an understanding not only of the order of a genetic code, but also of the context in which that code builds the equivalent of molecular biological sentences.

A critical point for genetic sequences starts with a promoter, which is where genes become active. As it turns out, these areas have considerable variability, which affects the genetic information they produce.

“Most of the genetic variability we have so far observed in population-level genomic data is present near the promoter regions, with the highest density overlapping with the transcription start site,” he explained in an email.

Most of the work he does involves understanding the non-coding portion of genomes. The long-term goal is to understand the complex puzzle of gene-gene interactions at isoform levels, which means how the interactions change if one splice variant is replaced by another of the same gene.

“We are trying to prioritize variants by computational predictions so the experimentalists can focus on a few candidates rather than millions,” Davuluri added.

Most of Davuluri’s work depends on the novel application of machine learning. Recently, he has used deep learning methods on large volumes of data. A recent example includes building a classifier based on a set of transcripts’ expression to predict a subtype of brain cancer or ovarian cancer.

In his work on glioblastoma and high grade ovarian cancer subtyping, he has applied machine learning algorithms on isoform level gene expression data.

Davuluri hopes to turn his ability to interpret specific genetic coding regions into a better understanding not only of cancer, but also of the specific drugs researchers use to treat it.

He recently developed an informatics pipeline for evaluating the differences in interaction profiles between a drug and its target protein isoforms.

In research he recently published in Scientific Reports, he found that over three quarters of drugs either missed a potential target isoform or target other isoforms with varied expression in multiple normal tissues.

Research into drug discovery is often done “as if one gene is making one protein,” Davuluri said. He believes the biggest reason for the failure of early stage drug discovery resides in picking a candidate that is not specific enough.

Ramana Davuluri with his daughter Roopavi. Photo by Laskshmi Davuluri

Davuluri is trying to make an impact by searching more specifically for the type of protein or drug target, which could, prior to use in a clinical trial, enhance the specificity and effectiveness of any treatment.

Hiring Davuluri expands the bioinformatics department, in which Joel Saltz is chairman, as well as the overall cancer effort. 

Davuluri had worked with Saltz years ago when both scientists conducted research at Ohio State University.

“I was impressed with him,” Saltz said. “I was delighted to hear that he was available and potentially interested. People who are senior and highly accomplished bioinfomaticians are rare and difficult to recruit.”

Saltz cited the “tremendous progress” Davuluri has made in the field of transcription factors and cancer.

Bioinformatic analysis generally doesn’t take into account the way genes can be interpreted in different ways in different kinds of cancer. Davuluri’s work, however, does, Saltz said.

Developing ways to understand how tumors interact with non-tumor areas, how metastases develop, and how immune cells interact with a tumor can provide key advances in the field of cancer research, Saltz said. “If you can look at how this plays out over space and time, you can get more insights as to how a cancer develops and the different part of cancer that interact,” he said.

When he was younger, Davuluri dreamt of being a doctor. In 10th grade, he went on a field trip to a nearby teaching hospital, which changed his mind after watching a doctor perform surgery on a patient.

Later in college, he realized he was better in mathematics than many other subjects.

Davuluri and Lakshmi are thrilled to be raising their daughter, whose name is a combination of the words for “beautiful” and “brave” in their native Telugu.

As for Davuluri’s work, within the next year he would like to understand variants. 

“Genetic variants can explain not only how we are different from one another, but also our susceptibility to complex diseases,” he explained. With increasing population level genomic data, he hopes to uncover variants in different ethnic groups that might provide better biomarkers.

by -
0 1652
Partha Mitra at the Shanghai Natural History Museum in China where he was giving a talk to children on how birds learn to sing.

By Daniel Dunaief

Throw a giant, twisted multi-colored ball of yarn on the floor, each strand of which contains several different colored parts. Now, imagine that the yarn, instead of being easy to grasp, has small, thin, short intertwined strings. It would be somewhere between difficult and impossible to tease apart each string.

Instead of holding the strings and looking at each one, you might want to construct a computer program that sorted through the pile.

That’s what Partha Mitra, a professor at Cold Spring Harbor Laboratory, is doing, although he has constructed an artificial intelligence program to look for different parts of neurons, such as axons, dendrites and soma, in high resolution images.

Partha Mitra at the Owl Cafe in Tokyo

Working with two dimensional images which form a three dimensional stock, he and a team of scientists have performed a process called semantic segmentation, in which they delineated all the different neuronal compartments in an image.

