Tay scandal taught us to take accountability, says Microsoft CEO

Satya Nadella says Redmond has learned from its disastrous racist chatbot

Microsoft CEO Satya Nadella stressed today that it is the job of AI companies to ensure that artificial intelligence in kept under control.

Speaking at the London launch of his new book, Hit Refresh, IT Pro heard Nadella address the question posed by many AI sceptics regarding what happens if tech companies create an AI that gets out of hand.

Advertisement - Article continues below

"It's up to us. In other words, how do we approach this with a set of design principles that allow us to control what AI we create? Just like good user experience, I would claim there is good AI," he said. "As designers of AI, it's our responsibility."

Microsoft has invested heavily in the space and Nadella considers it to be one of the three main pillars of the company's future, along with quantum computing and mixed reality. It has made progress with AI most notably through its digital assistant Cortana, but in other areas as well, including machine vision and advanced analytics.

Some of its AI experiments, however, have been less successful. One particularly embarrassing failure was Tay, a Twitter-based chatbot powered by machine learning. Designed to emulate a teenage girl, Tay's conversation was supposed to become more natural through learning from social interactions with real users.

Advertisement - Article continues below

This quickly went off the rails, as trolls exploited the system in order to teach Tay to parrot racial slurs, conspiracy theories and other objectionable comments.

Advertisement - Article continues below

Nadella acknowledged that the experiment proved problematic, but said that the company has learnt from the incident.

"One of the things that has really influenced our design principles is that episode; we have to take accountability. First and foremost, we need to be able to in fact foresee these attacks," he said.

"But the idea that we need to keep the broader goal of having this AI behave properly is our accountability. So how can we test it, how can we make sure that it does not lose control is a lot of places where we're working now."

Featured Resources

Top 5 challenges of migrating applications to the cloud

Explore how VMware Cloud on AWS helps to address common cloud migration challenges

Download now

3 reasons why now is the time to rethink your network

Changing requirements call for new solutions

Download now

All-flash buyer’s guide

Tips for evaluating Solid-State Arrays

Download now

Enabling enterprise machine and deep learning with intelligent storage

The power of AI can only be realised through efficient and performant delivery of data

Download now


Business operations

Amazon and Microsoft join NHS project battling pandemic

27 Mar 2020
Microsoft Windows

Microsoft puts Windows development on lockdown

25 Mar 2020
backup software

Windows File History and Backup review: Useful but limited

25 Mar 2020
web browser

Microsoft halts Edge browser updates in light of coronavirus pandemic

24 Mar 2020

Most Popular

Server & storage

HPE warns of 'critical' bug that destroys SSDs after 40,000 hours

26 Mar 2020
video conferencing

Zoom beams iOS user data to Facebook for targeted ads

27 Mar 2020

These are the companies offering free software during the coronavirus crisis

25 Mar 2020
Mobile Phones

Apple lifts iPhone purchase restrictions

23 Mar 2020