As the development and deployment of AI systems has become more of a regular occurrence, so have stories about biased data sets and model training failures. One recent example came earlier this month when Reuters reported that Amazon had terminated a project to create a recruitment ML algorithm because it was penalizing women due to the biased data sets it had been trained on.
Creating ethical AI is set to be an ongoing issue across many industries. So, ahead of her presentation at this year's DATAx New York Festival, we spoke with Francesca Lazzeri, AI and ML scientist for Microsoft's cloud developer advocacy team, about how companies can create safe, ethical AI and what some of the most disruptive technologies of the New Year will be.
DATAx: How do we as a society work toward democratizing AI?
Francesca Lazzeri: As AI begins to augment human understanding and decision-making in fields like education, healthcare, transportation, agriculture, energy and manufacturing, it will increase the need to solve one of the most crucial societal challenges we face nowadays: The advancement of inclusion in our society.
As a society, we should work together to ensure that AI-based technologies are designed and deployed in a way that not only earns the trust of the users who use it, but of those from whom the data is being collected in order to build those AI solutions. It is vital for the future of our society that we design AI to be both reliable and reflect ethical values which are deeply rooted in important and timeless principles.
We need to also realize that AI solutions learn from training data and so we need to create mechanisms to improve their training data. Training data can be imperfect, skewed, often drawing on incomplete samples that are poorly defined before use. Additionally, because of necessary labeling and feature engineering processes, human biases and cultural assumptions can also be transmitted by classification choices. All these technical challenges can result in the exclusion of sub-populations from what AI is able to see and learn from.
Register today to see Francesca Lazzeri and a host of other AI and machine learning experts share their thought and views at the Chief Data Officer Summit, part of the DATAx New York Festival on December 12–13, 2018
DATAx: What do you feel are some key directives everyone working in AI needs to follow to ensure they are creating safe, ethical AI?
FL: I believe the following key directives will support the creation and utilization of healthy AI solutions:
- Systematic evaluation of the quality and fitness of the data and models used to train and operate AI-based products and services.
- Involvement of domain experts in the design and operation of AI systems used to make substantial decisions about people.
- A robust feedback mechanism so that users can easily report performance issues they encounter.
Finally, I believe that when AI applications are used to suggest actions and decisions that will impact other lives, it is important that affected populations understand how those decisions were made, and that AI developers who design and deploy those solutions become accountable for how they operate.
These standards are critical to addressing the societal impacts of AI and building trust as the technology becomes more and more a part of the products and services that people use at work and at home every day.
DATAx: If you could recommend one way any firm could become more data-driven today, what would it be?
FL: I think that is important to have the right cloud technology platform in place. A modern cloud AI environment will make it easier to collect data, analyze, experiment and quickly put things into production. This sort of capability is becoming a must-have for any organizations, either technical or non-technical.
Moreover, collecting the right data is also crucial. Although companies are collecting petabytes of data, the key question is: Do you have the relevant data for the problems you are trying to solve? For example, many companies who are attempting predictive maintenance have piles of data sourced from all sorts of sensors, but not enough data about their failure history. The lack of this data makes it is very difficult to do predictive maintenance because models need to be trained on failure history data in order to be able to predict future failure incidents.
DATAx: In your opinion, what will be the most disruptive innovations of 2019?
FL: 3D metal printing and blockchain privacy.
3D printing is a great example of disruptive innovation and I think it will represent the next big wave in this industry. 3D metal printing will be useful to many manufacturing businesses.
Thanks to GDPR, blockchain will play a crucial role in today's economy because it can be used for shopping security, whether online or in person.
Francesca Lazzeri (@frlazzeri) will be speaking on Day Two of Innovation Enterprise's Chief Data Officer Summit, part of DATAx New York on December 12–13, 2018. To attend and hear more great insights from AI and ML professionals across some of the biggest organizations, register here.