Hewlett Packard Enterprise Data Science Institute

With more people using artificial intelligence tools like ChatGPT, C. T. Bauer College of Business researchers have found that middle-class workers may be the most receptive to incorporating the tool on the job. 

The rise of large language models (AI models including ChatGPT) have led people to ask what they mean for society, and what benefits they offer, explained Professor and Endowed C. T. Bauer Chair of AI Meng Li. These questions are part of the thought process behind Li’s research, which looks at workers’ social class backgrounds and how that impacts their adoption of LLMs in place of supervisor assistance. 

“We already understand that AI tools, ChatGPT, self-driving cars — they are not going away,” Li said. “They are going to be there regardless of if we like it or not, so just answering this question will be critical for our society.”

In their quest to see the relationship between social classes and adopting AI instead of getting help from supervisors, researchers did large scale surveys and behavioral experiments. This included early career professionals from various social class backgrounds.   

The paper, co-authored by Li, Bauer Ph.D. student Yao Yao and Boston College Assistant Professor Lai Wei, defines “early career” as workers with less than two to three years of experience. This group was selected because of their typical reliance on their supervisors, and because their relatively standardized current social class allows for more context to examine their social class background. 

The research insights found that middle-class workers are more willing to use LLMs instead of asking their supervisor for help compared to their lower-class and upper-class peers. 

“For the upper class, they are comfortable to talk to humans,” Li said. “They have the resources. For the lower class, they don't have the literacy or the knowledge of large language models. So, it turns out the middle class is more willing to adopt AI tools such as large language models.”

The results could be seen as an inverted U-shaped pattern and bring a light to the middle-class workers. The middle-class group stood out because they were most inclined to use the LLMs in this way. 

“They have the knowledge,” Li said. “They know how to use it. They know how the large language model can help them, or they are comfortable with technology. I think this unique advantage makes their adoption easier.” 

Li says it’s important to understand that LLMs impact people differently. 

“Trying to help the people who are not adopting AI, or have the need to, like the lower class, but don't know how to use it,” Li said. “I think finding a way to help them will be important.” 

Moving forward, how these dynamics will impact workplace inequality is a question the paper says is still up for future research. 

The next step to continue advancing research on how AI impacts the workplace is an exploration of how these dynamics may impact workplace inequality, Li said. 

On Thursday, April 18, 2024, the University of Houston Hewlett Packard Enterprise Data Science Institute hosted an online lecture titled "Computational Genomics Reveal Insights into the Biology of Non-model Organisms." Richard Meisel from UH's Department of Biology and Biochemistry presented his research on applying genomic methods to non-traditional study organisms. Meisel highlighted how recent technological advancements have made genomic analysis more accessible across various biological fields.  

The lecture focused on projects from Meisel's lab, a key example being their work on the house fly, Musca domestica, which possesses a complex sex determination system. Meisel’s team used advanced sequencing technologies and computational methods to assemble a high-quality genome for the house fly. This improved genome assembly allowed them to investigate the genetic basis of sex determination in this species, revealing variations in the male-determining chromosomes across different populations. The research uncovered how environmental factors, particularly temperature, play a role in the distribution of different male-determining chromosomes.  

Another significant aspect of the talk was the demonstration of how computational genomics can bridge the gap between model and non-model organisms. Meisel’s team used data from the well-studied fruit fly, Drosophila melanogaster, to inform experiments on the house fly, leading to the identification of genes involved in mating behavior.  

Throughout the presentation, Meisel emphasized the importance of integrating various approaches, including computational genomics, organismal experiments and comparative studies between model and non-model organisms. This multifaceted strategy allows researchers to address fundamental questions in biology from both mechanistic and evolutionary perspectives. Meisel highlighted the technical challenges of genomic assembly, such as dealing with repetitive DNA sequences. Long-read sequencing technologies and novel computational approaches have helped overcome these obstacles.

Richard Meisel, Ph.D. completed his postdoctoral training in the Department of Molecular Biology and Genetics at Cornell University. The Meisel lab uses lab experiments, genomic data and molecular biology techniques to study a broad range of questions in evolutionary biology. The research addresses how environmental variation and sex differences influence genetic and phenotypic diversity within populations and between species. Meisel’s lecture is part of an ongoing series hosted by the UH Hewlett Packard Enterprise Data Science Institute, aimed at showcasing cutting-edge research that leverages high-performance computing resources for scientific discovery. 

