Is Deep Learning Going To Be Illegal Post GDPR?

GDPR came in effect across the EU on May 25. But what is GDPR?
GDPR - General Data Protection Regulation. GDPR is a new data protection law passed in the European Union (EU). GDPR regulates the control of data outside and for the EU citizens. This is pretty strict for all organizations having customer or prospect data. People can rejoice as they have greater control over who has their data. Businesses have to be transparent about what data they have. While this is welcomed by majority of people, some do not welcome this change. This is because any violation by tech or other companies will attract fines for them. They can be charged with a fine of 4 percent of their annual global turnover.
In this article, we are going to discuss what impact GDPR has on Deep Learning. Deep Learning faces a black box problem - meaning it is not easily interpretable. A neural network is a black box. This means we cannot explain what is happening inside the layers of such a framework. As there are so many hidden layers, it is not easy to explain what goes on. But the neural networks rank high when it comes to accuracy. Researchers are trying to figure out how to explain a neural network's decision-making.

Solving the black box problem in AI
Image source: https://analyticsindiamag.com/deep-learning-going-illegal-europe
It is important to know how the deep learning concepts work on a neural network to explain anything. There is a statute in the GDPR addressing the Artificial Intelligence/Automation issues.

The Right to Explanation of Automated Decision:
This rule gives the right to people to get explanations for automated decisions. If they are not satisfied or do not want to be part of such an algorithm, they can drop out. It can be an email marketing, loan approval or any kind of data based automated decision. This is a limitation for tech companies that depend on data for their decisions. So, in a way it limits the use of Artificial Intelligence development and marketing.
Neural networks learn by universal approximation and not model tuning. The more data they have, the greater the accuracy of predictions and decision-making. Neural networks are not explainable because of the black box issue. This causes a technical gap when it comes to GDPR statutes. Tech companies cannot explain how they use deep learning to come to a decision. So, unless this problem resolves, it is a loophole. This can definitely make it illegal for companies to practice AI and deep learning.

The way we make decisions based on LIME explanations
Image source: https://analyticsindiamag.com/deep-learning-going-illegal-europe

Conclusion:

The problem of AI and deep learning's black box issue is not resolved yet. Professors, data scientists and professionals are trying to address the black box problem. There has been a couple of developments - LIME and DARPA's Explainable AI. LIME (Local Interpretable Model-Agnostic Explanations) is a method to produce explainable models. DARPA aka Explainable AI (XAI) creates a suite of explainable ML algorithms. So, in the end the only issue that remains is being able to explain and consent. Deep learning can only be illegal if these regulations are not followed. GDPR & Deep Learning will have a tough relation until these issues are resolved. To learn more about Deep Learning, look up our Deep Learning Indian centers.

Resources:

https://analyticsindiamag.com/deep-learning-going-illegal-europe/
https://en.wikipedia.org/wiki/General_Data_Protection_Regulation
 

Are You A Deep Learning Aspirant?



 

Leave a Reply

Your email address will not be published. Required fields are marked *