What is Whitebox AI and is it the answer to better transparency and technology for the future?
Robert Ruff, President of Sovren, continues the conversation with Chad & Cheese in this episode about GDPR, Free AI, Blackbox and Whitebox AI, and more.
Enjoy this Voices Series podcast from The Chad & Cheese - HR's Most Dangerous Podcast.
PODCAST TRANSCRIPTION sponsored by:
Voices Intro:
Voices. We hear them every day. Some voices, like mine, are smooth and comforting. While on the other hand, The Chad and Cheese Podcast is like listening to a Nickelback album. You'd rather stab yourself in the ears with an ice pick. Anyway, you're now listening to Voices, a podcast series from Chad and Cheese that features the most important and influential voices within the recruitment industry. Try not to fuck it up boys.
HR's Most Dangerous:
Hide your kids, lock the doors. You're listening to HR's most dangerous podcast. Chad Sowash and Joel Cheesman are here to punch the recruiting industry right where it hurts. Complete with breaking news, brash opinion and loads of snark, buckle up boys and girls. It's time for The Chad and Cheese Podcast.
Chad:
Welcome back. We're picking the conversation back up with Robert Ruff, president of Sovren, industry veteran, and all around smart dude.
Chad:
So over the years we've had these sharing consortiums to an extent where the Job Foxes of the world and there have been some other AllianceQ, where companies that actually... They had pooled the resumes of individuals who didn't make it into their organization and they would be able to feed off of those.
Chad:
Why has that never worked? I know that there's competition that money on those individuals, but still, wouldn't you rather see one of those individuals get a job whether it's with you or somebody else as opposed to having a shitty experience and just going into a black hole?
Robert:
I think to a certain extent the laws have changed those. So if you look at privacy laws, especially outside of the United States, I think things where you take someone's data that was submitted to a particular place for a particular purpose, and gets Senate just for that and you repurpose it into a larger thing that's probably not legal anymore.
Joel:
Yeah.
Robert:
And I think it's getting less legal here. So I think maybe the day of that idea ever being able to get traction, it never did in the past, I think it's, I think it's probably gone forever.
Chad:
What are some of your, I guess bigger thoughts in addition to that in regards to GDPR privacy and where you see the world going and how it impacts our industry?
Robert:
So I have a couple of different thoughts about it. I don't think the GDPR is nearly as honorous as people think. I think it makes a lot of sense. And I wish we had protections to like that over here.
Joel:
Yes.
Robert:
That said, what actually dismays me is that the only job function that is called out in the GDPR,
and it's called out twice as being explicitly subject to it, is applying for jobs.
Robert:
And what I can't get my head around about that is that a resume is actually a person's advertisement, publicly, for a job. So yes, you may send it just to one person at a time, but by no means are you trying to let people have this idea about your data that like, "Don't ever share this with somebody who might use it in a way that would benefit me by offering me a fabulous job and benefits."
Joel:
It's not bank account info.
Robert:
It's not bank account info. It's not your health. It's not, do you have a contagious disease?
Robert:
Yeah. But I think if we had a more transparent just like you're going back to the kind of like the pooling of resumes and whatnot, if we had a more transparent idea and allowed candidates to opt into something like that. So yeah, I would definitely love to be able to... AT&Ts, Sprints or maybe all these other big companies. I would love to have my resume provided to those companies as well.
Robert:
I think a lot of it has to do with transparency. We've been shit for transparency in this industry and if we just focused on being able to treat that actual resume like a human, and maybe not like shit like a human, but treat like a customer because they might spend money with us, then it might actually make sense. But I think the transparency piece makes the most sense because that's really what GDPR is. It's about what you own and being transparent about what's going on.
Chad:
Yeah, I think it's about being honest about what we're going to do with your data.
Robert:
Yeah.
Chad:
Now you don't have to go read some privacy policy and get an attorney. So Apple's newest privacy policy I think was 70 pages that comes out on your iPhone.
Joel:
Like sure.
Robert:
Okay. But I agree with you that... So one of the things about GDPR that's not well known is that if you are subject to an automated algorithm, well all algorithms are actually automated. But if you're subject to an algorithm for a selection, maybe it's for credit extension, maybe it's for hiring. They have to let you know how that algorithm worked. Well, most AI is Black-Box AI.
Chad:
Yes.
Robert:
And no one knows what it's doing and you're like, well it's just 20 to a thousand layers of statistics of statistics and we don't really know how it got to the answer or why and we can't really control it.
Robert:
Well that on its face doesn't really fly for the GDPR unless you have an opt out where a human gets involved. With our software, a little advertisement, we don't have that problem because we built White-Box AI. So it does what humans do the way humans do it. Humans can be told what it's doing in human terms and can change it and control it.
Robert:
But I think that the idea that algorithms are something that need to be disclosed is a good idea. Because if they're not disclosed and they're not verifiable there's a tremendous amount of hidden and unknown bias that is lurking in those, right? Because we're doing is we're looking at data that's always been there about who we hired in the past and we're saying, "Oh look like the people we've always hired so we will want to hire you."
Robert:
It's basically locking people in into a vision of the future that is exactly like their past, which is the opposite of competing in the global society we're in today where things change so much.
Joel:
Yeah.
Chad:
Well exactly. And those data sets, right? That big data. Remember we always used to talk about big data?
