“Trust, but verify” must be replaced with “Verify, before you trust” in this brave new world.
In the final part of this series on AI, David goes over the problems presented by AI and what to look out for when you start using it in your business.
In this episode, David goes over five essential points when introducing and using AI in your business:
Never miss an episode when you follow us on social media:
Instagram: @tbshpodcast
Twitter: @tbshpodcast
Facebook: The Buck Stops Here Podcast / @tbshpodcast
All music is provided by Soundstripe.
One of the most famous episodes of The Twilight Zone is entitled “To Serve Man.” And in the episode, these aliens come to Earth and offer us all this amazing technology and life-extending benefits. And then the last scene of the episode, someone runs in and says, “Stop everything! I’ve translated it. To serve man is a cookbook for humans!” That, my friends, is the kind of dangers that await you if you don’t start thinking about the pitfalls that await you with artificial intelligence as you incorporate it into your business. Today on The Buck Stops Here, we’re going to go through what we consider the top five pitfalls and traps that await you with AI if you don’t put the right workflow and plans in place to deal with it. All that and more, today on The Buck Stops Here.
So, on this episode of The Buck Stops Here, we’re going to look at the traps and pitfalls that may await you in dealing with artificial intelligence. And that is really important that you understand that there are a whole bunch of them. The second part of this three-part series, I was illustrating a lot of the ways that you could use the artificial intelligence to your benefit.
How it’s an amazing “game changer,” to use the cliché, for a lot of businesses out there. And that’s really exciting. That’s really exciting. But here’s the part, and probably I would think this would be the least popular part of these three-part episodes, is the one where you rain on everyone’s parade. You say, hey, hold on a minute. I think that’s great. But there’s some things we haven’t thought about in here quite yet on how we need to use it. Now, here’s the problem with this is that this is going to move at such an accelerated pace over the next few months even. You cannot wait until the dust settles because then you’ll be left behind. You need to start the race now. The first thing I’m going to say here is that this is truly a place where the buck stops with you in running your organization or your division of your company. You are in charge of this. You must be the one on point leading the expedition through the jungle. You’re going to have to be cutting away the vines and you’re going to have to be pointing out the traps and pitfalls, and you’re going to have to be setting the stage for those who are following you.
You have no choice in this matter, and if you don’t, you’re going to have a lot of heartache, my friend, when you start thinking about how this technology works. And there’s five topics we’re going to hit on briefly in this podcast episode that you really need to think about. Number one is going to be introducing this technology to your employees. We talked a little bit about that in episode two, and we think that that is so important that it needs to be hidden to the traps and pitfalls of the technology. We need to talk about that now. The second piece is going to be the… This is a gold rush! It’s a veritable gold rush. You will be inundated with companies making claims that their software will do X. I’ve already tested out, I think, close to two dozen different pieces of software, and I found, with the exception of two pieces, that the claims being made differ widely from what is actually observed in the software. And because there’s going to be so much money thrown into this from Silicon Valley and everything else, there’s going to be a lot of false gold and a lot of false claims made about the software.
I think the fundamental thing is that a lot of these claims are being made, not because the software is being powered by AI, but people are literally putting on the mantle of AI. And it doesn’t matter what your technology is, even if it was just basic machine learning or something like that. Or, there was a piece of sound software I was trying to use the other day that said that it would take the sound background noise out of a room and it was just a basic sound gate. But it claimed it was done by artificial intelligence, and they had a shiny YouTube video that went along with it. Ultimately, I’m not reviewing that software here, but if it’s powered by AI, I am deeply surprised. The third piece we’re going to talk about… So, there’s going to be the overselling you on capabilities. The third piece we’re going to talk about is the problem with the veracity, or truthfulness, of the information provided by the AI itself, and how you’re going to have to address that. The fourth one is the confidentiality problems. There’s a whole bunch of things on there. These are the last two are coupled together into the legal regime is what I’d call this.
The last one is ownership and copyright. There’s infringement questions based on this. There’s privacy questions based on this. We’ve already seen the first shots be fired by… The lawsuits are coming. There is a veritable tsunami of lawsuits that are coming in regards to artificial intelligence. A, because of protectionist type things, which we always see with new technology. But, as importantly is, there’s a whole bunch of legal questions that haven’t been answered, and no one is seeking to answer those. It’s only wrong if you get caught, or the idea is easier to ask for forgiveness and permission, and I don’t think that is a place I would want to stake the reputation and livelihood of my business upon. Those shifting sands. You want to make sure you have bedrock for some of that stuff, or at least are aware of those challenges. Without any further ado, let’s just dive right in. The first one to deal with is introducing it to your employees. This technology will creep out or scare most of your employees. Some will think it’s very cool, but they’re probably a minority of the people who work for you. What I want to say on that is it’s very, very important you think about when you approach this to your employees and you say, hey, I want you guys to be aware of this new technology.
