For over three decades, a highly accurate early diagnostic test for ovarian cancer has eluded physicians. Now, scientists in the Georgia Tech Integrated Cancer Research Center (ICRC) have combined machine learning with information on blood metabolites to develop a new test able to detect ovarian cancer with 93 percent accuracy among samples from the team’s study group.

John McDonald, professor emeritus in the School of Biological Sciences, founding director of the ICRC, and the study’s corresponding author, explains that the new test’s accuracy is better in detecting ovarian cancer than existing tests for women clinically classified as normal, with a particular improvement in detecting early-stage ovarian disease in that cohort.

The team’s results and methodologies are detailed in a new paper, “A Personalized Probabilistic Approach to Ovarian Cancer Diagnostics,” published in the March 2024 online issue of the medical journal Gynecologic Oncology. Based on their computer models, the researchers have developed what they believe will be a more clinically useful approach to ovarian cancer diagnosis — whereby a patient’s individual metabolic profile can be used to assign a more accurate probability of the presence or absence of the disease.

“This personalized, probabilistic approach to cancer diagnostics is more clinically informative and accurate than traditional binary (yes/no) tests,” McDonald says. “It represents a promising new direction in the early detection of ovarian cancer, and perhaps other cancers as well.”

The study co-authors also include Dongjo Ban, a Bioinformatics Ph.D. student in McDonald’s lab; Research Scientists Stephen N. Housley, Lilya V. Matyunina, and L.DeEtte (Walker) McDonald; Regents’ Professor Jeffrey Skolnick, who also serves as Mary and Maisie Gibson Chair in the School of Biological Sciences and Georgia Research Alliance Eminent Scholar in Computational Systems Biology; and two collaborating physicians: University of North Carolina Professor Victoria L. Bae-Jump and Ovarian Cancer Institute of Atlanta Founder and Chief Executive Officer Benedict B. Benigno. Members of the research team are forming a startup to transfer and commercialize the technology, and plan to seek requisite trials and FDA approval for the test.

Silent killer

Ovarian cancer is often referred to as the silent killer because the disease is typically asymptomatic when it first arises — and is usually not detected until later stages of development, when it is difficult to treat.

McDonald explains that while the average five-year survival rate for late-stage ovarian cancer patients, even after treatment, is around 31 percent — but that if ovarian cancer is detected and treated early, the average five-year survival rate is more than 90 percent.

“Clearly, there is a tremendous need for an accurate early diagnostic test for this insidious disease,” McDonald says.

And although development of an early detection test for ovarian cancer has been vigorously pursued for more than three decades, the development of early, accurate diagnostic tests has proven elusive. Because cancer begins on the molecular level, McDonald explains, there are multiple possible pathways capable of leading to even the same cancer type.

“Because of this high-level molecular heterogeneity among patients, the identification of a single universal diagnostic biomarker of ovarian cancer has not been possible,” McDonald says. “For this reason, we opted to use a branch of artificial intelligence — machine learning — to develop an alternative probabilistic approach to the challenge of ovarian cancer diagnostics.”

Metabolic profiles

Georgia Tech co-author Dongjo Ban, whose thesis research contributed to the study, explains that “because end-point changes on the metabolic level are known to be reflective of underlying changes operating collectively on multiple molecular levels, we chose metabolic profiles as the backbone of our analysis.”

“The set of human metabolites is a collective measure of the health of cells,” adds coauthor Jeffrey Skolnick, “and by not arbitrary choosing any subset in advance, one lets the artificial intelligence figure out which are the key players for a given individual.”

Mass spectrometry can identify the presence of metabolites in the blood by detecting their mass and charge signatures. However, Ban says, the precise chemical makeup of a metabolite requires much more extensive characterization.

Ban explains that because the precise chemical composition of less than seven percent of the metabolites circulating in human blood have, thus far, been chemically characterized, it is currently impossible to accurately pinpoint the specific molecular processes contributing to an individual's metabolic profile.

However, the research team recognized that, even without knowing the precise chemical make-up of each individual metabolite, the mere presence of different metabolites in the blood of different individuals, as detected by mass spectrometry, can be incorporated as features in the building of accurate machine learning-based predictive models (similar to the use of individual facial features in the building of facial pattern recognition algorithms).

“Thousands of metabolites are known to be circulating in the human bloodstream, and they can be readily and accurately detected by mass spectrometry and combined with machine learning to establish an accurate ovarian cancer diagnostic,” Ban says.

