The Ethics of AI

LIVE

It seems like you can't take a step these days without running into the subject of Artificial Intelligence (AI). Machine learning and AI are here to stay but what exactly are the benefits to this technology and more importantly, what are the risks? What ethical questions to we need to ask ourselves when leveraging AI?

In this panel discussion we will be speaking with Rob High - CTO/IBM Fellow/VP for IBM Watson & Cloud Platform and Amy Gross - CEO/Founder WineSleuth/wine4.me about the applications of AI and the associated risks.

Welcome Amy and Rob! Thanks for joining me today.
So I would like to start with a general question for Rob. Can you give us a bit of a background on your work with AI and the application of the technology? I think many people don’t quite understand that this is not machines taking over the world.
I like to think of AI as *augmented* intelligence. What I mean by that is that AI is at it’s most beneficial when what it is doing is helping us (people!) think better. To help us see things we might otherwise miss. To find relationships between things that are not obvious.
To help us discover other points of view that will change the way we think about things, and result in us making better decisions or solving problems in a better way because of it.
That’s a really easy to understand explanation! (I’m going to steal your “augmented” intelligence phrase)
I like what you said about finding relationship between things that are not obvious, Rob. AI can be a great tool to help us find those things and benefit from them.
I think that’s very much what you’re doing at VineSleuth as well.
Amy and Rob, you have worked together on solutions - Amy, can you tell us what AI is doing for you at VineSleuth?
Then we will get into some of the ways to leverage the tech and how to do that in the most ethical way.
At VineSleuth, we are using AI to build flavor profiles for our users and guide them toward wine and food choices they will most enjoy, helping them whittle down a wine list of thousands to a list that will work best for them, personally.
Finding the needle in the haystack
Is there a lot of work that a consumer has to do in order to leverage the building of the profiles?
Our users can build their flavor profiles quite simply, without much effort. We built a mobile app called Wine4.Me where users can simply tell us what wines they enjoy (or don’t enjoy) and from that we can begin building a profile. .
So the big work for the users is tasting all those wines... and then simply loving or dumping them. Our data and AI do the rest, and allow continuous input.
Ah - got it. And I assume that as they spend more time on wine4.me and tastings, it continues to learn from your choices and can hone in more closely.
Does it also teach you as you go on your wine journey?
Absolutely, Carree. The more wines one loves or dumps, the more data the system has and the more accurate the suggestions become. Also, our system was also built to change as the users’ preferences change in time... always evolving.
For example - I like Aussie reds, but know almost nothing about California reds that I might like. I would love to learn how they are similar, how they differ and why I might chose a domestic over an import?
Users can also see a snapshot of each wines’ flavor profile-- and of their own personal profile characteristics, so they can begin to learn what it is they do like in wines. Sometimes the world of wine uses confusing, flowery vocabulary. We try to make it easier to understand.
Yes - the vocabulary - that has always made me laugh actually - so putting that into a better context is great!
Wine4.Me works with specific wines, so if you look at an Aussie red you love, and pull up a California red, you can start to see similarities. We used unbiased data to train our system which makes those comparisons possible.
So all of this seems fantastic for the consumer and also makes a company smarter about offerings and ultimately helps the bottom line, but on the flip side, do either of you find that people are scared of the technology and about what information machines are gathering on us?
What guide rails should we be looking to put on tech such as this to make sure it is used for good and not evil?
We are finding that people are rather curious and willing to give the application a try. What we try to continually explain, though, is that the personalized wine rankings are not influenced by a winery or advertising or a retailer’s desire to sell.
I think there is a growing awareness amongst end users that we have to take responsibility for the information that we share.
There is often a tradeoff between what we share and the value we get back from that information. But any app that does use data to provide value MUST fulfill its obligations to protect the privacy of that data.
We do not modify our rankings based on advertising That is something that concerns users. They want to trust the AI, but are wary of motivations.
It’s one of the core principles of IBM’s cloud based services ... that data used in the cloud is YOUR data, and will be protected as such.
As leaders in the space, it’s great to hear both of you echoing the same sentiment. Privacy is key as well as consumer trust.
And we are encourage and like to reinforce the respect that companies like Amy’s have for their end-user’s data
Rob, knowing that our data is safe is very important to us. We appreciate how IBM’s systems allow us to use your power and still keep our data safe.
