Yes, we can build a better future with Artificial Intelligence – But only if we have more female leadership.
We sat together with five Women in AI who are doing inspiring work and talked about why we need more women in leadership roles, how to empower more women to join the field, and lastly what learnings and insights all five want to share with aspiring female AI engineers. Enjoy!
Yemissi B. Kifouly, Udactiy Mentor
Kulsoom Abdullah, Data Science Consultant
Ecem Yılmazhaliloğlu, President of Technoladies
Rebeca Moreno Jiménez, UNHCR Innovation Officer
Sofia Kyriazi, UNHCR AI Engineer
More about Omdena
Omdena is the collaborative platform to build innovative, ethical, and efficient AI and Data Science solutions to real-world problems.
Using GAN networks for satellite image quality augmentation to identify trees next to power stations more accurately. The solution from this project helps to prevent power outages and fires sparked by falling trees and storms.
Using Generative Adversarial Network (GAN) for Data Augmentation
The GAN stands for Generative Adversarial Network, which is essentially applying game theory and put a couple of artificial neural networks to compete with each other while they are trained at the same time. One network tries to generate the image and the other tries to detect if it is real or fake. Actually, it is something very simple, but pretty effective too. This is clearer with an image:
But again, how can we use this to accomplish our goal? It turns out that there is a kind of GAN Network named pix2pix for Data Augmentation. This kind of GAN can be used as an input, a pre-defined sketch of the real one. Like take a doodle and from there build a picture like a landscape or anything you want. An example of this is the application that Nvidia did to generate artificial landscapes. The Link for the video is given here.
Ok, so maybe this can work. At that moment the label team has already labeled some images, so if we use these labels to build some doodles, then we can use this to train a GAN to generate the images. It actually works!
So now we just need to find a way to generate random doodles to feed the pix2pix GAN. So here is another GAN to the rescue, a DCGAN in this case. So, in this case, the idea was to generate a random doodle from random noise. Getting something like this:
And finally putting all the pieces together, with the help of some Python and Opencv code, we end up with a script that generates a 100% random image from pure noise with the corresponding labels. At the moment we can generate thousands of synthetic images with their corresponding labels in a JSON file in coco format. For the labels, we use the doodle to get labels by masking the colors and then build the synthetic images from the doodle.
For now, the results look promising, but they are just preliminary results and can be enhanced, for example, the labels that we use, only had labels for trees or not trees, this can be enhanced by another label to make the model more specific and accurate, like for example also label roads, fields, buildings, lakes, rivers and so on, to make the model generate this stuff.
Killing the myth of the lone genius, why command-and-control work systems are failing to build AI solutions that are adopted.
You may not have heard of top-down management or bottom-up management, but you are definitely familiar with the former, as it’s a traditional management style. Essentially, the boss or leader makes all of the decisions, and the employees or team members carry them out.
According to Stanford research, top-down approaches often result in a lack of participation, input from the wrong people, and not enough creative conflict.
Now, before we get into the why let us understand the differences between both approaches.
Top-down all bad? No, but…
The advantage of top-down management is that decisions can be made and implemented very quickly. This is particularly important when time is limited. The other benefit of top-down project planning is that it helps align the project goals with the organization’s strategic goals as upper management is giving the directions.
But here comes the problem!
Artificial Intelligence applied in and by organizations to solve real-world problems is a new field and there is no one size fits all model that can be replicated across companies and sectors.
It is just common sense that organizations do not have a strategy in place and solving real-world problems with AI requires innovative minds, lots of creativity, and cross-functional collaboration.
More than 50 percent of AI solutions fail due to a misalignment of stakeholders, no metrics, and no contextual AI strategy.
And there are more reasons why top-down does not work:
AI teams are by nature multi-disciplinary consisting of data scientists, Machine learning engineers, data engineers, and domain experts. All of which have different tasks to focus on but need to work together to integrate each other’s work and build a functional solution.
AI systems built in the lab are often biased and lack real-world approval. Bias comes from a lack of diversity and top-down means a single person or only a few make decisions. Not a good idea if your system is going to impact thousands or even millions of diverse users. Amazon’s sexist algorithm is an example of how it should not be done.
Running AI projects requires solving multiple problems in various domains such as accessing data, preparing the data, fine-tune parameters, testing different data models, and communicating results in a story-driven and value-oriented way. No single person has all the knowledge to cover all tasks needed.
The alternative? Bottom-up collaboration
The advantage of bottom-up planning is that the team members have a voice in the project planning and decisions are made collaboratively. This improves team communication and team building and also empowers the team members.
Identifying tasks first also leads to a more detailed project plan, with a potentially more accurate schedule, and a better understanding of the problem to be solved.
Benefits with regards to AI projects
Bottom-up development sets the ground for innovation by combining diversity of thought with a safe environment where information exchange happens easily and openly. Especially in AI projects, bottom-up teams require people from different backgrounds and experience levels. This high degree of diversity not only prevents biased solutions but also enables breakthroughs via fast iteration cycles and constant perspective sharing, which connects to the next point.
Less risk of bias
More perspectives, higher gender diversity, and cross-cultural teams result in a more inclusive approach to preparing datasets and building solutions.
More and better data
Another potential benefit of bottom-up driven teams is to they find creative ways to access new data sources and augment existing ones. In AI the team with the best data wins.
Higher user adoption
Many AI solutions and products fail in the real world due to a lack of customer trust and no contextual applicability. A bottom-up approach takes the development process out of the lab and into the real world by embedding diverse perspectives and open dialogue where chances are higher to build inclusive and trustworthy solutions.
How it works in reality: a real-world case study
A company that wants to become AI-powered and build solutions for real-world problems has to overcome several challenges along the way.
In theory, bottom-up collaboration sounds all good but many companies struggle to get access to AI experts and data scientists that can help to translate a problem into a deployable AI solution or prototype.
In addition, many organizations have little or no data, to begin with. And even if data is plentiful, the question remains, how to leverage the raw data to solve problems or gain valuable insights. If an organization made the step to develop an AI system, the next wave of challenges is just around the corner.
In order to help organizations, we started Omdena — an innovation platform where organizations host AI projects that are being solved by global and collaborative teams of up to 50 engineers.
Omdena runs AI Projects with organizations that want to get started with AI, solve a real-world problem, or build deployable solutions within two months.
In our eight-week projects, a selected team of 40 to 50 engineers and domain experts delivers deployable solutions using real-world data. A team consists of various roles, backgrounds, and involves people from around the world, which mitigates not only bias but also results in fast and agile development.
Decisions are being made within different task groups and are discussed within the team to find the best-fit solution.
Our community has compiled many impactful examples of applying AI technology ranging from use cases in the environment to health, finance, safety, justice, and many more.
If you want to learn more about us you can check out all our projects and real-world case studies here.
Omdena is an innovation platform for building AI solutions to real-world problems through the power of bottom-up collaboration.