A new probabilistic approach

The researchers developed their integrative approach by combining metabolomic profiles and machine learning-based classifiers to establish a diagnostic test with 93 percent accuracy when tested on 564 women from Georgia, North Carolina, Philadelphia and Western Canada. 431 of the study participants were active ovarian cancer patients, and while the remaining 133 women in the study did not have ovarian cancer.

Further studies have been initiated to study the possibility that the test is able to detect very early-stage disease in women displaying no clinical symptoms, McDonald says.

McDonald anticipates a clinical future where a person with a metabolic profile that falls within a score range that makes cancer highly unlikely would only require yearly monitoring. But someone with a metabolic score that lies in a range where a majority (say, 90%) have previously been diagnosed with ovarian cancer would likely be monitored more frequently — or perhaps immediately referred for advanced screening.

Citation: https://doi.org/10.1016/j.ygyno.2023.12.030

Funding

This research was funded by the Ovarian Cancer Institute (Atlanta), the Laura Crandall Brown Foundation, the Deborah Nash Endowment Fund, Northside Hospital (Atlanta), and the Mark Light Integrated Cancer Research Student Fellowship.

Disclosure

Study co-authors John McDonald, Stephen N. Housley, Jeffrey Skolnick, and Benedict B. Benigno are the co-founders of MyOncoDx, Inc., formed to support further research, technology transfer, and commercialization for the team’s new clinical tool for the diagnosis of ovarian cancer.

Researchers at the Georgia Institute of Technology have developed a light-based means of printing nano-sized metal structures that is significantly faster and cheaper than any technology currently available. It is a scalable solution that could transform a scientific field long reliant on technologies that are prohibitively expensive and slow. The breakthrough has the potential to bring new technologies out of labs and into the world.

Technological advances in many fields rely on the ability to print metallic structures that are nano-sized — a scale hundreds of times smaller than the width of a human hair. Sourabh Saha, assistant professor in the George W. Woodruff School of Mechanical Engineering, and Jungho Choi, a Ph.D. student in Saha’s lab, developed a technique for printing metal nanostructures that is 480 times faster and 35 times cheaper than the current conventional method.

Their research was published in the journal Advanced Materials.

Printing metal on the nanoscale — a technique known as nanopatterning — allows for the creation of unique structures with interesting functions. It is crucial for the development of many technologies, including electronic devices, solar energy conversion, sensors, and other systems.

It is generally believed that high-intensity light sources are required for nanoscale printing. But this type of tool, known as a femtosecond laser, can cost up to half a million dollars and is too expensive for most research labs and small businesses.

“As a scientific community, we don’t have the ability to make enough of these nanomaterials quickly and affordably, and that is why promising technologies often stay limited to the lab and don’t get translated into real-world applications,” Saha said.

“The question we wanted to answer is, ‘Do we really need a high-intensity femtosecond laser to print on the nanoscale?’ Our hypothesis was that we don’t need that light source to get the type of printing we want.”

They searched for a low-cost, low-intensity light that could be focused in a way similar to femtosecond lasers, and chose superluminescent light emitting diodes (SLEDs) for their commercial availability. SLEDs emit light that is a billion times less intense than that of femtosecond lasers.

Saha and Choi set out to create an original projection-style printing technology, designing a system that converts digital images into optical images and displays them on a glass surface. The system operates like digital projectors but produces images that are more sharply focused. They leveraged the unique properties of the superluminescent light to generate sharply focused images with minimal defects.

They then developed a clear ink solution made up of metal salt and added other chemicals to make sure the liquid could absorb light. When light from their projection system hit the solution, it caused a chemical reaction that converted the salt solution into metal. The metal nanoparticles stuck to the surface of the glass, and the agglomeration of the metal particles creates the nanostructures. Because it is a projection type of printing, it can print an entire structure in one go, rather than point by point — making it much faster.

After testing the technique, they found that projection-style nanoscale printing is possible even with low-intensity light, but only if the images are sharply focused. Saha and Choi believe that researchers can readily replicate their work using commercially available hardware. Unlike a pricey femtosecond laser, the type of SLED that Saha and Choi used in their printer costs about $3,000.

“At present, only top universities have access to these expensive technologies, and even then, they are located in shared facilities and are not always available,” Choi said. “We want to democratize the capability of nanoscale 3D printing, and we hope our research opens the door for greater access to this type of process at a low cost.”