Being curious humans - we often want to hear that one “horror story” so that we can put into context how the tech could be abused and give us direction on how not walk that path.
Do either of you have a story to share here?
Amy, you shared how WineSleuth uses AI - so I want to ask Rob what one of the best applications of the technology that you have worked on or seen.
I don’t have a story to share, but I do know that it is critical that we make true recommendations for each user based on his or her profile and not mix in paid placements or even come close to that. The moment we do that, our integrity is gone and we cannot be trusted.
My favorite, by far, is the work we did with Memorial Sloan Kettering on building a system to assist Doctors on identifying treatment options for their oncology patients. It is the reason that I agreed to take on the CTO role for Watson.
The paid piece is super important to point out - glad that you mentioned it. I know that I am definitely more apt to use a platform that does not see my data and use it for retargetting and advertising purposes.
In that system, Watson is identifying the most relevant treatment options for a patient, personalized to that specific patient, and factoring in the current clinical evidence for what that treatment is relevant.
Def Rob! Having sat in the Watson Tour here in NYC as seeing how AI is changing the landscape of healthcare is astounding
It doesn’t replace the Doctor or their judgement, but it does offer them other perspectives that can change the way they think of the patient’s needs.
Even if it only changes things a small fraction of the time -- that is enough to significantly improve outcomes for many patients. It can lead to faster and better recovery, or even save lives.
As you said at the beginning of this discussion - it makes you smarter. And I would imagine that healthcare professionals find that extra “brainpower” extremely beneficial
And appreciate the fact that AI can do so much heavy lifting from a research and investigation perspective
Yes. Funny story: My mom went to her doctor, and as mom’s do on occasion, was bragging to her doctor about what I do. The doctor reacted, “Those guys are going to take our jobs!” She has to back out of the discussion as the doctor got more agitated about it.
But the point is, this isn’t about taking jobs. This is about helping people get the information they need to make better decisions.
What I think is really great here is that there are really not limitations to the applications of AI
Most people that get close to the technology realize that benefit very quickly. The ones that only listen to remote stories, and then rely on their own imagination are the ones that often build up a fear of it that is not at all accurate.
We can save lives, find you the perfect wine, tell you the best driving route, etc. The possibilities really are endless.
I get overwhelmed sometimes by the number of different ways that people have found to leverage this technology to advance their own lives.
That’s great ... looking forward to continuing this further.
We’ve discussed some of the innovative ways that AI is being applied, but it would be great to give folks some direction on how to use this technology in the most responsible way.
Our personal data is in so many companies hands today - how do we ensure we are mindful of the lines that we cannot cross?
And do organizations really understand what those lines are - I know that sounds silly - but even the most innocuous application of AI could cross those lines unknowingly.
The question of responsible use is really important. It’s also complex -- at least in part because we all tend to have slightly different views of what we consider to be “good use” vs. “bad use”, and then further by the fact that there are actually multiple parties involved.
Amen to that. It’s mind blowing how much of our personal “stuff” is out there.
First, let me just assume there is some middle ground for what we would all agree is good vs. bad. Solutions that help improve our health are generally good (presuming they don’t do so at the expense of someone else’s ill-health).
All of that personal stuff has the potential to make life better in so many ways, provided it is used well. It is critical that we all act accordingly..and that consumers are wise about their data while still trusting.
And solutions that can be used to steal from people are bad.
I would agree, Rob.
Let’s also recognize that there are providers of basic technology, those that create solutions by assembling and purposing that technology, and those that consume (benefit?!) from those solutions.
I believe it us up to those that are using data and AI to be transparent (and honest) as to their motives and data sources.
I know it’s early days here, and common sense should always be applied, but are there any good resources out there to help organizations with this?
All of us have a role and responsibility in creating good use. Technology providers (like IBM Watson) have a responsibility to design our technology to encourage good use.
Solution providers have a responsibility for protecting user data and being forthright about their intentions.
And consumers have a responsibility to demand good solutions, to demand the integrity from the solution providers, and to be diligent about what information they provide to the solution for the benefit they want to get.
Rob, you bring up something interesting there - IBM and other larger and more seasoned orgs have pretty strict data privacy policies - but what about smaller companies or start-ups that might not have that structure in place?
