Managing Director & Computer Science PhD Student @ The Diasporic Group & Cornell University
Sponsor
1361
12
Similar sessions
Share this session
This Session has closed
About Chinasa T. Okolo
Chinasa T. Okolo is a Computer Science Ph.D. student at Cornell University. Before coming to Cornell, she graduated from Pomona College with a degree in Computer Science. Her research interests include computer vision, human-AI interaction, global health, and ICTD. Within these fields, she leverages recognition techniques to improve the rapid diagnosis of infectious and tropical diseases and analyzes the applications, implications, and perceptions of AI-enabled technologies in the Global South.
Below is a summary of some of the interesting things Chinasa is involved in.
HID Health Incubation Intern @ Apple, May 2021 - Present
Biological Computation Research Intern @ Microsoft, May 2018 - Aug 2018
Investor @ HoaQ (hoaq.club), 2020 - Present
Managing Partner & Founder @ The Diasporic Group, 2020 - Present
Strategic Advisor @ Doing Good Work in Africa (DOWA), 2020 - 2021
INTERVIEW
QUESTIONS
These interview questions were brought to you by TwoCents and Blessing Guembe
TwoCents
What, in your opinion, are the most important factors driving machine learning demand?
Chinasa T. Okolo
"Most important" is definitely relative. I believe that nowadays, many companies have so much data they've been collecting from customers over the years and ML is a great way to derive insights in a much faster way. For non-commercial applications like academic research, I find similar motivations where people want to use ML to find new ways to interpret and derive insights from their data.
TwoCents
There have been numerous publications regarding machine learning being harmful in terms of its decision logic.Since it is almost impossible to justify their decision logic because they primarily edit different features by trial and error until they find the ideal decision. Are you in agreement with this viewpoint? Do you believe that explainable artificial intelligence will be able to bridge this gap?
Chinasa T. Okolo
I definitely believe that explainable artificial intelligence is the next big step in AI/ML! As these technologies become more integrated within our lives, practitioners have to ensure that the people using these systems actually understand them. This is super important for high risk domains like healthcare. Additionally, I believe human-centered AI/ML will become even more important and is an area that I am actively pursuing in my work.
TwoCents
In the machine learning domain, how important is statistics, linear algebra and calculus? As a beginning, what topics should I cover within these sub-fields of mathematics?
Chinasa T. Okolo
I believe all of these topics are extremely important since they provide the fundamentals of machine learning algorithms. In linear algebra, you want to ensure that you focus on matrices and vectors, notation, and operations. For statistics, I believe combinatorics, probability, Bayes' Theorem, variance and expectation, and conditional, joint, and standard distributions are all important. There's so many things to cover within all of these math topics and I hope this helps!
TwoCents
What are the benefits of machine learning for small firms, and what will be the most significant impact of machine learning on small enterprises in the next five years?
Chinasa T. Okolo
If small firms have the necessary data and infrastructural capacity to train ML algorithms, then they can benefit from added insights on their customers and respective operations. I don't work with small enterprises, so it's a bit hard for me to detail what the most significant impact will be.
TwoCents
What industries are actively utilizing machine learning capabilities to increase their product efficacy, and how advanced is machine learning technology today?
Chinasa T. Okolo
Pretty much all industries from healthcare to education to manufacturing! ML technologies have definitely come a long way from where they were in the 1960s but there is still a long way to go. In terms of "artificial intelligence", we still see a huge gap between humans and AI. For example, Tesla's autopilot system is susceptible to false alarms that humans wouldn't experience and hasn't been able to account for the large amount of edge cases that can occur in driving. Tesla is doing a great job but I feel that we are a long way from a true autopilot. The same with other fields like medicine where AI is being used in a variety of ways but still doesn't beat a doctor when all things are considered.
TwoCents
There are so many machine learning techniques today, and if given a real-world project, how can one determine which technique to utilize?
Chinasa T. Okolo
Working your way backwards is a really great step to understanding what ML techniques to use? Are you working with images or videos? If so, use computer vision. Working with text? Then use NLP! You could go on and on with this process, breaking down the techniques further and further based on the specifications of your problem and the desired results.
TwoCents
Currently what is the biggest highlight of machine learning? In five years what will machine learning look like?
Chinasa T. Okolo
The biggest highlight for me is the ability to save lives and also democratize access to healthcare, education, agriculture and more! AI/ML has so much potential to improve the lives of those living in under resourced regions but it has to be leveraged properly. As the field is changing so fast, it's impossible to make solid predictions but I hope to see the field of ML become more inclusive and readily accessible to those without expensive equipment. We'll definitely see more industries take advantage of these technologies and AI/ML becoming more ubiquitous in human life for sure!