The researchers say their technique will be particularly useful for people working in the fields of electronics, optics, and plasmonics, which all require a variety of complex metallic nanostructures.

“I think the metrics of cost and speed have been greatly undervalued in the scientific community that works on fabrication and manufacturing of tiny structures,” Saha said.

“In the real world, these metrics are important when it comes to translating discoveries from the lab to industry. Only when we have manufacturing techniques that take these metrics into account will we be able to fully leverage nanotechnology for societal benefit.”

Citation: J. Choi, S. K. Saha, Scalable Printing of Metal Nanostructures through Superluminescent Light Projection. Adv. Mater. 2024, 36, 2308112.

DOI: https://doi.org/10.1002/adma.202308112

Funding: Funding includes grants from the G.W.W. School of Mechanical Engineering and the EVPR’s office at the Georgia Institute of Technology. Imaging was performed at the Georgia Tech Institute for Electronics and Nanotechnology, a member of the National Nanotechnology Coordinated Infrastructure (NNCI), which is supported by the National Science Foundation (ECCS-2025462).

This press release is shared jointly with the Event Horizon Telescope newsroom.

The Event Horizon Telescope (EHT) Collaboration has released new images of M87*, the supermassive black hole at the center of the galaxy Messier 87, using data from observations taken in April 2018.

With the participation of the newly commissioned Greenland Telescope and a dramatically improved recording rate across the array, the 2018 observations give researchers a view of the source independent from the first observations in 2017. 

“This persistence is a remarkable confirmation of our earlier interpretation that the EHT images do reveal the shadow of the black hole — and strengthens the tests of Einstein’s theories that we have performed,” says Dimitrios Psaltis, professor in the School of Physics at the Georgia Institute of Technology who served as EHT project scientist at the time of the 2019 announcement.

Psaltis and Georgia Tech School of Physics Chair Feryal Özel are members of the EHT collaboration, and are the scientists who developed many of the theoretical tools to analyze and interpret the images.

“Modeling the image features of this black hole across observations that span many years — and comparing them to the images of the black hole in the center of our Milky Way — already provide powerful checks on our plasma models” says Özel, who led the 2022 announcement of the image of the Milky Way black hole.

New era of black hole imaging

A recent paper published in the journal Astronomy & Astrophysics presents the team’s new images from the 2018 data that reveal a familiar ring the same size as the one observed in 2017. This bright ring surrounds a deep central depression, “the shadow of the black hole,” as predicted by general relativity. Excitingly, the brightness peak of the ring has shifted by about 30º compared to the images from 2017, which is consistent with our theoretical understanding of variability from turbulent material around black holes.

“A fundamental requirement of science is to be able to reproduce results,” says Keiichi Asada, an associate research fellow at Academia Sinica Institute for Astronomy and Astrophysics in Taiwan. “Confirmation of the ring in a completely new data set is a huge milestone for our collaboration and a strong indication that we are looking at a black hole shadow and the material orbiting around it.”

In 2017, the EHT took the first image of a black hole. This object, M87*, is the beating heart of the giant elliptical galaxy Messier 87 and lives 55 million light years away from Earth. The image of the black hole revealed a bright circular ring, brighter in the southern part of the ring. Further analysis of the data also revealed the structure of M87* in polarized light, giving us greater insight into the geometry of the magnetic field and the nature of the plasma around the black hole.

The new era of black hole direct imaging, spearheaded by the extensive analysis of the 2017 observations of M87*, opened a new window that let researchers investigate black hole astrophysics and allow them to test the theory of general relativity at a fundamental level.

“Our theoretical models tell us that the state of the material around M87* should be uncorrelated between 2017 and 2018,” EHT researchers explain. “Thus, multiple observations of M87* will help us place independent constraints on the plasma and magnetic field structure around the black hole and help us disentangle the complicated astrophysics from the effects of general relativity.”

Greenland Telescope

To help accomplish new and exciting science, the EHT is under continuous development. The Greenland Telescope joined the EHT for the first time in 2018, just five months after its construction was completed far above the Arctic Circle. This new telescope significantly improved the image fidelity of the EHT array, improving the coverage, particularly in the North-South direction. The Large Millimeter Telescope also participated for the first time with its full 50 m surface, greatly improving its sensitivity. The EHT array was also upgraded to observe in four frequency bands around 230 GHz, compared to only two bands in 2017.

