作为埃默里人工智能学习中心的一名学生，大三学生alisha morejon正在使用机器学习来分析乌干达学生画的艺术品。 一旦提取出有意义的模式，就可以设计一种干预措施来提高发展中国家的识字率和入学率。 图中（左-右）是瓦莱达·登特、托米·奥托林、艾丽莎·莫雷洪和杰夫·古德曼。 --贝基斯坦摄影。 如果大三学生alishamorejon在成为emory人工智能学习中心的一名学生时被告知，在不到一年的时间里，她将共同撰写一篇研究论文，并前往加拿大在一次国际会议上展示她的工作，她永远不会相信。 “我仍然对我作为一名学生所接触到的所有新经历感到惊讶。 它真的把我带到了我从未想象过的地方，并在这个过程中教会了我很多关于自己的东西。 为了结束她在人工智能学习中心工作的第一年，morejon计划今年夏天前往渥太华参加心理治疗研究学会的年度国际会议。 在那里，她将帮助发表论文“使用机器学习探索乌干达儿童的入学准备情况”，这对本科生来说是一个难得的机会。 人工智能学习中心于2023年秋季开业，是埃默里人工智能的关键组成部分。 人性倡议全企业致力于塑造未来的道德人工智能，为人类服务。 作为在埃默里建立人工智能知识和社区计划的一部分，该中心为埃默里的学生提供体验式学习机会，为他们提供将人工智能概念应用于现实世界项目的实践经验。 当被雇佣时，学生会报名参加一个每周10-15小时的带薪职位。 在伍德拉夫图书馆的中心所在地，他们会花一定的时间迎接来访者，并回答有关中心及其服务的问题。 每个学生都被分配到体验式学习项目中，在这些项目中，他们直接与赞助商合作，使用人工智能解决问题。 他们还被要求每学期亲自教授一个选定主题的研讨会。 该中心的高级项目协调员tommy ottolin与项目赞助商valeda f进行了配对。 图书馆和博物馆副教务长dent和埃默里大学医学院精神病学和行为科学教授、坎德尔神学学院心理学和精神关怀副教授geoff goodman。 dent和goodman联系了该中心，帮助开展一个项目，该项目源于他们在早期扫盲干预方面的长期研究合作。 他们的项目“一个故事生长在乌干达”已经从乌干达农村的3到5岁的儿童那里创作了40000多幅绘画作品。 他们想利用人工智能来帮助他们对大量的艺术品进行分类，并想出一些模式，让他们深入了解孩子们的发展和入学准备情况。 “我认为tommy之所以选择让我参与这个项目，是因为我对人文学科感兴趣，”未修宗教专业的morejon说。 “我倾向于着眼于大局和战略，而不是超级技术方面。 我分析了所有现有的研究，从不同的角度看，然后说，‘嘿，我们想过这个吗？’”每张图纸都需要基于特征的编码，这是一个非常耗时的前景。 目标是使用深度学习模型自动编码这些图片。 morejon和她的另一位有着丰富编程经验的同学遇到了一些障碍。 大约有200张图纸已经由dent和古德曼团队的研究人员进行了分析和编码——这是一个相对较小的训练数据样本。 他们设计的过程包括使用监督学习，即向模型提供70%的带有标签的训练数据，然后提供30%的未编码图纸来测试准确性。 到目前为止，结果并不完美。 莫雷洪说：“我们仍处于试错阶段。”。 但潜在的短期和长期影响值得它们坚持下去。 一旦实现，这个项目将使编码数千张图纸的艰巨任务成为可能。 此外，随着团队不断调整模型以检测新的图案，以数字方式存储每张图纸的数据将是一项巨大的资源。 如果提取出有意义的模式，就可以设计一种干预措施来提高发展中国家的识字率和入学率，这就是人工智能面临的现实挑战。 人性是为了解决。 morejon不仅学习了机器学习方法，还学习了自己。 “我玩得很开心。 她说：“在这个项目中发现交叉性真的很有趣。”。 许多好处该中心的学生研究员项目的美妙之处在于，学习远远超出了项目本身。 ottolin称这些项目为“真实世界的冒险”，他解释说，每个项目都包含了学生积极参与中介关系、管理项目进展、与客户沟通和故障排除的期望。 这些非技术性技能对于培养学生在快速变化的世界中蓬勃发展同样重要，这是学生蓬勃发展倡议的关键原则。 项目请求以多种方式处理：电子邮件、电话、预约。 一旦一个项目被中心工作人员评估为学生主导的项目的可行性，项目建议书就完成了。 该文件包括项目发起人的详细信息，哪些学生将被分配到项目中，他们的角色是什么，项目目标和成功的衡量标准。 鉴于该中心及其研究生项目相对较新，项目的设置和实施方式也在不断调整。 “我告诉我们采访的每一个学生，我们边走边学。 这真是个很酷的工作场所。 如果你能灵活、透明，并准备好推动变革，那么这将非常适合你，”ottolin说。 ottolin补充道，项目发起人明白这是学生的学习机会，有时也包括挫折和失败。 “我们正在研究新技术并回答新问题。 有时事情并不像我们最初想象的那么简单。”ottolin说。 “我必须把它交给赞助商。 他们是很好的合作伙伴，非常了解我们想要为这些学生营造的环境。 吸取的教训也可以是不该做什么的教训。 “在中心，我想运用我的技能，但我也想让自己沉浸在人工智能的世界中。 随着我们日常生活中技术的普及，保持领先是很重要的。 我很欣赏emory推动人工智能教育的倡议。 dylan parkermaking the theory tangib该中心体验式学习项目的另一个不可否认的好处是将理论转化为有形事物的前景。 计算机科学专业的迪兰·帕克和拉斐尔·帕拉西奥正在为埃默里监察员办公室做一个聊天机器人项目。 帕拉西奥还选修了《cs329：计算语言学》，帕克现在正在选修；虽然该课程的主要项目之一是设计和构建聊天机器人，但帕克表示，他很感激有机会在课堂之外运用自己的技术。 “监察员聊天机器人项目很好地补充了我在课堂上学到的东西。 palacio说：“这学期我将介绍人工智能，所以我可以学习人工智能的理论方面，然后应用这些知识。”。 解决问题，节省资源通过运用所学技能，学生也有机会通过解决真实的问题产生持久的影响。 这正是郑同学在为埃默里大学莫兰实验室及其种群生物学、生态学和进化项目工作时所做的。 莫兰实验室的研究人员使用c。 用于共同进化和交配系统研究的秀丽隐杆线虫。 研究人员必须每周在显微镜下手动计数和分类数千种蠕虫。 这个繁琐的过程可能需要几个小时甚至几天的时间，而且错误率变化很大。 郑是生物学和计算机科学双专业的学生，他的任务是创建一个深度学习模型，使蠕虫的计数和分类过程自动化，同时最大限度地减少错误。 郑说她在这个项目中遇到了两大障碍。 首先是缺乏足够的训练数据，这是人工智能模型构建中的一个常见主题。 在这种情况下，现有数据将训练模型来分析模式、特征和阈值。 郑和她的另一位ai研究生伙伴不得不通过从其他来源寻找可访问的数据集来缓解数据短缺的问题。 另一个挑战是目标检测和卷积神经网络（cnns）中的自主学习过程。 郑在这些方面几乎没有经验，所以她和她的搭档必须从头开始学习底层逻辑、数学和实现。 “自我教育过程是艰难的，有时令人沮丧。 这需要很大的耐心和毅力，因为我们必须消化复杂的概念并将其应用到我们的工作中，”郑说。 “这个问题没有一个简单的解决方案，但我确实通过打磨学会了，现在我有了一套全新的技能，这将在未来为我服务。”。 “结果是一种工具，可以为研究小组每周节省四个小时的计数时间，让他们有时间花在研究的其他方面。 自c。 秀丽隐杆线虫是实验室科学中最常见的生物之一，其他研究人员可能会使用这种新工具来节省时间、视力和消除人为错误。 该项目在人工智能、编码、专业沟通和团队合作方面进行了深入的学习。 它扩展了我对机器学习的局限性、实际应用和影响的理解。 我喜欢从头开始构建解决方案，将理论与实践相结合是有价值和令人满意的。 iris郑中心体验式学习的未来随着夏季的临近，中心的体验式学习机会正在扩大。 该中心与定量理论与方法系等学术单位以及cs370：计算机科学实践课程llm、数据分析和可视化等课程合作，提供课外和课堂体验式学习。 亚特兰大市的技术和创新办公室以及更多的合作伙伴关系也带来了外部机会。 “当我们考虑人工智能时。 人性化倡议，我们正在努力将我们的学生培养成最全面的人工智能专业人士。 ottolin说：“所以这不仅仅是一个人在一个房间里编写代码，我们想让他们了解团队合作、培养客户关系、管理项目、解决问题的全部策略。”。 “在解决这些问题的同时，他们正从道德和公平的角度看待这个过程。 这就是emory和该中心的体验式学习项目为人类带来的学习、成长和改变的机会。 “对探索体验式学习感兴趣，或者对某个项目有想法？联系ai学习团队中心。 “对探索体验式学习感兴趣，或者对某个项目有想法？联系ai学习团队中心。
