RD-11.15-16.2022-Themis-AI

Startup Exchange Video | Duration: 4:14
November 15, 2022
  • Interactive transcript
    Share

    ELAHEH AHMADI: Hi, everybody. My name is Elaheh Ahmadi and I'm the CEO and co-founder of Themis AI. Themis AI was founded and born out of MIT Computer Science and Artificial Intelligence Lab at MIT, and by Professor Daniela Rus, Dr. Alexander Amini, and me. And our mission is to empower the world to create, deploy, and advance trustworthy AI.

    So what are we doing exactly? What is the problem that we're tackling? With the rise of AI in industry, we have seen a lot of catastrophic consequences of using AI without having any safety measures. And we are seeing the rise of AI in settings such as autonomous vehicle, facial detection, and securities, financial models, and health care. And on top of that, a lot of executives, based on our research, around 75% of executives, even though they recognize that trustworthy AI is an issue, they still haven't taken any actions to solve that within their organization.

    On top of that, there is huge loss in untapped revenue, more than $13 trillion, because of using biased and untrustworthy AI models. And what makes it worse is that around 62% of organizations are facing talent shortage to hire people to actually tackle these problems. With that, we innovated a technology at MIT for more than five years, and at Themis, that is robust and reliable, which means we can tell you when you should trust and shouldn't trust your model.

    It's fair and ethical, which means we can ensure that your model has consistent performance on all subgroups within your data. And it's explainable, which means we can tell you what is happening exactly for each decision, and how decisions are being made. In one of our case studies, we helped our customers to deploy a complex risk-aware autonomous system. One of their challenges is that their complex system consists of more than thousands of different sub-machine learning algorithms doing subtasks.

    But one of their blockers is how can we make these models be risk-aware, meaning how can we know when each model is actually trustworthy and when it isn't. With our unified solution, we were able to help them convert all of their models into risk-aware, with only one platform. And that helped them to cut down 12 months of R&D into only one week of development. And on top of that, they were able to deploy safe autonomous systems.

    The product that we're bringing into the industry is an end-to-end risk-aware AI workflow. What we believe is that the only way that you can deploy a safe and trustworthy AI is by identifying and mitigating risk in every single step of your machine learning cycle. One of our products, which we have a free and open source version available, is the risk-aware AI diagnostic tools.

    Our product comes on top of your existing workflow and converts it into a risk-aware one. So on top of the prediction that your model was designed to provide to you, now we can tell you risk metrics about that prediction, of how trustworthy that is, how biased that prediction is, and whether there was input noise into your model or not. And what's more, we make sure to make usage of our product as easy as possible.

    So in order to use our product Capsa, all you need to do is add two lines of code into your existing workflow and you can get all of the benefits of our product. With Themis, we can help you to deploy risk-aware AI in different applications within your company and on different data types such as visual, tabular, and language models. And what we're looking to do is to work with you through a partnership or a proof-of-concept project.

    Our table's also out there if you want to see a demo of our product or learn more about what we're doing. Thank you so much for your time and attention.

  • Interactive transcript
    Share

    ELAHEH AHMADI: Hi, everybody. My name is Elaheh Ahmadi and I'm the CEO and co-founder of Themis AI. Themis AI was founded and born out of MIT Computer Science and Artificial Intelligence Lab at MIT, and by Professor Daniela Rus, Dr. Alexander Amini, and me. And our mission is to empower the world to create, deploy, and advance trustworthy AI.

    So what are we doing exactly? What is the problem that we're tackling? With the rise of AI in industry, we have seen a lot of catastrophic consequences of using AI without having any safety measures. And we are seeing the rise of AI in settings such as autonomous vehicle, facial detection, and securities, financial models, and health care. And on top of that, a lot of executives, based on our research, around 75% of executives, even though they recognize that trustworthy AI is an issue, they still haven't taken any actions to solve that within their organization.

    On top of that, there is huge loss in untapped revenue, more than $13 trillion, because of using biased and untrustworthy AI models. And what makes it worse is that around 62% of organizations are facing talent shortage to hire people to actually tackle these problems. With that, we innovated a technology at MIT for more than five years, and at Themis, that is robust and reliable, which means we can tell you when you should trust and shouldn't trust your model.

    It's fair and ethical, which means we can ensure that your model has consistent performance on all subgroups within your data. And it's explainable, which means we can tell you what is happening exactly for each decision, and how decisions are being made. In one of our case studies, we helped our customers to deploy a complex risk-aware autonomous system. One of their challenges is that their complex system consists of more than thousands of different sub-machine learning algorithms doing subtasks.

    But one of their blockers is how can we make these models be risk-aware, meaning how can we know when each model is actually trustworthy and when it isn't. With our unified solution, we were able to help them convert all of their models into risk-aware, with only one platform. And that helped them to cut down 12 months of R&D into only one week of development. And on top of that, they were able to deploy safe autonomous systems.

    The product that we're bringing into the industry is an end-to-end risk-aware AI workflow. What we believe is that the only way that you can deploy a safe and trustworthy AI is by identifying and mitigating risk in every single step of your machine learning cycle. One of our products, which we have a free and open source version available, is the risk-aware AI diagnostic tools.

    Our product comes on top of your existing workflow and converts it into a risk-aware one. So on top of the prediction that your model was designed to provide to you, now we can tell you risk metrics about that prediction, of how trustworthy that is, how biased that prediction is, and whether there was input noise into your model or not. And what's more, we make sure to make usage of our product as easy as possible.

    So in order to use our product Capsa, all you need to do is add two lines of code into your existing workflow and you can get all of the benefits of our product. With Themis, we can help you to deploy risk-aware AI in different applications within your company and on different data types such as visual, tabular, and language models. And what we're looking to do is to work with you through a partnership or a proof-of-concept project.

    Our table's also out there if you want to see a demo of our product or learn more about what we're doing. Thank you so much for your time and attention.

    Download Transcript