Repeated observations with an improved array are essential to demonstrate the robustness of our findings and strengthen our confidence in our results.  In addition to the groundbreaking science, the EHT also serves as a technology testbed for cutting-edge developments in high-frequency radio interferometry.

"Advancing scientific endeavors requires continuous enhancement in data quality and analysis techniques," says Rohan Dahale, a Ph.D. candidate at the Instituto de Astrofísica de Andalucía (IAA-CSIC) in Spain. "The inclusion of the Greenland Telescope in our array filled critical gaps in our earth-sized telescope. The 2021, 2022, and the forthcoming 2024 observations witness improvements to the array, fueling our enthusiasm to push the frontiers of black hole astrophysics."

Remarkably similar

The analysis of the 2018 data features eight independent imaging and modeling techniques, including methods used in the previous 2017 analysis of M87* and new ones developed from the collaboration’s experience analyzing Sgr A*.

The EHT team explains that the image of M87* taken in 2018 is remarkably similar to what they saw in 2017. “We see a bright ring of the same size, with a dark central region and one side of the ring brighter than the other. The mass and distance of M87* will not appreciably increase throughout a human lifetime, so general relativity predicts that the ring diameter should stay the same from year to year. The stability of the measured diameter in the images from 2017 to 2018 robustly supports the conclusion that M87* is well described by general relativity.”

Mass matters, brightness peak

“One of the remarkable properties of a black hole is that its radius is strongly dependent on only one quantity: its mass,” says Nitika Yadlapalli Yurk, a former graduate student at the California Institute of Technology (Caltech), now a postdoctoral fellow at the NASA Jet Propulsion Laboratory (JPL) in California. “Since M87* is not accreting material (which would increase its mass) at a rapid rate, general relativity tells us that its radius will remain fairly unchanged over human history. It’s pretty exciting to see that our data confirm this prediction.”

While the size of the black hole shadow did not change between 2017 and 2018, the location of the brightest region around the ring did change significantly, the team adds. The bright region rotated about 30º counterclockwise to settle in the bottom right part of the ring at about the 5 o’clock position.

Historical observations of M87* with a less sensitive array and fewer telescopes also indicated that the shadow structure changes yearly (Wielgus 2020, ApJ, 901, 67) but with less precision. While the 2018 EHT array still cannot observe the jet emerging from M87*, the black hole spin axis predicted from the location of the brightest region around the ring is more consistent with the jet axis seen at other wavelengths.  

“The biggest change, that the brightness peak shifted around the ring, is actually something we predicted when we published the first results in 2019,” says Britt Jeter, a postdoctoral fellow at Academia Sinica Institute for Astronomy and Astrophysics in Taiwan. “While general relativity says the ring size should stay pretty fixed, the emission from the turbulent, messy accretion disk around the black hole will cause the brightest part of the ring to wobble around a common center. The amount of wobble we see over time is something we can use to test our theories for the magnetic field and plasma environment around the black hole.”

2024 and beyond

“While all the EHT papers published so far have featured an analysis of our first observations in 2017,” the research team adds, “this result represents the first efforts to explore the many additional years of data the EHT collaboration has collected.” In addition to 2017 and 2018, the EHT conducted successful observations in 2021 and 2022 and is scheduled to observe in the first half of 2024. Each year, the EHT array has improved in some way, either through the addition of new telescopes, better hardware, or additional observing frequencies. “Within the collaboration, we are working very hard to analyze all this data and are excited to show you more results in the future.”

###

DOI: https://doi.org/10.1051/0004-6361/202347932 

ABOUT EHT

The EHT collaboration involves more than 300 researchers from Africa, Asia, Europe, and North and South America. The international collaboration is working to capture the most detailed black hole images ever obtained by creating a virtual Earth-sized telescope. Supported by considerable international investment, the EHT links existing telescopes using novel systems, creating a fundamentally new instrument with the highest angular resolving power that has yet been achieved.

The individual telescopes involved are ALMA, APEX, the IRAM 30-meter Telescope, the IRAM NOEMA Observatory, the James Clerk Maxwell Telescope (JCMT), the Large Millimeter Telescope (LMT), the Submillimeter Array (SMA), the Submillimeter Telescope (SMT), the South Pole Telescope (SPT), the Kitt Peak Telescope, and the Greenland Telescope (GLT).  Data were correlated at the Max-Planck-Institut für Radioastronomie (MPIfR) and MIT Haystack Observatory.  The postprocessing was done within the collaboration by an international team at different institutions.