Through out all of that, transparency and integrity are a common theme.
As a smaller business, at VineSleuth we do have strict policies in place, too. It is critical as an early adopter of AI for us to set and follow standards... and explain those to our clients and users, as well.
Amy, do you find that you spend more time than anticipated explaining that?
That’s right. Amy and her team are a great example of the fact that integrity and transparency are not conditioned by the size of the organization, it’s really a policy and value that any and all organizations should embrace.
Yes. Often users assume that our technology might recommend a wine based on sales goals or advertising dollars of wineries or a retailer, instead of based on true personal preference, as we do. The moment we do that, our integrity is lost and we cannot be trusted.
And another AMEN to that! - I hope everyone has that approach
Wow - Amy, you are making me think of a million things here. The advertising aspect. It’s great that you make that distinction.
We have had potential clients ask us about weighting recommendations to favor sales or promotions. Once we open the discussion, they quickly understand the ethics behind it and appreciate our approach even more.
It can sometimes be difficult to understand where a company “ends” and advertising “starts”
Two things we do at IBM to help with this equation are: a) we ensure all of our services are able to both represent their confidence, as well as cite their sources of evidence for any conclusion it provides; and b) we also enforce our commitment to protecting customer’s data.
You don’t want AI to be perceived as Big Brother
I really do appreciate IBM’s approach.. and that you are very clear that each client’s data is its own and is protected.
We do that internally as well - not just for our clients. So its cool to hear it being applied in both directions
Artificial Intelligence is very honest in its results-- that is why it is so critical to get the data fed into it right.
Amy, do you have a WineSleuth + IBM Watson version 2 planned?
meaning taking what you have learned with the technology already and taking it to a new place?
We are working through a few options right now, actually. But I can’t reveal much just yet. :)
I figured!
It’s all an evolution!
We are about to launch an expanded web version of the Wine4.Me application very soon. THAT is quite exciting for us!
Rob - I know that the health care side of this is what most consumers know when they think of IBM and Watson - but what are some of the other applications that you have seen or worked on that are moving the needle
Everything with AI is kind of exciting, isn’t it?!!
For the last couple of years, the most common interest has been around conversational agents -- especially around customer support and service, as well as product selection.
These have had the benefit of increasing customer sat, service efficiency, as well as reducing agent churn.
Smart chat bots and the like, I agree, I see that a lot internally. Are they really ready for prime time?
But there is a re-emerging interest in applying knowledge reasoning -- to find the information you are missing -- types of scenarios.
I think the “connecting the dots you didn’t know you could connect” part to be fascinating
Chatbots are really very easy to create in their most basic form. Where people are really seeing a difference is when the conversation with the client can get even more involved in solving for their problem -- get to the “problem behind the question”.
The technology is mature for this sort of thing, but it does require that the organization think about how to accomplish that in their particular domain.
Who is doing that well Rob? There are a lot of bad bots out there these days.
The kinds of problems that the conversational agent needs to seek and solve in banking are different than they are different retail segments, are different than different healthcare segments, or legal segments, etc.
Two that are very notable are Royal Bank of Scotland and Autodesk.
Both of these are examples of organizations that really tackled the value of a deeper engagement with their clients. And they addressed the importance of this by putting in place teams dedicated to achieving, evolving, and enhancing that effect.
Totally agree on that Rob, and wow - for a bank to be notable in this space is great to hear. Banks always take heat from consumers and to have a conversation agent succeed in that space is pretty massive
Amy, since your adoption of AI, do you find that you have companies reach out to you for guidance?
Since you seemed to have “cracked it”
The economy generally relies on us having trust in our banking institutions, and because of that they tend to be very conservative. But that also makes them vulnerable to not serving their client needs as much as they should.
That’s starting to change, and at least a few banks are taking the lead in making a difference.
Yes, we have found that other businesses interested in AI have reached out to learn about our experiences. It’s been exciting to be a startup in AI and to also work with IBM!
It’s an honor to have this opportunity, and to do so with Amy. Looking forward to digging in more.
I learned a lot and I am sure that your presence at Mobile World Congress will be a smashing session!
Well you have both ignited my brain with how I might apply this in my marketing work in social! THANK YOU FOR THAT!
If you are a member of this panel, please sign in to contribute.