Bob Friday Talks - AI Alchemy: Building Trust in the Digital Wizard

0:00 hi I'm emine and welcome to another

0:02 episode of poob Friday talks AI has been

0:05 associated with doomsday scenarios and

0:08 stories and pop culture however there

0:10 are more basic concerns about AI than

0:13 Terminators ethical concerns governance

0:16 trust become important so today I'm

0:19 joined by Bob Friday chief AI officer

0:23 here at Juniper to talk about the ethics

0:25 behind AI thanks for being here Bob y

0:28 thanks for having me so so this is a

0:30 very interesting topic so let's dive

0:33 into our first question why is AI

0:36 different than other intelligent

0:38 solutions that we have built in the past

0:40 yeah you know what I tell people is AI

0:42 is really the next step in the evolution

0:44 of automation you know and what's really

0:46 different about this generation of

0:49 automation is in the past we were

0:50 Building Solutions that were very

0:52 deterministic right you know the robot

0:54 that worked on your car did the same

0:56 thing day in and day out you know the

0:58 script that the network it guy wrote was

1:01 basically doing the same thing day and

1:03 out we're now building AI solutions that

1:05 are really doing Solutions on par with

1:09 humans right we're basically Building

1:10 Solutions that are starting to do

1:12 cognitive reasoning right and that is

1:14 why people are starting to look at the

1:16 issues around when you build a solution

1:19 that is basically changing Behavior

1:21 right just like a real person you know

1:24 when you hire someone that person learns

1:26 and changes his behavior and skills

1:28 we're building AI Solutions that look

1:30 and feel like an actual person or what a

1:32 human's doing and same ethical concerns

1:36 that we see with humans now are starting

1:38 to apply to AI when you actually see

1:39 their behavior change you don't know

1:41 what exactly is going to happen on the

1:42 next day so there are a bunch of uh

1:45 ethical questions around AI but to your

1:48 you know to your opinion in your opinion

1:50 what's uh some top concerns about Ethics

1:53 in AI yeah I you know I think what

1:55 people are looking at right now you look

1:57 like the self-driving car no we're

2:00 approaching a day some point where we're

2:02 going to find out that these AI

2:03 algorithms and self-driving cars are

2:05 actually safer than humans right and so

2:09 there may be a day when all of a sudden

2:10 you know driving becomes illegal you

2:13 know so there'll be eth and it's not so

2:15 much an ethical concern is it's going to

2:17 get more into liability you know when

2:19 you have a car wreck inherently you're

2:21 liable for that wreck what it's going to

2:23 happen in the days when you're south-

2:25 driving car wrecks who's liable is it

2:27 going to be you or the manufacturer

2:29 after the car so those are the issues

2:31 we're starting to face with these AI

2:33 solutions that are starting to do more

2:35 and more things that are really on par

2:37 with humans that we have humans doing

2:39 right

2:39 now Bob that's an excellent example on

2:42 the self-driving cars now as more Fields

2:45 let's say healthc care HR hiring OR tech

2:50 jobs more F to adopt AI how do we train

2:53 AI models to have the same level of

2:55 Ethics as humans and how do we ensure

2:58 that it's unbiased that it's fair and

3:01 who should be held accountable yeah you

3:03 know a great example of this would be

3:05 you know if you look at what we're doing

3:07 with chat gbt you know if I were to give

3:09 chat TPT the mission to optimize a

3:13 website for profit and actually give it

3:15 access to a bank account which we are

3:17 close to doing right now the same

3:19 business ethics that we apply to a

3:20 business would want to be applied to

3:22 chat TBT right if chat gbt is on the

3:26 mission to basically optimize this

3:28 website for profit and it decides that

3:30 it's going to put its competition out of

3:31 business you want to make sure that AI

3:34 has the same ethics that we apply to our

3:36 business standards apply to that AI

3:38 assistant right and if you look in our

3:40 HR departments right the same ethics

3:42 against discrimination right if we're

3:45 using an AI tool for hiring we want to

3:48 make sure the same ethics that apply to

3:49 that human HR department also apply to

3:52 that AI we don't want any discrimination

3:54 to be built into our AI Solutions in the

3:56 future same as we would apply to any

3:59 human situ ation MH now maybe let's deep

4:02 dive into AI governance can you expand

4:04 more on that part you know AI governance

4:07 is almost a topic similar to ESG right

4:10 you know where our customers are

4:11 starting to ask us what are we doing on

4:14 par with carbon right so when we get our

4:17 rfps a lot of those things we have to

4:19 explain how we're helping the global

4:21 world around lowering carbon emissions

4:24 we're starting to see AI governance show

4:25 up in that same ESG type of category

4:28 they want to understand what is our

4:29 policy and our governance and our

4:31 principles around Ai and adopting AI

4:34 inside of juniper Juniper currently has

4:37 a set of AI principles around the

4:38 network so the first step is really

4:40 making sure that you have a set of AI

4:42 principles when you're developing now

4:45 let's talk about trust we cannot really

4:47 adopt or work with a tool if we don't

4:50 trust it right so what is explainable AI

4:54 uh or explainability and how is it

4:56 related to trusting AI yeah another

4:59 great question and usually I probably

5:02 give the example of the solutions we're

5:04 really building or building doing things

5:06 on par with humans again and really just

5:09 like any other intern or new hire you

5:12 want to be able to understand what that

5:14 person can do you have to get to know

5:16 its skills and what it's capable of and

5:18 even more important like a human these

5:21 AI Solutions are going to start adapting

5:23 and changing their behavior as they

5:25 actually get more data and learn so that

5:28 is why I call the level of trust the

5:30 same trust we have to find in a new hire

5:33 is going to be similar to when I bring

5:35 on this AI assistance I'm going to want

5:38 to understand its skills what it's

5:40 capable of and I'm going to want to

5:42 learn that as it trust as it learns new

5:45 behaviors that actually it does not

5:47 change going into bad behaviors now

5:50 going back to training Technologies

5:51 right what measures do you take to

5:54 deliver Expendable Ai and transparency

5:57 yeah a good example of this is what

5:58 we're doing with zoom and teams right

6:01 now where we're taking Network features

6:03 we're taking Zoom data we're training

6:06 models to actually predict your Zoom

6:09 performance your latency right now what

6:12 we use now to explain what features are

6:14 relevant to those predictions is

6:16 something called shapl This is a data

6:18 science technique is once you've trained

6:20 a model to actually predict something I

6:23 want to understand what features are

6:25 relevant in making those predictions so

6:28 that is exact transparency

6:30 we're starting to help our customers

6:32 understand how the model is making

6:33 predictions and what network features

6:35 are really relevant in those predictions

6:37 thanks V for walking us through the

6:39 ethics of AI this is a very heated topic

6:42 and thank you for listening until next

6:50 time