The EHT consortium consists of 13 stakeholder institutes: the Academia Sinica Institute of Astronomy and Astrophysics, the University of Arizona, the University of Chicago, the East Asian Observatory, Goethe-Universitaet Frankfurt, Institut de Radioastronomie Millimétrique, Large Millimeter Telescope, Max Planck Institute for Radio Astronomy, MIT Haystack Observatory, National Astronomical Observatory of Japan, Perimeter Institute for Theoretical Physics, Radboud University, and the Smithsonian Astrophysical Observatory.

ABOUT GEORGIA TECH

The Georgia Institute of Technology, or Georgia Tech, is one of the top public research universities in the U.S., developing leaders who advance technology and improve the human condition.

The Institute offers business, computing, design, engineering, liberal arts, and sciences degrees. Its more than 45,000 undergraduate and graduate students, representing 50 states and more than 148 countries, study at the main campus in Atlanta, at campuses in France and China, and through distance and online learning.

As a leading technological university, Georgia Tech is an engine of economic development for Georgia, the Southeast, and the nation, conducting more than $1 billion in research annually for government, industry, and society.

IMAGE:

The Event Horizon Telescope Collaboration has released new images of M87* from observations taken in April 2018, one year after the first observations in April 2017. The new observations in 2018, which feature the first participation of the Greenland Telescope, reveal a familiar, bright ring of emission of the same size as we found in 2017.  This bright ring surrounds a dark central shadow, and the brightest part of the ring in 2018 has shifted by about 30º relative from 2017 to now lie in the 5 o’clock position. (Credit: EHT Collaboration)

Machine learning (ML) has transformed the digital landscape with its unprecedented ability to automate complex tasks and improve decision-making processes. However, many organizations, including the U.S. Department of Defense (DoD), still rely on time-consuming methods for developing and testing machine learning models, which can create strategic vulnerabilities in today’s fast-changing environment. 

The Georgia Tech Research Institute (GTRI) is addressing this challenge by developing a Machine Learning Operations (MLOps) platform that standardizes the development and testing of artificial intelligence (AI) and ML models to enhance the speed and efficiency with which these models are utilized during real-time decision-making situations.   

“It’s been difficult for organizations to transition these models from a research environment and turn them into fully-functional products that can be used in real-time,” said Austin Ruth, a GTRI research engineer who is leading this project. “Our goal is to bring AI/ML to the tactical edge where it could be used during active threat situations to heighten the survivability of our warfighters.” 

Rather than treating ML development in isolation, GTRI’s MLOps platform would bridge the gap between data scientists and field operations so that organizations can oversee the entire lifecycle of ML projects from development to deployment at the tactical edge. 

The tactical edge refers to the immediate operational space where decisions are made and actions take place. Bringing AI and ML capabilities closer to the point of action would enhance the speed, efficiency and effectiveness of decision-making processes and contribute to more agile and adaptive responses to threats. 

“We want to develop a system where fighter jets or warships don’t have to do any data transfers but could train and label the data right where they are and have the AI/ML models improve in real-time as they’re actively going up against threats,” said Ruth.   

For example, a model could monitor a plane’s altitude and speed, immediately spot potential wing drag issues and alert the pilot about it. In an electronic warfare (EW) situation when facing enemy aircraft or missiles, the models could process vast amounts of incoming data to more quickly identify threats and recommend effective countermeasures in real time. 

AI/ML models need to be trained and tested to ensure their effectiveness in adapting to new, unseen data. However, without having a standardized process in place, training and testing is done in a fragmented manner, which poses several risks, such as overfitting, where the model performs well on the training data but fails to generalize unseen data and makes inaccurate predictions or decisions in real-world situations, security vulnerabilities where bad actors exploit weaknesses in the models, and a general lack of robustness and inefficient resource utilization.

“Throughout this project, we noticed that training and testing are often done in a piecemeal fashion and thus aren’t repeatable,” said Jovan Munroe, a GTRI senior research engineer who is also leading this project. “Our MLOps platform makes the training and testing process more consistent and well-defined so that these models are better equipped to identify and address unknown variables in the battle space.” 

This project has been supported by GTRI’s Independent Research and Development (IRAD) Program, winning an IRAD of the Year award in fiscal year 2023. In fiscal year 2024, the project received funding from a U.S. government sponsor. 