Scientists who design machine learning programs generally take two approaches: they either train the machine to learn from data or they tailor them based on prior knowledge. “There is a larger debate going on in the machine learning community,” Mitra said.

His effort attempts to take this puzzle to the next step, which hybridizes the earlier efforts, attempting to learn from the data with some prior knowledge structure built in. “We are moving away from the purely data driven” approach, he explained.

Mitra and his colleagues recently published a paper about their artificial intelligence-driven neuroanatomy work in the journal Nature Machine Intelligence.

For postmortem human brains, one challenge is that few whole-brain light microscopic data sets exist. For those that do exist, the amount of data is large enough to tax available resources.

Indeed, the total amount of storage to study one brain at light microscope resolution is one petabyte of data, which amounts to a million megapixel images.

“We need an automated method,” Mitra said. “We are on the threshold of where we are getting data a cellular resolution of the human brain. You need these techniques” for that discovery. Researchers are on the verge of getting more whole-brain data sets more routinely.

Mitra is interested in the meso-scale architecture, or the way groups of neurons are laid out in the brain. This is the scale at which species-typical structures are visible. Individual cells would show strong variation from one individual to another. At the mesoscale, however, researchers expect the same architecture in brains of different neurotypical individuals of the same species.

Trained as a physicist, Mitra likes the concreteness of the data and the fact that neuroanatomical structure is not as contingent on subtle experimental protocol differences.

He said behavioral and neural activity measurements can depend on how researchers set up their study and appreciates the way anatomy provides physical and architectural maps of brain cells.

The amount of data neuroanatomists have collected exceeds the ability of these specialists to interpret it, in part because of the reduction in cost of storing the information. In 1989, a human brain worth of light microscope data would have cost approximately the entire budget for the National Institutes of Health based on the expense of hard disk storage at the time. Today, Mitra can buy that much data storage every year with a small fraction of his NIH grant.

“There has been a very big change in our ability to store and digitize data,” he said. “What we don’t have is a million neuroanatomists looking at this. The data has exploded in a systematic way. We can’t [interpret and understand] it unaided by the computer.”

Mitra described the work as a “small technical piece of a larger enterprise,” as the group tries to address whether it’s possible to automate what a neuroanatomist does. Through this work, he hopes computers might discover common principals of the anatomy and construction of neurons in the brain.

While the algorithms and artificial intelligence will aid in the process, Mitra doesn’t expect the research to lead to a fully automated process. Rather, this work has the potential to accelerate the process of studying neuroanatomy.

Down the road, this kind of understanding could enable researchers and ultimately health care professionals to compare the architecture and circuitry of brains from people with various diseases or conditions with those of people who aren’t battling any neurological or cognitive issues.

“There’s real potential to looking at” the brains of people who have various challenges, Mitra said.

The paper in Nature Machine Intelligence reflected a couple of years of work that Mitra and others did in parallel with other research pursuits.

A resident of Midtown, Mitra, his wife Tatiana and their seven-year-old daughter have done considerable walking around the city during the pandemic.

The couple created a virtual exhibit for the New York Hall of Science in the Children’s Science Museum in which they described amazing brains. A figurative sculptor, Tatiana provided the artwork for the exhibition.

Mitra, who has been at Cold Spring Harbor Laboratory since 2003, said neuroanatomy has become increasingly popular over the last several years. He would like to enhance the ability of the artificial intelligence program in this field.

“I would like to eliminate the human proofreading,” he said. “We are still actively working on the methodology.”

Using topological methods, Mitra has also traced single neurons. He has published that work through a preprint in bioRxiv.

by -
0 4143
New documentary examines the future of artificial intelligence               and the impact it will have on our world.

Reviewed By Jeffrey Sanzel

We Need to Talk About AI is an intriguing and occasionally alarmist documentary that explores the historical and current developments in Artificial Intelligence. It raises far more questions than it even attempts to answer and that, most likely, is its point.  The title’s urgency is appropriate to this peripatetically engaging ninety minutes.

Director Leanne Pooley has conducted extensive interviews with scientists, engineers, philosophers, filmmaker James Cameron, and a whole range of experts, along with dozens of clips from news broadcasts and nearly one hundred years of science fiction movies. The film plays at a breakneck pace, fervently bouncing from one opinion to an alternate point-of-view.