This is how it’s going to work, etc. You need to approach that, I think, in a gentle fashion, a transparent fashion with your employees. But I also believe that you’re going to have to dictate the way your people are going to use it, just like you would any new policy or procedure. What you found, and this goes in the confidentiality thing, is if you don’t, if you don’t get out there in front and do that, you’re going to have issues. For example, Amazon currently has forbidden their employees to use ChatGPT because they noticed that ChatGPT was generating responses that involved some of Amazon’s confidential information. I guess there was no policy or procedure in place initially for that, and people were feeding confidential documentation into what is a public framework. Remember, you’re training that framework, and then all of a sudden you could start asking it questions, and it will regurgitate your confidential information to anyone who comes asking, and you need to be aware of that. So, when you introduce it to your employees, you’re going to have to say, here’s the new workflow, here’s the plan, here’s how we’re going to use it.
And, at the same time, here’s how we’re not going to utilize these things. And you’re going to need to talk about the different softwares you’re using, and you need to tell your employees that they can or cannot use these certain things. And you’re going to have to put teeth in this, because if someone violates your HIPAA requirements, or your GDPR requirements, or your right to be forgotten, that’s a GDPR requirement. And there’s an open question right now of whether these softwares will be compliant. I will say the vast majority of them are not compliant today. They might be after they get hit with a few million dollars in fines and lawsuits, but they aren’t today. And so, if your people are out using that, like the Wild West, you don’t want to be on the hook for them using it on behalf of your company. So, you need to think about how you’re going to introduce it to your employees, and you need to recognize that you are the one who’s going to have to set the rules and the tone and how that’s going to work. And you cannot, let me repeat, you cannot abdicate responsibility to someone else, in this particular situation.
Point two is the gold rush problem. It’s the overselling you on the capabilities of this. This is moving so quickly right now, and everyone and their brother has thought they’re going to add on AI. Just if you have a piece of software right now, it’s powered by AI. You’re going to start seeing this across the board. And the problem with that is that a lot of it just ain’t so. A lot of it is not powered by artificial intelligence. But because of this, everybody’s excited about it. And you’re going to have people making claims that are not, in fact, true. And you will try the software and you’ll find there’s a wide variety of ways in which the software works and does not work. So that being said, if you were to look online right now at the different artificially intelligent powered by AI writing suites, every one of them claim that they are the best for whatever reason. And I understand that from a business standpoint, there’s a certain amount of confidence you have to have in your own product. But a lot of the things are not true. There’s one very popular writing platform out there, and I’m not going to name it, but it has the ability to detect artificial intelligence.
Okay. I decided to run a whole bunch of prompts through it the other day with its artificial intelligence detecting regime, and it was wrong 80 to 90% of the time, and I took things that were wholly written by AI. It said that they… And there were a couple of those that it grabbed out of 10 prompts, let’s just say. It’s a small sample size. And it said that they were written by artificial intelligence, and I removed a comma. I removed a comma, folks. And it literally said, oh, 100% written by human. I misspelled a word. I used the wrong version of there, their, and they’re, the three different versions in the English language, and it all of a sudden said, oh, 100% written by a human just because I added in a mistake. It wasn’t really high-quality content. I would have told you, if I had read it, I wouldn’t have been able to identify it as written by a machine, but I did recognize that it wasn’t the level of work I expect out of my writing team, and I would have pushed back on that material. So, that’s the whole thing.
You’re going to be oversold on a lot of capabilities. The third piece I want to bring up is the problem with the information. Depending on the data set you’re using, ChatGPT, Bard, whatever it is, they’ve all been trained on a different amount of information, and they all have different shelf lives on that. So, ChatGPT, in particular, says that it doesn’t know things about the world before 2021, other than things that have been programmed into it right now. So, there’s an old moniker that was popularized by the American President Ronald Reagan called “Trust, but verify.” And I think with artificial intelligence, you can’t do that at all. I think you’re going to have to verify first and never trust, because the fact of the matter is, as it currently sits in the current incarnations of the material is that it’s just not right. It’s just not right. Google lost $100 billion in market value in a day because its public unveiling of Bard was wrong on a data point. And you can’t risk everything on that. It’s like you want to read it first. And, at the same time, I don’t think you can let it go alone.