Writer: Anna Akins 
Photos: Sean McNeil 
GTRI Communications
Georgia Tech Research Institute
Atlanta, Georgia

The Georgia Tech Research Institute (GTRI) is the nonprofit, applied research division of the Georgia Institute of Technology (Georgia Tech). Founded in 1934 as the Engineering Experiment Station, GTRI has grown to more than 2,900 employees, supporting eight laboratories in over 20 locations around the country and performing more than $940 million of problem-solving research annually for government and industry. GTRI's renowned researchers combine science, engineering, economics, policy, and technical expertise to solve complex problems for the U.S. federal government, state, and industry.

Four faculty members have been selected for the second cohort of the Faculty Executive Leadership Academy (FELA) program, which is designed to identify and develop senior faculty members for leadership. The new cohort includes:    

• David Ballantyne, Professor and Associate Chair for Academic Programs, School of Physics  

• Martha Grover, Professor and Associate Chair for Graduate Studies, School of Chemical and Biomolecular Engineering, and ADVANCE Professor in the College of Engineering  

• Aaron Levine, Professor, School of Public Policy, and Associate Dean for Research and Outreach, Ivan Allen College of Liberal Arts  

• Han Zhang, Steven A. Denning Professor of Technology and Management, Scheller College of Business

This program will build on the Fellows’ previous leadership experiences by providing access to senior leadership and their decisions; creating opportunities for them to interact with academic leaders from across the nation; offering close, and accessible mentoring with a cohort learning model; and participating in the formulation of project-based solutions and policies related to real problems and ongoing issues. FELA is led by former provost and executive vice president for Academic Affairs Rafael Bras. 

“More than ever before, the health and future of higher education depends on good, experienced leaders with a broad understanding of many issues,” said Bras. “FELA is designed as an immersive program to develop that experience and understanding. Georgia Tech is investing in its future by developing the leadership talents of its senior faculty.” 

This cohort of FELA Fellows will serve a two-year term beginning in July 2024. During their FELA experience, Fellows will rotate every six months into one of the four Executive Leadership Team offices – the Office of the President, Office of the Provost, Office of the Vice President for Research, and Office of the Vice President for Administration and Finance. In addition to shadowing and project work during these rotations, Fellows will participate in regular cohort meetings with Bras for leadership guest speakers, case studies, and role play activities to further ground their experience.  

“Georgia Tech fosters the careers of thousands of the brightest faculty minds in higher education,” said Steve McLaughlin, provost and executive vice president for Academic Affairs. “These four new members of our Faculty Executive Leadership Academy represent those who wish to step forward as future leaders and learn what it takes to support the work of a university on an executive level. We look forward to working alongside them in the coming months.” 

Learn more about the FELA program and members of the first and second cohorts. 

In 1950, Alan Turing asked, “Can machines think?” More than 70 years later, advancements in artificial intelligence are creating exciting possibilities and questions about its potential pitfalls.  

A recent executive order issued by President Joe Biden seeks to establish "new standards for AI safety and security" while addressing consumer privacy concerns and promoting innovation. Georgia Tech experts have examined the key elements of the order and offer their thoughts on its scope and what comes next.  

A Precautionary Tale 

The order calls for the development of standards, tools, and tests to ensure the safe use of AI. From voice scams and phishing campaigns to larger-scale threats, the technology’s potential dangers have been widely documented. But Margaret Kosal, associate professor in the Ivan Allen College of Liberal Arts, says that additional context is often needed to dispel hysteria. 

"No one is going to be hooking up AI to launch nuclear weapons, but AI capabilities may enable targeting, or enable the command and control and the decision-making time to be compressed,” she said.  
 
The order will create an AI Safety and Security Board tasked with addressing critical threats. Companies developing foundation models that "pose a serious risk to national security, national economic security, or national public health and safety” will be required to notify the federal government when training the model and required to share the results of all red-team safety tests — a simulated cyberattack to test a system's defenses.  

Since the launch of ChatGPT in 2022, a CNBC report details a 1,267% rise in phishing emails. Srijan Kumar, assistant professor in the College of Computing, attributes the increase to the technology's availability and an inability to rein in "bad actors."  

He says these scams will only continue to get more sophisticated and personalized. They “can be created by knowing what you might be willing to fall prey to versus what I might fall prey to,” said Kumar, whose systems have influenced misinformation detection on sites like X (formerly Twitter) and Wikipedia. “AI is not going to autonomously do all of those bad things, but this order can ensure there are consequences for people who misuse it.”  