as a student fellow at emory’s center for ai learning, junior alisha morejon is using machine learning to analyze artwork drawn by ugandan school children. once meaningful patterns are extracted, an intervention could be designed to increase literacy outcomes and school readiness in the developing world. pictured (l-r) are valeda dent, tommy ottolin, alisha morejon and geoff goodman.— photos by becky stein photography.if junior alisha morejon was told when she became a student fellow for emory’s center for ai learning that in less than a year, she would be co-authoring a research paper and traveling to canada to present her work at an international conference, she never would have believed it.“i’m still surprised at all the new experiences i’ve been exposed to as a student fellow. it really has taken me to places i never imagined and taught me so much about myself in the process,” says morejon, who is a joint computer science and mathematics major.to cap off her first year working at the center for ai learning, morejon plans to travel to ottawa this summer for the annual international meeting for the society for psychotherapy research. there she will help present the paper “using machine learning to explore ugandan childrens school readiness”a rare opportunity for an undergraduate student.the center for ai learning, which opened its doors in fall 2023, is a key component of emory’s ai.humanity initiative an enterprise-wide commitment to shaping the future of ethical ai to serve humanity. as part of its programming to build ai literacy and community at emory, the center offers experiential learning opportunities to emory students, providing them with practical experience applying ai concepts to real-world projects.when hired, student fellows sign up for a paid position of 10-15 hours per week. a certain number of hours are spent in the center’s physical location in woodruff library where they greet visitors and answer questions about the center and its offerings. each student is assigned to experiential learning projects in which they work directly with sponsors to solve a problem using ai. they are also required to teach an in-person workshop on a chosen subject each semester.the center’s senior program coordinator, tommy ottolin, matched morejon with project sponsors valeda f. dent, vice provost of libraries and museum, and geoff goodman, professor of psychiatry and behavioral sciences in emory university school of medicine and associate professor of psychology and spiritual care in candler school of theology.dent and goodman approached the center to help with a project that grew out of their long-term research collaboration in early literacy intervention. their project, “a story grows in uganda,” has generated over 40,000 drawings from children in rural uganda aged three to five. they wanted to use ai to help them categorize the vast trove of artwork and come up with patterns that could give them insight into the children’s development and readiness for school.“i think tommy chose to have me work on this project specifically because of my interest in the humanities,” says morejon, who is a religion minor. “i tend to look at the big picture and strategy over the super technical aspects. i’m the one who analyzed all the existing research, looked at different angles and said, ‘hey, did we think about this?’”each drawing requires feature-based coding, an extremely time-consuming prospect. the objective was to automate the process of coding these pictures using deep learning models.morejon and her partner, another student fellow who had more programming experience, did encounter several roadblocks along the way.around 200 drawings came to them already analyzed and coded by researchers on dent and goodman’s team a relatively small sample of training data. the process they designed involves using supervised learning in which the model is fed 70% of the training data with labels and then given 30% of uncoded drawings to test for accuracy. the results so far have been less than perfect.