I don’t think you can ask it or trust it to respond on your behalf. It’s like we’ve had self-driving cars, or cars that supposedly or allegedly are able to self-drive, and they’re great when everything’s normal and everything is going according to plan. But the minute something weird happens, a ball bounces in the street, or something else happens there, they run over things, they have problems. And I think you’re going to find that, with artificial intelligence, when nuance and context are important, I think that’s when you’re going to have the biggest issues with it. So, I think you’re going to have to verify this. Think of this as a very earnest intern that you have working for you, and the other thing about it, you can… This intern is never going to think they’re wrong. This intern is going to tell you something with the absolute sincerity. I was using one of the AI tools the other day, and it absolutely insisted that I give it inputs in a certain fashion. Then I found out after researching stuff for 30 minutes, I was like, why isn’t this working? Why isn’t this working? I found out 30 minutes later that those inputs don’t work at all.
But it told me with absolute conviction that you could do the following thing with its software and it wasn’t true at all, and it was ridiculous. So, you’re going to have to verify a lot of things. It’s an incredible tool and you’re going to have to work this in your business. You cannot dismiss it, but you’re going to have to check it. It’s going to save you a lot of time, but it’s going to create other work for you at the same time. The fourth piece, and this enters the legal regime questions we want to talk about, is there are massive confidentiality problems embedded in the artificial intelligence segments. If you feed in, and I would say 99% of the platforms, I’ve only found one that allegedly segregates your data and doesn’t train the data and provide that to other people within the system or the network. It doesn’t automatically add to the database. I found one, and I’ve looked at dozens and dozens of these pieces of software now at this point. And the funny thing that I want to bring up about that is that software has a free trial. During that 14-day free trial, it’s allowed to use any data you give it to train itself on systems.
Even there, and you tell me in those first 14 days if you aren’t going to start shoving stuff in there that’s not confidential. It’s rough to say, hey, try my software out, see if you like it. But just to let you know, the first two weeks when you’re testing this software out and finding out if you like it, I’m not going to keep your stuff confidential during that time. So, it’s a major issue. So, there are massive confidentiality problems. And then, at the same time, you’ve got to start thinking about this in like, how does this work for legal stuff? How about in the United States? We have HIPAA we have to deal with. When you start talking about confidential information, what if you, oops, I copy and pasted patients’ information into this platform and now it’s publicly available? I say that John Doe is subject to the following, or I reveal some fiduciary information from an accounting standpoint. There’s a lot of pitfalls out there in this. And there’s a question, GDPR, the European Union’s Privacy Regulation requires there’s a thing embedded in it that you have a right to be forgotten.
How, if you provide something in a third party database, how do you verify that it’s ever been removed or cleaned from there? And I will say that a lot of these shoppers right now don’t have any way to solve this problem. Not yet. They’ll have something in the next year or two or something like that. There will be lawsuits and then they’ll figure it out. But until then, we don’t know. We just don’t know on some of these shoppers yet what that looks like. And I think you can’t go out there and expect to do all the stuff and make sure that information rings confidential. If you have confidentiality clauses in your contracts, that’s something you need to think about. If, and I will go ahead and say that, if you actually put confidential information out there, like if you asked it to help you come up with a pricing schema or something for whatever it is, and you feed it into one of these open databases, and your competitor comes along and asks that information, you’re going to be informing your competitors on how they can compete with you. So there’s major issues with this.
Eventually, at some point in the future, maybe having your own deep learning neural network is probably part and parcel of what advanced companies will have. Maybe everybody will have their own private Jarvis, Tony Stark that can help you make the next Iron Man suit. But until we get there, no one’s going to have that right now. And the last piece of this I really want to advise people on or talk about is there’s open questions dealing with copyright and trademark. Copyright, primarily, not really trademark so much. There’s a question right now over if any artificially intelligent lead generated content is susceptible to copyright. And it depends on every legal system is going to have a different answer to this question. In the United States, the Copyright and Trademark Office has said that solely generated artificially intelligent content is not susceptible to copyright. So if you use it to write computer code, if you use it to write a sonnet or a song or something else or a piece of artwork, just on its own, it’s not protectable. So I’ve always said this would be funny. So someone creates artwork for the front cover of their new novel using artificial intelligence, using something like Dolly, and they put it out there on their New York Times best selling novel.
Aside from the infringement of trade complaints and the other legal regimes, but from a pure copyright standpoint, I could put my book out there next year’s, which is not a best seller, get same bookshelf space and have the same cover on it and maybe a different title. But what percentage of business would I steal from you? There’s a question right there as if you’ll have any claim against me because it’s certainly not copyright if your image was not copyrightable on its own. And there have been some questions where the copyright office has generated copyright if there’s enough human direction on the robots. And there’s a question there, like, who owns the copyright? Did the person who invented the code that made the machine, will they hold the copyright? Is it not copyrightable at all? Will you have copyright if you gave it enough direction? How will you have to document this information? Because right now there’s somebody saying they did a whole bunch of steps. That’s a lot harder to copyright than just something you sit down and sketch out on the back of a napkin. With that, you just say, I did that. It’s not a napkin.