TwoCents
What are the skills required in machine learning and how can I develop them, and what are the tools required for beginners to kickstart a machine learning career?
Chinasa T. Okolo
There are so many possible skills used in machine learning that it's somewhat hard to answer this question. Obviously having a strong mathematical foundation helps, so brushing up on your linear algebra and statistics will be helpful. You can develop these skills by hard work through self-studying, taking courses on online platforms like Coursera, Udemy, and LinkedIn learning. Nowadays, there are so many blogs and resources that are a quick Google search away! As a PhD student, my career in machine learning has been mostly research based, so I am unfamiliar with the tools "required" to kickstart a machine learning career. Again, this will depend on the type of career you pursue (academic vs. industry).
TwoCents
What are the components of a successful machine learning deployment, as well as the infrastructure required to enable them?
Chinasa T. Okolo
A typical answer would probably state that "good" data and an well-built model would be components of a successful machine learning deployment, but I believe that the more qualitative sides of ML development are much more necessary. As issues of bias, fairness, and explainability have become more prominent in AI/ML development, ensuring that your systems are usable and equitable for all populations is extremely important. Bias can be encoded both in data and the models themselves, so it's also important that this be prioritized for successful ML deployment.
Obviously, you'll need GPU/TPU capacity to efficiently train your algorithms, but this is definitely a barrier for those coming from low-resource regions. Efforts like Google Colaboratory have helped somewhat but can be challenging to use if you don't have sufficient electricity or internet access.
Lagos, Nigeria
Hello Chinasa,
Tell us a bit about the work you do at Apple as a Health Incubation Intern, and about the work you did at Microsoft as a Biological Computation Research Intern.
How advanced is the current state of AI as regards assisting medical practitioners with the diagnosis of infectious diseases? As an example, could cutting-edge AI correctly identify a particular type of skin disease and even proffer treatment just by processing a scanned image?
Lastly, Do you share Elon Musk's concern about "general" or "strong" AI with an open-ended utility function? In your opinion, what is likely to be the first mainstream product powered by "general" or "strong" AI?
4 Answer requests
Hi Cyrus, I'm unable to talk much about my work at Apple but I am currently developing new ways to leverage Responsible AI methodologies in health sensing applications. At Microsoft Research, I developed domain-specific computational models for bacterial quorum sensing which is how bacteria communicate with each other.
I would say that the current state of AI is pretty advanced with regards to the specific techniques but many of these implementations are not production-level ready or haven't been tested in real-life clinical settings to actually be of any use. Yes, what you described is possible but there still remain many issues with bias in these and other types of applications.
Hi Chinasa, hope you are well? My question revolves around prospects for those migrating to ML from social science fields. Specifically, what are the possibilities of obtaining a masters and hopefully a PhD in Machine Learning despite being a Political Science major at undergraduate level?
3 Answer requests
There is a good chance that you will be able to do this, you will most definitely need to start off with a Master's program then transition to a traditional PhD program. Many programs favor students with relevant degrees (Computer science, electrical/computer engineering, math, etc.) for admission into their programs since they have the necessary background to understand the basics of computer science.
Hello Chinasa, so recently I've been hearing a lot about machine intelligence. Is this the same thing as machine learning? And if not, what is the difference?
3 Answer requests
I would say machine learning is the process that leads the way to machine intelligence!
- What is the possibility of conducting ML experiments (especially in the area of biomedical Image Analysis) without access to high-end Gpu or computing equipment and how can one overcome this challenge?
- How can young researchers with a trained model deploy it into production?
- What were your struggles if you faced any in building AI-based systems or starting out in the field of AI and how you overcame it.
- How do you get to determine what real-life problem that will require an AI solution?
3 Answer requests
Fortunately, some large tech companies have begun to democratize access to compute and offer free cloud services. I would check out Google Colaboratory that offers free GPU/TPU access to train models.
There's no specific rule for determining what real-life problems can be solved with AI, if you have sufficient data and a solid understanding of what models can be used to solve the problem, then you can test it out!
I have a group of questions.
1. How did you get there like what major decisions did you make why you did them (these are career choices)
2. What advice do you have to give to a comp science year 3 student?
0 Answer requests
A major decision for me was majoring in Computer Science, I always had an interest in biology and medicine but knew that CS could be leveraged to enhance these fields even further. Also choosing to focus on research vs. software engineering internships in undergrad really gave me a strong research foundation that helped me in my successful admission to PhD programs!
As for advice, if you want to do a PhD, I would suggest you start looking into research opportunities or programs on campus as soon as possible or look into AI residency programs. It's hard to admit students into PhD programs without research experience and I find that many students put this off until later.