“we are still in the trial-and-error phase,” says morejon.but the potential short- and long-term impacts are worth their persistence. once realized, this project will make the herculean task of coding thousands of drawings achievable. in addition, storing the data of each drawing digitally will be a great resource as the team continues tweaking the model to detect novel patterns. if meaningful patterns are extracted, an intervention could be designed to increase literacy outcomes and school readiness in the developing world and that’s the kind of real-world challenge that ai.humanity was designed to address.morejon learned not only about machine learning methodology but also about herself. “i was having a lot of fun. finding the intersectionality in this project has been really interesting,” she says.a multitude of benefitsthe beauty of the center’s student fellows program is that the learning goes far beyond the projects themselves. ottolin, who calls the projects “real-world adventures,” explains that built into each is the expectation that the students actively participate in brokering relationships, managing the evolution of the project, communicating with clients and troubleshooting. these non-technical skills are equally important in preparing students to flourish in a rapidly changing world a key tenet of the student flourishing initiative.project requests are fielded in many ways: emails, calls, walk-ins. once a project has been evaluated by the center staff for feasibility as a student-led project, a project proposal document is completed. the document includes details on the project sponsor, which student fellows will be assigned to the project and what their roles will be, project objectives and measures of success.given that the center and its student fellows program are relatively new, there are constant adjustments being made to how projects are set up and carried out.“i tell every student we interview that we are learning as we go. this is a really cool place to work. if you can be nimble and transparent and be ready to incite change, then this would be a good fit for you,” says ottolin.ottolin adds that project sponsors understand that this is a learning opportunity for students and that sometimes includes setbacks and failures.“we’re working with new technologies and answering new questions. sometimes things aren’t as straightforward as we originally thought,” says ottolin. “i have to hand it to our sponsors. they’ve been great partners and have been very cognizant of the environment we want to build for these students. lessons learned can be lessons about what not to do, too.”at the center, i wanted to apply my skills but i also wanted to immerse myself in the world of ai. with the prevalence of technologies that are part of our everyday life, it’s important to stay ahead of the times. i appreciate emory’s initiative to push forward ai education.dylan parkermaking the theoretical tangibleanother undeniable benefit of the center’s experiential learning projects is the prospect of turning the theoretical into something tangible.student fellows dylan parker and raphael palacio, both computer science majors, are working on a chatbot project for the emory ombuds office. palacio has also taken cs 329: computational linguistics, and parker is taking it now; while one of the major projects in that course involves designing and building a chatbot, parker says he appreciates the opportunity to apply his technical skills outside of the classroom.