Currently streaming On Demand, the documentary is appropriately hosted by Keir Dullea, who gives a dry menace to the narration and occasionally appears walking through crowded streets like a being from an alternate universe. Dullea is best known as astronaut Dave Bowman in Stanley Kubrick’s landmark 2001: A Space Odyssey (1968).  Pooley uses the film’s HAL (Heuristically Programmed ALgorithmic Computer) as the example of man’s greatest fear in the world of AI: a computer that becomes sentient and will no longer obey its human creators.

The early days of AI work seems almost quaint in comparison to latter-day capabilities.  Much of this can be traced to the advancement in the computer technology and the rise of the internet. The internet’s considerable expansion in the last two decades has been the greatest gamechanger. 

A constant refrain is that the dialogue surrounding AI has been “hijacked” by Hollywood: the majority of the populace associate AI in negative terms. It is about the rebellion of manmade machines (e.g., The Terminator). The scientists are in agreement that this is a misrepresentation. That is, they are for the most part. As the film progresses, the views on the dangers of AI diverge.

It all comes down to the question of conscience and autonomy. There is a dissection of the issues behind self-driving cars and how to embed ethics into the machine. The Trolley Problem — how do you decide who to save —  is used to demonstrate the challenge. To make the decision, the machine would have to be a conscious being. 

Furthermore, can a machine be conscious or have a conscience? The idea of conscious and conscience becomes central.  As it is almost impossible to define what “conscious” is, it creates additional conflicts in the narrative. This leads to conversations on emotion and whether machines will ever be able to feel and react to social cues.  

The film poses many hypotheses and explores the predicament from all sides. There is rarely uniform agreement. Can a machine make itself smarter by programming itself? Will the evolution be gradual or exponential?

Even now, robot surgery, agriculture, and even Facebook’s suicide awareness algorithms are examples given of the recent uses of AI. Computers can now beat the world’s greatest chess players. Not that many years ago, these were considered impossible outside of speculative fiction.

Throughout, Pooley returns to the teaching of Baby X, an intelligent toddler simulation that is both fascinating and chilling. Baby X almost seems human and appears to be learning. It is a strange and exciting phenomenon.

Already, the argument is made that we carry less in our brains because we carry parts of them in our pockets in the form of cell phones. In essence, they are the merging of minds with computers. They are an augmentation and a symbiotic integration. 

Ultimately, it comes down to not so much how we build AI but what we do with it. The unifying position of the interviewed is the fear that this power will be used for evil — or at least negative purposes. (Pooley unsubtly does a quick montage of the world’s foremost demagogues.) 

The consensus is that it should not be about who arrives first but who gets there safely. They hope but doubt for regulation. If it is corporations or business (Google, Microsoft, etc.) that get primary control, it will be driven by greed. If it is the military, it will be about killing. They say we only have one chance to get it right, and the leader in the field must, in essence, be the good parent. AI will dominate the economy and, therefore, the world. 

There are myriad questions raised: What it means to be human? If machines become more, will we be become less? Is AI going to do something for you or to you? Is science fiction the canary in the coal mine? That is, do we face the apocalypse if AI doesn’t play out the positive scenarios?

And then there are the moral questions. Can machines be made accountable? Does a machine have rights? If so, is this a form of slavery, where conscious beings are created and then dehumanized? There is a brief section about the rise of sex robots that is twinned with a clip from the 1927 silent film Metropolis. Can a machine say, “No?”

Perhaps we have come a long way from the science fiction movies of our past. Maybe we will never face the voice of HAL saying, “I’m sorry, Dave. I can’t do that.”

Or perhaps we will.  

The final line sums up the entire journey:  “What do we want the role of humans to be?” We Need to Talk About AI is a great place to start.

by -
0 3489
From left, graduate students Prakhar Avasthi, Alisa Yurovsky, Charuta Pethe and Haochen Chen with director Steven Skiena, center. Photo by Gary Ghayrat/Stony Brook University

By Daniel Dunaief

Steven Skiena practices what he teaches. Named the director of the Institute for AI-Driven Discovery and Innovation in the College of Engineering and Applied Sciences at Stony Brook University, Skiena is using artificial intelligence to search for three staff members he hopes to hire in this new initiative.

He is looking for two tenured professors who will work in the Department of Computer Science and one who will be a part of the Department of Biomedical Informatics.