Robert:
Right.
Chad:
Now we can crunch that big data with the algorithms, but all that data is how all the bias methodologies we used before. So it's like how do we become unbiased? Well we don't use all that bias shit.
Robert:
So we've got an engine that we scrub out all the data about the person before we use it in the engine. So you know, I can't look at your name and guess it's about who you are, either as a sex or a nationality or anything because I don't see your name. We scrubbed your name out, your address, your email addresses, your date of birth, how old you are, your gender, everything that we know about you we scrub out before we use it on our engine.
Robert:
In addition, we get out of it or allow you to take out of it, what I think is the hidden bias that needs to just stop and it is rampant in the U.S. And before we have the next recession, this type of bias needs to be made illegal. And it is the bias on educational attainment.
Joel:
Which really sucks because my email address, bigstud@aol used to get me a lot of places and now it's totally scraped out from everything. So whatever.
Chad:
Well there, there's an app for that, but...
Joel:
There many actually. What a Great segue to AI and I'm curious, you were at HR Tech this year-
Robert:
Yeah.
Joel:
... and AI is pretty much on every trade show booth in the hall. As someone who's really dedicated to AI, how do you feel or what do you think when you look at all these companies, touting AI and what that really means? If anything?
Robert:
Well I think that finally people are starting to wake up and realize that if the home builder is telling you, you should buy his house because he uses 18 volt cordless tools to build it, maybe that's not the compelling argument you're looking for.
SFX:
That is one big pile of shit.
Chad:
Okay, so real quick. So I mean what you're saying is AI's the standard, everybody's using it. So putting it on your booth really doesn't mean anything. The big question that we keep asking in this industry is AI real in recruiting?
Joel:
And how do you define it?
Robert:
So AI is... It's not an application, it's not a solution, it's not even a tool. It's actually a toolkit. So you download the free AI software. I don't know if this is widely known, but the greatest AI in the world is free. Google will give you the exact stuff they use in their business for free.
Joel:
Open source.
Robert:
Yes, it's open source. So what do we know about that? We know that the AI toolkit itself has no value. If it had value, they'd sell it for something. What is the value in it? The value in it is what it can do. And what it can do is going to be limited by how much data you have, how well you have curated it and how well you have finessed the toolkit around that data in multiple, multiple iterations.
Robert:
So that gets you to a working algorithm which then has to be validated in the real world and more feedback mechanisms. But that only gets you Black-Box AI. To do White-Box AI, you have to take it many steps further and you have to get the good data and the bad data and put them into separate buckets and start using those explicitly and algorithms on top of algorithms rather than having one magic thing that is uncontrollable.
Robert:
You have to bring it up to a level of, we started with this level of competency and we built on top of that and that's what people don't understand about AI. They think that it's some type of magic that you just point it to data and it figures out everything about the data. It's like, no, that's not how it works.
Chad:
So that's one of the biggest issues I think we have though with AI is that if you take a look at profiles, a lot of times it's garbage, unstructured garbage, which is a very hard job for you, which you guys obviously do well. But then you'd take a look at the job description and that's just as much garbage as the person's profile or their resume. So how do you actually take two sources that are incomplete sets of data, let's say, okay, in bad data and actually make that match happen?
Robert:
So I don't want to answer that question, but I want to answer a question that's related to it. So I'm acting like a politician now. You asked me a question I'm going to give you-
Chad:
You don't want to give her away the secret sauce. That's what I'm hearing.
Robert:
So there's another question about what's the proper use of AI in chatbots and the industry and stuff like that. And I get tired of listening to people saying that this is going to change everything. Well, let's admit that most of the hiring in any society is of the type of hiring that actually doesn't even need a resume. It's application-based.
Joel:
Right.
Robert:
And a lot of places and a lot of jobs that you might find a thousand feet from where you're sitting right now. The only thing really required is, can you show up on time? Do you have reliable transportation? Can you pass a drug test? So those kinds of things. You don't need matching software, you don't need AI for that. You know what you need? You need an app. An app that reaches people in that vicinity and says, "Hey, you got a job for you. You want to show up tomorrow? Can you do these things? Great."
Joel:
Is that the Gig Economy you're talking about?
Robert:
So not necessarily the Gig Economy, but if I was going to hire somebody for a slot where if somebody quit today, McDonald's, I know that I can't hire somebody 40 miles away, they're not going to reliably get there all the time, right? I've got a small radius and I should be able to reach people in that radius and say, "Look, this is a 11.50 an hour, this is what you do, it starts tomorrow. You need to be able to do this, this, and this. So pass a drug test, don't show up if you're going to fail the drug test, whatever those two or three little drivers are.
Robert:
That completely can be done with an app because as you're describing, there is no data that needs to be matched there. What we need to do is know are you available and do you have these attributes and these attributes are never going to be on a resume.
Joel:
All right. So almost a Tinder for jobs.
Robert:
Yeah. Exactly
audience:
Ooh.
Robert:
Oh, I felt-
Chad:
I love it.
Robert:
... from the comments earlier. We were going to go that direction.
Chad:
Look for more episodes of Voices. This Chad and Cheese Podcast series devoted the stories and opinions of industry leaders. Subscribe on iTunes, Google Podcast or wherever you get your podcasts so you don't miss a single show. For more visit chadcheese.com.