You file it with the copyright office in DC in the United States, and you get copyright protections from the moment you’ve created it. There’s a big question about how this will even work in this new regime. And as we see, Getty Images just filed lawsuit because a lot of their images were being lifted wholesale. There’s a big privacy question about this. This information of yours that you put on the internet that you had copyright to has been taken wholly without your permission. There was no opt out. There was no way for you to say that you didn’t have to do this. And there will be, like I said, a tsunami of legal questions about this coming out in the days and coming months. Now, this is not to say you shouldn’t adopt and use artificial intelligence in your business. And this brings us to the no BS segment of the podcast. I’ve said it before, I’ll say it again, and if I could brand it across your forehead so you’d see it in the mirror every morning, I would. You must lead from the front. The buck truly stops with you. You cannot rely on other people who are not signing the checks at your company, who are not making the big decisions in your organization to take the lead on this initiative.
You cannot tell yourself, I’m too old to learn a new technology. If you are, it’s time to pass the torch to somebody who will, because these big decisions must be made by you. You must set a framework and a regime up for how you’re going to use these things in your company. And if you don’t have an idea of how the technology is used, you can’t make intelligent decisions about it. Let me repeat that. If you do not know how the technology is used, if you have not used it yourself, you cannot and you will be unable to make intelligent decisions about it. And saying that you’re not going to use the technology is not an answer. As we said in the previous podcast dealing with this, your company will never be worth more than it is today. If you are not going to adopt artificial intelligence, it’s time to look at selling it or moving it on to a new generation. That’s where you’re at. And that is the no BS segment of the pitfalls and perils of artificial intelligence. So where does this leave us? Well, there’s a lot of open questions out there about this.
Right now, the point of this three part series dealing with artificial intelligence was to give people a framework they can work in, to do some of the heavy lifting for you, at least initially, to let you know what you’re looking at. But at the end of the day, what you’re going to need, and this is your three takeaways. Number one, you’re going to have to build a plan and a framework. What technologies you’re going to use, how you’re going to use them, and how you’re going to introduce them to your team. That is a requirement of you using artificial intelligence in your business. And if you’re going to use this new technology, which I encourage everyone out there to start using in some way today in their business, you’re going to need a plan and framework, and you’re going to have to revisit and update your plan. This is going to change so quickly over the coming days, weeks, and months, that you’re going to have to revisit this plan at a minimum, probably once a quarter on the outside, I would tell everybody probably once a month is enough. If you’re using this technology heavily in your business currently, you may need to look at it weekly.
But this is something you’re going to have to look at right now and it will look and revolutionize industries. And you’re going to have to make part of your plan in this extra time it’s going to save you because it’s helping you do certain things in your business. You’re going to have to go ahead and carve out time to look at the new technologies to test them and see what’s happening. You’re going to need to look at in any piece of technology you get, you’re going to have to look at those, though you may never have read a license agreement before for any software you’re using in your company, you’re going to have to read it and say, Okay, what’s the privacy rules on this? What do those look like? Maybe you should contact an attorney to help you look at those things. If you’re making big decisions and dealing with confidentiality, I think with confidential information like HIPAA, etc, you probably need to get legal counsel involved to make sure that you’ve at least covered all the bases. And that’s beyond the scope of this podcast today, but it’s at least letting people know what they need to look at.
And the third part of this is you’re going to have to lead from the front. This is one of those places that you as an owner, a decision maker, a board member on the company where you cannot just say, Hey, I’ve deputized so and so to look at this. This is not like just buying, getting a new vendor or something like that. This is something that you could really get yourself in hot water with very quickly. You need to ask the people in your organization to help you with these things, but you yourself need to be fully versed on this new technology and see how you’re using it. At the very minimum, you’re going to have to look at either outsourcing this to a vendor who is the dedicated expert who can help shield you from some of that liability. But I think like anything else, just like having someone in your own organization do it, I think just because you’ve got the plucky person who’s coming up in the mail room and she’s a hard worker and everything else, you can’t just say, Hey, you’re young, you like new technology, will you look at this piece of technology for me and tell me how I need to use it?
You cannot abdicate your responsibility. This is not just like using a new social media platform. This has so many pitfalls attached to it and everything else, and you have to use it that you’re going to have to figure out how to walk through this minefield and you’re going to have to lean from the front. That brings us to a close on our three part series, at least the first one dealing with artificial intelligence. If you’ve liked what you’ve heard here and you’d like to hear more, please like and subscribe on Spotify, Apple, or Google, or wherever great podcasts are hosted. Again, I’m your host, David Maples. I hope you enjoyed what you’ve heard here. Please give us a thumbs up or a five star review. With that, go out there, use some AI and make it a great year.