0:04 responsible AI refers to developing and
0:06 using AI in ways that are ethical
0:09 trustworthy and socially responsible
0:12 lots of developers businesses and
0:13 governments care about this and have
0:16 been having conversations and also been
0:18 working hard to make sure that AI is
0:20 built and used
0:22 responsibly because of all this
0:24 attention and effort on responsible AI
0:26 we've actually made quite a lot of
0:28 progress on this in the last few years
0:30 with for example many governments and
0:32 companies publishing Frameworks for
0:35 responsible AI but a lot of work Still
0:38 Remains let's take a look at what
0:40 responsible AI means while we're still
0:42 figuring out a lot of details of how to
0:45 build responsible AI some common themes
0:48 have emerged these are some of I think
0:50 the key dimensions of implementing
0:53 responsible AI first is fairness to
0:56 ensure the AI doesn't perpetuate or
0:59 amplify biases transparency to make sure
1:01 AI systems and the decisions are
1:03 understandable to the stakeholders to
1:04 the people
1:07 impacted privacy protecting user data
1:08 and ensuring
1:10 confidentiality security safeguarding AI
1:13 systems for malicious attacks and LLY
1:16 ethical use ensuring the AI is used for
1:18 beneficial purposes one of the
1:22 challenges of these Dimensions or these
1:24 principles is that the implementation is
1:27 not always straightforward for example
1:30 for I think at least a couple thousand
1:32 years now Humanity has been debating
1:35 what is ethical and what is not ethical
1:37 there is unfortunately no clear
1:39 mathematical definition of ethical
1:42 versus unethical Behavior although of
1:44 course there are many clear-cut cases as
1:47 well but that's why for individuals
1:49 organizations even countries to adopt
1:51 responsible AI there are certain
1:54 emerging best practices to Hope have the
1:57 discussion and debate that will lead to
1:59 better and more responsible decisions
2:01 even when sometimes the right thing to
2:03 do could be ambiguous I want to share a
2:06 few tips first I think is important to
2:08 build a culture that encourages
2:10 discussion and debate on ethical issues
2:12 so if someone on your team has a concern
2:14 about the use of responsible AI it'd be
2:17 great if they have the freedom to raise
2:19 that issue to enable the team to maybe
2:22 make a better decision second tip is to
2:24 brainstorm either by yourself or with
2:26 your team or with an even broader group
2:28 of stakeholders how things could go
2:32 wrong a found on many projects that this
2:35 brainstorming can help identify
2:37 potential problems and allow the team to
2:40 mitigate them in advance a checklist for
2:42 brainstorming could be the five
2:44 Dimensions I described on the previous
2:47 slide could the AI system have issues
2:49 with fairness transparency privacy
2:52 security or ethical use for example on
2:54 some of the projects I've worked on my
2:56 team has brainstormed in advance if the
2:59 LM we deployed could have fairness
3:01 issues such as if it might exhibit some
3:04 of the biases that you saw earlier in
3:06 discourse finally I encourage you to
3:09 work over diverse team and include
3:11 perspectives from all stakeholders
3:14 impacted by the AI system for many
3:16 projects seeking a diverse set of
3:19 opinions as well as speaking with people
3:20 that could be quite different than
3:24 myself has allowed my team to understand
3:27 better the impact of an AI system and
3:30 led us to make better decisions for
3:33 example Building Systems in healthcare I
3:35 found that talking to patients and
3:37 doctors gave perspectives different in
3:40 mind and really changed the direction we
3:42 took our projects in and working on
3:45 retail applications talking to some of
3:47 the customers as well as the sellers
3:49 gave my team new ideas that we wouldn't
3:52 have had otherwise and I think this
3:55 pattern is true for many projects if you
3:57 work in a specific industry such as
4:00 Healthcare or Finance or me media OR
4:02 tech there may be emerging best
4:05 practices for responsible AI specific to
4:07 your industry they could be useful to
4:10 consult as well as you embark on your
4:12 project I think we all want to use AI to
4:15 make people better off there have been a
4:18 few times that I've kill projects that I
4:20 assess to be financially sound on
4:23 ethical grounds as you decide what to
4:25 work on and what not to work on I hope
4:28 you keep on considering responsible Ai
4:30 and only work on projects that you think
4:33 are ethical and that make people better
4:36 off and now we're approaching the end of
4:39 this course let's go to the next video
4:41 to see a summary of what we've covered