Hi
I'm working on a project to advise farmers on climate smart crops, aside from identifying diseases in plants, is there a way that computer vision can be used in a small scale school project to achieve this considering limited access to datasets and have you worked on any cool project involving computer vision.
0 Answer requests
Yes, this is a image recognition process that has been done in multiple ways before. I would suggest looking up relevant literature for inspiration and browsing around competition sites like Kaggle or Zindi to get access to datasets!
The coolest project I have worked on probably was something I did with a startup to recognize different hairtypes and hairstyles!
Over the last couple of years, ML modelling skills have become ubiquitous such that anyone can literally create a regression model, or some classification model (i.e, model.fit()). However, I feel the needs in the industry are broader. In recent years we have heard about model deployment alot. I'd like it if you can talk about relevant knowledge and skills that make the difference in the industry.
Secondly, for someone in a backgroud in Engineering say Operations Research, what are some research areas in Applied ML one can look into for a PhD, that may eliminate some of the more mathematical aspects of ML and focus more on Application?
0 Answer requests
Industry is very broad since there are so many different applications large tech companies are working on. The needs are pretty much the same, it just depends on what specific project it is (ads targeting, email autocompletion, heart rate detection, etc.). I would say the major difference is learning how to scale these algorithms for use across thousands, maybe millions of users but this has pretty much been figured out.
Pretty much everything out of theoretical ML is applied, so if you're interested in working with text data pursue projects in NLP or if you like image + video data, pursue Computer Vision!
What should I be worried about most when thinking of applying machine learning techniques in solving a problem.
0 Answer requests
I think the first question should be: does this problem actually need ML?
Hi. As an expert in machine learning, I would really love to hear your thoughts on data-centric machine learning, especially as it relates to the concepts of fairness, security, and explainability. Thanks.
0 Answer requests
You can say that all ML is data-centric since this is the basic facet machine learning models need to function. Fairness is a growing topic that is fortunately receiving much needed attention. Companies and individuals are realizing the harms that ML can pose and are actively working to address this. Explainability is also a big factor in making ML models fair but is critically understudied, especially in the context of making ML "explainable" to those with lower technology or AI knowledge. Security can be a big issue but is usually handled on the networking side of things (data storage, transmission, etc.).
Hello Chinaza, I am a master's student concentrating in software engineering, I have tried to break into the ML field but faced challenges especially with my math background, what would you advice I do to close the skill gaps, or are there ways I could find a niche for myself between ML and Software Engineering
0 Answer requests
I answered this somewhat in a previous question, so I'll take a bit of my answer from there:
I haven't worked as a Software Engineer, but ML engineers are being heavily recruited. You should look up a few of these job roles and see what the requirements are, this could help shape your priorities for what to focus on in your learning journey!
do you think we always need to apply machine learning to every problem we are trying to solve using data or can we fully optimize the value of a data set without applying machine learning? if so, can you give a real life experience, if any?
0 Answer requests
No I don't! I think today, ML is being used a crutch to solve major problems when simple solutions are much more effective. For example, if you talk about leveraging ML to create better educational outcomes but you don't have enough schools or even electricity to run these systems, there is obviously a mismatch in priorities.
Hello Chinasa, glad to have you here. Lately, the area of quantization is beginning to gain traction in deep neural networks (DNN), what's your take on this exciting new domain and do you think the tradeoff between the marginal drop in accuracy for a reduced model size is worth it?
Also, with my experience in DNN, I noticed that most people see DNN as a one stop point to solve almost any classification or Regression tasks (This should normally not be the right thing to do, as one may be going to a stick fight with a bazooka), so my question is this, at what point do we draw the line between opting for a more traditional ML algorithm/technique or diving right into using neural nets.
0 Answer requests
I have no experience in quantization, so I won't be able to answer this well. However, I believe it is important that ML be as accessible as it possibly can be and am looking forward to seeing improvements within this domain.
I think it's totally fine to explore as many techniques as you can since you don't know what works until you try it!
's Wall
Getting answers & insights from doesn't stop with this session.
Beyond sessions, TwoCents makes it possible for you to keep on getting answers and insights from through the Wall feature. (Coming soon)
The Wall is a section of 's profile where you can ask questions and get answers, as well as read insights.
This feature is only available to users who subscribe to 's Wall.
If you participated in this session, TwoCents will notify you once this feature is available.
This session is a program related session.
What does this mean?
Program related sessions are hosted in conjuction with our partners to facilitate easy question curation and answering.
We give our host partners the opportunity to use TwoCents to take questions during a program, conference or online class or meeting they are conducting.
Interested in becoming a host partner?
We'd love to have you on board!. Reach out to us via any of the channels below
Question
Question
Question
Share this answer to your feed
TwoCents
Chinasa T. Okolo
Share this answer to your feed
Chinasa T. Okolo
Comments