On March 21, 2024, the University of Houston Hewlett Packard Enterprise Data Science Institute hosted an online research lecture titled "Accelerated Materials Design Driven by Machine Learning" featuring Mingjian Wen, an assistant professor of Chemical and Biomolecular Engineering at UH's Cullen College of Engineering.  

Wen’s lecture highlighted two main challenges in creating reliable artificial intelligence models for materials science: uncertainty quantification for neural network models and the incorporation of prior physical knowledge and constraints into such models.  

Wen's research employs graph neural networks to model the elastic properties of materials. From the predicted elastic tensor, one can obtain various mechanical properties, such as bulk modulus (resistance to volume change) and shear modulus (resistance to sliding deformation) across different crystal systems. This capability is particularly valuable because many materials exhibit anisotropy, behaving differently depending on the direction of applied force.

Wen stressed the importance of incorporating physical constraints and uncertainty quantification into these models. He explained how a dropout technique can be used to assess the reliability of predictions, especially when extrapolating beyond the training data. This approach allows researchers to identify when a model might be unreliable for certain materials or conditions.

The lecture highlighted practical applications of these AI-driven methods. Wen demonstrated how the models could accelerate materials innovation by rapidly screening thousands of potential candidates and identifying those with exceptional properties such as strength. Wen emphasized that the true power of these methods lies in combining traditional scientific knowledge with machine learning techniques. This synergy between human expertise and AI tools has the potential to revolutionize materials design and discovery across various industries, creating more effective and physically-justified models for materials science.

Mingjian Wen, Ph.D is a Presidential Frontier Faculty Fellow at the University of Houston. His current work focuses on developing data-driven approaches for computational understanding and design of energy materials and contributing to open cyberinfrastructure for materials modeling. Prior to UH, Wen worked as a postdoctoral scholar on the Materials Project at Lawrence Berkeley National Lab. Wen's lecture is part of an ongoing series hosted by the HPE Data Science Institute, aimed at showcasing cutting-edge research that leverages high-performance computing resources for scientific discovery. 

Machine learning techniques are revolutionizing nuclear data evaluation and reactor design. Vladimir Sobes, assistant professor in the Department of Nuclear Engineering at the University of Tennessee, Knoxville, has dedicated his career to exploring the application of modern artificial intelligence and machine learning algorithms to current problems in nuclear engineering. 

In the realm of nuclear data evaluation, Sobes underscored the significance of ML algorithms in automating tasks such as hyper-parameter tuning and learning complex functions, particularly in the realm of uncertainty quantification. Sobes illustrated the structured nature of nuclear data, stressing the importance of automated feature identification—a task traditionally performed manually by experts. ML/AI, he argued, offers the promise of enhancing efficiency, reproducibility, and uncertainty quantification in nuclear data evaluation, thereby enabling experts to devote their time to more critical tasks. 

Transitioning to the topic of reactor design, AI has the potential surpass human capabilities in autonomous optimization, although Sobes cautioned against the indiscriminate use of surrogate models.  

Sobes demonstrated the effectiveness of the Gaussian Process Learning Algorithm for design optimization using a scenario related to the development of a spacecraft antenna. NASA engineers asked an AI algorithm to imagine an antenna. Despite an initial skepticism of its unintuitive design, rigorous testing demonstrated superior performance compared to human-designed alternatives. In 2006, the NASA ST5 spacecraft antenna, the same model proposed by an evolutionary computer design, was sent to space. 

AI/ML technologies are reshaping the landscape of nuclear engineering, from streamlining nuclear data evaluation processes to pushing the boundaries of reactor design beyond what was previously thought possible. Sobes' insights highlighted the symbiotic relationship between human expertise and AI, emphasizing the potential for collaboration to drive innovation in nuclear science. 

The presentation concluded with a quote by Arthur C. Clarke, adapted to signify the transformative potential of advanced optimization methodologies: "Any sufficiently advanced optimization methodology is going to be indistinguishable from intelligence."  

Sobes envisions a future where AI supplements human expertise, enabling breakthroughs in nuclear science and technology.