A Delicate Balance 

Building an AI platform requires large amounts of data regardless of its intended application. Two primary goals of the executive order are protecting privacy and advancing equity.  

To protect personal data, the order tasks Congress with evaluating how agencies collect and use commercially available information and address algorithmic discrimination.  

Acknowledging that everyone should be allowed to have their voice represented in the outputs of AI data sets, Deven Desai, associate professor in the Scheller College of Business, noted, "There are people who don't want to be part of data sets, which is their right, but this means their voices won't be reflected in the outputs.”   

The order also includes sections to address intellectual property concerns among inventors and creators, though legal challenges will likely set new precedents in the years ahead.  

When that time comes, Kosal says that defining “theft” in the context of AI becomes the true challenge and that, ultimately, money will play a significant role. "If you spit out a Harry Potter book and read it yourself, nobody will care. It's when you start selling it to make money, and you don't share proceeds with the original people, then it becomes an issue," she said.   

What Does AI-Generated Mean? 

The order instructs the Department of Commerce to develop guidelines for content authentication and watermarking to label AI-generated content. Desai questions what it means for something to be truly created by AI.  

An important distinction lies between using AI to assist a writer in organizing their thoughts and using the technology to generate content. He likens the trend to the music industry in the 1980s.  

"Synthesizers really changed people's ability to generate music and, for a while, people thought that was horrible. They can just program the music. They're not. I am still the human responsible for that music, or that article in this case, so what is the point of the label?" he asks. 

As AI assistance becomes commonplace in content creation, trusting the source of information is increasingly important. Recently, articles published on Sports Illustrated's website featured AI-generated content provided by a third-party company that had used a machine to write the content and create fake bylines. Sports Illustrated, which may not have known of the problem, ran the material without disclosure to readers. CEO Ross Levinsohn was ousted shortly after the story broke.  

“Perhaps if the third party had disclosed its use of AI software, SI would have been able to assess how much AI was used and then chosen not to run the material, or to run it with a disclaimer that AI helped write the material,” Desai said. "Of course, even if they label the content as AI-generated, a reader still won't know exactly how much of the content came from AI or a human.” 

AI and the Workforce 

As AI systems and models become more sophisticated, workers may become more concerned about being replaced. To counteract these concerns, the order calls for a study to examine AI’s potential impact on labor markets and investments in workforce training efforts.  

Kumar compares the rise of AI to similar technological innovations throughout history and sees it as an opportunity for workers and industries to adapt. "It's less a matter of AI replacing workers and more of reskilling people to use the new technology. It's no different from when assembly lines in the auto industry were created."  

Promoting Innovation and Competition 

The power to harness the full potential of AI has initiated a race to the top. Desai believes that part of the executive order providing resources to smaller developers can help level the playing field.   

"There is a possibility here for markets to open up. Current players using models that weren't built with transparency in mind might struggle, but maybe that's OK." 

The issue of reliability and transparency comes into focus for Desai, especially as it relates to government usage of AI. The order calls on agencies to "acquire specified AI products and services faster, more cheaply, and more effectively through more rapid and efficient contracting."  

When taxpayer dollars are at stake, government can’t afford to trust a technology it doesn’t fully understand — a topic Desai has explored elsewhere. "You can’t just say, ‘We don’t know how it works, but we trust it.’ That’s not going to work. So that’s where there may be a slowdown in the government’s ability to use private sector software if they can’t explain how the thing works and to show that it doesn’t have discriminatory issues.” 

What's Next 

Promoting and policing the safe use of AI cannot be done independently. Georgia Tech experts agree that participation on a global scale is necessary. To that end, the European Union will unveil its comprehensive EU AI Act, which includes a similar framework to the president's executive order.  

Due to the evolving nature of AI, the executive order or the EU's actions will not be all-encompassing. Law often lags behind technology, but Kosal points out that it's crucial to think beyond what currently exists when crafting policy.  

Experts also agree that AI cannot be regulated or governed through a single document and that this order is likely the first in a series of policymaking moves. Kosal sees tremendous opportunity with the innovation surrounding AI but hopes the growing fear of its rise does not usher in another AI winter, in which interest and research funding fade. 

You may be familiar with yeast as the organism content to turn carbs into products like bread and beer when left to ferment in the dark. In these cases, exposure to light can hinder or even spoil the process. 