“We hope to use an artificial intelligence screen,” which Skiena calls a Poach-o-matic to “identify candidates we might not have thought of before. Ideally, the program will kick up a name and afterward, we’d bump our hand on our head and say, ‘Of course, this person might be great.’”

Steven Skiena. Photo from SBU

Artificial intelligence and machine learning have become popular areas in research institutions like Stony Brook, as well as in corporations with a wide range of potential applications, including in search engine companies like Google.

Skiena, who is a distinguished teaching professor, said he has “several candidates and we’re now actively interviewing,” adding that many departments on campus have faculty who are interested in applying machine learning in their work.

“There’s been an explosion of people from all disciplines who are interested in this,” Skiena said. He recently met with a materials scientist who uses machine learning techniques to improve experimental data. He’s also talked with people from the business school and from neuroscience.

SBU students have also shown considerable interest in these areas. Last semester, Skiena taught 250 graduate students in his introduction to data science class.

“This is a staggering demand from students that are very excited about this,” he said. Machine learning has become a “part of the standard tool kit for doing mathematical modeling and forecasting in many disciplines and that’s only going to increase.”

In an recent email, Andrew Schwartz, a core faculty at the institute and an assistant professor in the Department of Computer Science at Stony Brook, said he believes bringing in new faculty “should attract additional graduate students that may become future leaders in the field.”

Increasing coverage of AI beyond the current expertise in vision, visualization, natural language processing and biomedical engineering can “go a long way. There are a large amount of breakthroughs in AI that seemingly come from taking an idea from one subfield and applying it to another.” Schwartz appreciates the impact Skiena, who is his faculty mentor, has had on the field.

Skiena has “managed to contribute to a wide range of topics,” Schwartz explained. His book, “The Algorithm Design Manual,” is used by people worldwide preparing for technical interviews. Knowing this book thoroughly is often a “suggested step” for people preparing to interview at Google or other tech companies, Schwartz added.

The students in Schwartz and Skiena’s labs share space and have regular weekly coffee hours. Schwartz appreciates how Skiena often “presents a puzzling question or an out-of-the-box take on a question.”

The core technical expertise at the institute is in machine learning, data science, computer vision and natural language processing.

The creation of the institute shows that Stony Brook is “serious about being one of the top universities and research centers for expertise in AI,” explained Schwartz.

A few years ago, researchers realized that the artificial intelligence models developed biases based on the kind of training data used to create them. “If you’re trying to build a system to judge resumes to decide who will be a good person to hire for a certain type of job” the system has a danger of searching for male candidates if most or all of the people hired had been male in the past, Skiena explained.

Unintentional biases can creep in if the data sets are skewed toward one group, even if the programmer who created the artificial intelligence system was using available information and patterns.

In his own research, Skiena, who has been at Stony Brook since 1988, works on natural language processing. Specifically, he has explored the meaning of words and what a text is trying to communicate.

He has worked on sentiment analysis, trying to understand questions such as whether a particular political figure who receives considerable media coverage is having a good or bad week.

Another project explores the quality of news sources. “Can you algorithmically analyze large corpuses of news articles and determine which are reliable and which are less so?” he asked. 

One measure of the reliability of a news source is to determine how much other articles cite from it. “It is important to teach skepticism of a source” of news or of data, Skiena said. 

“When I teach data science, a lot of what I teach includes questions of why you believe a model will do a good thing and why is a data source relevant,” he added.

A resident of Setauket, Skiena lives with his wife Renee. Their daughter Bonnie is a first-year student at the University of Delaware, where she is studying computer science. Their tenth-grade daughter Abby attends Ward Melville High School and joins her father for bike rides on Long Island.

Skiena, who grew up in East Brunswick, New Jersey, said he appreciates the university community. By working in the AI field, Skiena, who has seven doctoral students in his lab, said he often observes glitches in online models like article classification on Google News or advertisements selected for him on a website to try to figure out why the model erred. He has also developed a sense of how probability and random events work, which he said helps him not overinterpret unusual events in day-to-day life.

As for his work at the institute, Skiena hopes Stony Brook will be recognized as a major player in the field of machine learning and areas of artificial intelligence. “We have good faculty in this area already and we’re hiring more. The hope is that you reach critical mass.”

Prev12Page 2 of 2
TBR News Media is your local news and entertainment website.
We provide you with the latest breaking news and information for your community.
Contact us: [email protected]

Most Viewed

©