“the ombuds chatbot project is a good complement to what i’m learning in class. this semester i’m taking intro to ai, so i get to learn the theoretical side of ai and then get to apply that knowledge,” says palacio.solving problems, saving resourcesbeyond applying learned skills, students also have the chance to make a lasting impact by solving an authentic problem.that is just what student fellow iris zheng did in her work for emory’s morran lab and its population biology, ecology and evolution program. researchers in the morran lab use c. elegans worms for coevolution and mating systems research. researchers must manually count and classify thousands of worms under a microscope each week. this tedious process can take hours or even days and has a highly variable rate of error. zheng, a double major in biology and computer science, was given the task of creating a deep learning model that automates the process of counting and classifying the worms while minimizing errors.zheng says she faced two big hurdles during the project. the first was lack of adequate training data, a common theme in ai model building. in this instance, existing data would train the model to analyze patterns, features and thresholds. zheng and her partner, another ai student fellow, had to mitigate the data shortfall by finding accessible datasets from other sources.another challenge was the self-guided learning process in object detection and convolutional neural networks (cnns). zheng had little experience in these areas, so she and her partner had to learn about the underlying logic, math and implementation from the ground up.“the self-education process was arduous and at times, frustrating. it required a great deal of patience and persistence, as we had to digest complex concepts and apply them to our work,” says zheng. “there wasn’t an easy solution to this problem, but i did learn through grinding and now i have a totally new skill set that will serve me in the future.”the result is a tool that saves the research team four hours of counting each week, freeing them up to spend time on other aspects of their research. since c. elegans are one of the most common organisms used in lab science, the new tool could potentially be used by other researchers to save time, eyesight and eliminate human error.the project was an intensive learning experience in artificial intelligence, coding, professional communication and teamwork. it expanded my understanding of machine learnings limitations, practical applications and impact. i enjoyed building a solution from scratch, and blending theory with practice was valuable and gratifying.iris zhengthe future of experiential learning at the centeras summer approaches, the center’s experiential learning opportunities are expanding. the center has partnered with academic units like the department of quantitative theory and methods and courses like cs 370: computer science practicum course – llm, data analytics and visualization to offer co-curricular and classroom experiential learning. there are also external opportunities through the city of atlanta’s office of technology and innovation and more partnerships on the way.“when we think about the ai.humanity initiative, we are trying to shape our students into the most well-rounded ai professionals. so that’s not just writing code in a room by yourself we want to expose them to the full gambit of working on a team, nurturing a client relationship, managing a project, solving problems,” says ottolin. “and while solving these problems, they are looking at the process through a lens of ethics and fairness. that’s what emory and the center’s experiential learning projects bring to the table a chance to learn, grow and make a difference to humanity.”interested in exploring experiential learning or have an idea for a project? reach out to the center for ai learning team. .
.jpg){kind=logo}