In a new study published in Current Biology, researchers in Georgia Tech’s School of Biological Sciences have engineered one of the world’s first strains of yeast that may be happier with the lights on.

“We were frankly shocked by how simple it was to turn the yeast into phototrophs (organisms that can harness and use energy from light),” says Anthony Burnetti, a research scientist working in Associate Professor William Ratcliff’s laboratory and corresponding author of the study. “All we needed to do was move a single gene, and they grew 2% faster in the light than in the dark. Without any fine-tuning or careful coaxing, it just worked.”

Easily equipping the yeast with such an evolutionarily important trait could mean big things for our understanding of how this trait originated — and how it can be used to study things like biofuel production, evolution, and cellular aging.

Looking for an energy boost

The research was inspired by the group’s past work investigating the evolution of multicellular life. The group published their first report on their Multicellularity Long-Term Evolution Experiment (MuLTEE) in Nature last year, uncovering how their single-celled model organism, “snowflake yeast,” was able to evolve multicellularity over 3,000 generations.

Throughout these evolution experiments, one major limitation for multicellular evolution appeared: energy.

“Oxygen has a hard time diffusing deep into tissues, and you get tissues without the ability to get energy as a result,” says Burnetti. “I was looking for ways to get around this oxygen-based energy limitation.”

One way to give organisms an energy boost without using oxygen is through light. But the ability to turn light into usable energy can be complicated from an evolutionary standpoint. For example, the molecular machinery that allows plants to use light for energy involves a host of genes and proteins that are hard to synthesize and transfer to other organisms — both in the lab and naturally through evolution. 

Luckily, plants are not the only organisms that can convert light to energy.

Keeping it simple

A simpler way for organisms to use light is with rhodopsins: proteins that can convert light into energy without additional cellular machinery.

“Rhodopsins are found all over the tree of life and apparently are acquired by organisms obtaining genes from each other over evolutionary time,” says Autumn Peterson, a biology Ph.D. student working with Ratcliff and lead author of the study.

This type of genetic exchange is called horizontal gene transfer and involves sharing genetic information between organisms that aren’t closely related. Horizontal gene transfer can cause seemingly big evolutionary jumps in a short time, like how bacteria are quickly able to develop resistance to certain antibiotics. This can happen with all kinds of genetic information and is particularly common with rhodopsin proteins.

“In the process of figuring out a way to get rhodopsins into multi-celled yeast,” explains Burnetti, “we found we could learn about horizontal transfer of rhodopsins that has occurred across evolution in the past by transferring it into regular, single-celled yeast where it has never been before.”

To see if they could outfit a single-celled organism with solar-powered rhodopsin, researchers added a rhodopsin gene synthesized from a parasitic fungus to common baker’s yeast. This specific gene is coded for a form of rhodopsin that would be inserted into the cell’s vacuole, a part of the cell that, like mitochondria, can turn chemical gradients made by proteins like rhodopsin into energy. 

Equipped with vacuolar rhodopsin, the yeast grew roughly 2% faster when lit — a huge benefit in terms of evolution.

“Here we have a single gene, and we're just yanking it across contexts into a lineage that's never been a phototroph before, and it just works,” says Burnetti. “This says that it really is that easy for this kind of a system, at least sometimes, to do its job in a new organism.”

This simplicity provides key evolutionary insights and says a lot about “the ease with which rhodopsins have been able to spread across so many lineages and why that may be so,” explains Peterson, who Peterson recently received a Howard Hughes Medical Institute (HHMI) Gilliam Fellowship for her work. Carina Baskett, grant writer for Georgia Tech’s Center for Microbial Dynamics and Infection, also worked on the study.

Because vacuolar function may contribute to cellular aging, the group has also initiated collaborations to study how rhodopsins may be able to reduce aging effects in the yeast. Other researchers are already starting to use similar new, solar-powered yeast to study advancing bioproduction, which could mark big improvements for things like synthesizing biofuels.

Ratcliff and his group, however, are mostly keen to explore how this added benefit could impact the single-celled yeast’s journey to a multicellular organism. 

“We have this beautiful model system of simple multicellularity,” says Burnetti, referring to the long-running Multicellularity Long-Term Evolution Experiment (MuLTEE). “We want to give it phototrophy and see how it changes its evolution.”

Citation: Peterson et al., 2024, Current Biology 34, 1–7.

DOI: https://doi.org/10.1016/j.cub.2023.12.044