Episode 5: Can You Trust AI?
In this episode of AI at Work, host Brooke Torres sits down with Anthony Habayeb, Head of Partnerships at BotChain. Tune in to learn more about BotChain and how they are leveraging blockchain to provide trust and transparency for AI agents.
Brooke Torres, Director of Marketing, Talla
Anthony Habayeb, Head of Partnerships, BotChain
Brooke Torres: Hi everyone, welcome to AI at Work. I am Brooke Torres, and I'm our director of marketing at Talla, and one of your hosts for this podcast, where we talk about what's happening in the artificial intelligence space, in the world of work, and how the two are transforming each other. Today, I'm with Anthony Habayeb, who is the head of partnerships at BotChain. So welcome, Anthony, and thanks for joining us.
Anthony Habayeb: Good to be here, Brooke.
BT: Anthony, tell me a little bit about your background, and about BotChain.
AH: My background is not in the AI or data science space. I've spent a lot of time building businesses focused on growth, accelerating early stage businesses, companies, or business units. I came across Rob May, Talla and BotChain. Artificial intelligence was an area that I was super interested in as an emerging field. I was curious to see how I could apply some of my go-to-market skills into this new technology arena.
BotChain very much falls into that. BotChain is a project and a business opportunity that is trying to operate right at the intersection of artificial intelligence and blockchain. If you were to go around a room of a bunch of entrepreneurs in Boston or in Silicon Valley, and say, “What are two or three of the things that are most interesting industries or technologies that are coming up in the marketplace?” I think you would probably find AI and blockchain are sitting towards the top of that list. It is uniquely interesting to spend some time on a product that works with both AI and blockchain.
BT: To back up, I first want to ask you a question about AI in general. Given that the people who are joining us for this podcast are thinking about, and deploying AI products for their employees, and for their end users if they're a consumer-focused company. Can you trust AI? Should you be unleashing this on the people that really matter to your business most?
AH: Yeah. It's a great question, and I think I have a unique perspective, because I have not been spending years and years in data science and artificial intelligence. And what helps me first-- for someone that might be listening and thinking about AI in the early days-- it's a very broad term, right? So sometimes you could view artificial intelligence as automated software. As proscribed and automated processes and workflows, right?
Then other people think about Robocop, and think about truly non-predictive human-like machines. It's a very big span. So across that span, is there a benefit, and should we as business operators be thinking about how to automate things? 100%. Should we be trying to find things that can be fed information, and could be sort of scripted and developed in a way that has an expected output? Yes.
Are there fears around, how you build something where you don't know what's going to happen on the back end of it? That is a concerning thing, but you can also build in a way that has certain parameters. As a manager, you often think about trying to empower your employees-- with an ability to fail, but do too much damage, or hurt themselves, or hurt the company, right? Like, failure can be a good thing.
Having machines learn means that they will make mistakes sometimes, and I think that there are responsible development practices that I see happening at Talla and in the ecosystem that can be applied as a company's implementing some of the non-predictive AI. I think that's really where your question sits: artificial intelligence which has non-predictive outcomes. That's where we start to really think about the ethics, and the risks, and the concerns. All of those things, in time, I think are continuing to be managed responsibly.
BT: I know that BotChain is a solution to help with exactly some of the things that you're talking about. Talk to me about BotChain, what it is, and what it does, specifically
AH: One of the things, when you hear trust in artificial intelligence, you sometimes hear this phrase, "black box." People don't know exactly what happens inside the AI machine, or the AI brain. That's sometimes what people are afraid of. So, how do you remove the curtain, right? How do you reveal Oz? In that context, first it's: “Who are you talking to?”
You're talking to a machine, not a person. I'm sitting here with you, and I see you, and I can talk to you, and I know that it's you. But, if I'm engaging with an avatar, or I'm talking through a chat platform, how do I know that the thing on the other end is really the machine that it says it is? We've all dealt with spam in our email, we've all probably heard about Twitter bots and Facebook bots that are creating fake news, or scamming people, or getting credit card information.
What if you had a way as a consumer to confirm the thing you're talking to really is the thing that it says? Well, to do that, you would probably need some sort of a solution, or a technology, or BotChain to provide you with a way you can verify that this thing is what it says it is. That's the end user value of BotChain, but the person who brings that value to the consumer is the company.
A company registers its machine, registers its AI, and says for example, “Hey, BotChain, I'm Anthony, I've built this Facebook bot for Bank of America, and I want to register this thing.” Now that I've done that, I've given Bank of America the ability to let their millions of consumers across the country, when they're talking to Bank of America bot, they can verify that that bot really is Bank of America.
That's very powerful. As more and more engagements between humans and bots are happening, an ability for the company to give that consumer trust, that they can share sensitive information with a machine is hugely important. BotChain looks to provide that as one benefit and value. Then on a purely enterprise level, we all are held to standards of compliance and auditability as we operate in the marketplace.
One of the challenges we have in certifying ourselves as companies, when we have AI, is: what did my machine do? It's easy to do a SOX compliance or Sarbanes compliance on people processes, right? Or on auditing of systems that retain a lot of records. AI doesn't do that in the same way. BotChain provides a developer the ability, once something's registered, for all of the actions that the machine has taken to be registered or hashed.
When a machine does something, you can now send a hash of that action to BotChain, and that hash allows you to verify and certify your record set of the machine. That's very powerful. We've talked to companies like PwC and Ernst & Young who agree, if AI machines do that, that will help those companies to be compliant, and, when needed, to be audited and certified.
BT: It's interesting. It's really foundational, and also kind of surprising that right now there are all these big companies that have bots that they've deployed that are interacting with their end users, Bank of America as an example. There's no registry or real certification process for them.
AH: It doesn't exist.
BT: You mentioned that BotChain is a blockchain platform, tell me a little bit more about that. How does it work, and you know, some of the things that are important to our listeners about it.
AH: I'm actually just thinking about some of the prep that you had given me for this conversation, that this is an audience that's interested in artificial intelligence in the workplace. The idea of blockchain, some of the folks listening might not have any foundation for that. I didn't either, when I started with BotChain. What's interesting to me is the basic technology of blockchain gives the ability for something to be recorded in a way that cannot be changed.
The way that happens, really simply, is imagine it that you are at a bank. There's a ledger. What do you have in your account? When you put something in or take something out, it's recorded. The bank has that information. Well, why does the bank need to do that, right?
You could have that action happen, and that action recorded somewhere, let's say on millions and millions of servers, or instances all across the world. Every time we take a next action, that's appended to the thing you did before. Once again, it's on millions of instances. You have this history of all the stuff that's happened that can't be changed.
If you wanted to try to change a record, to go in and edit or delete something that's on this permanent record, you'd have to try to do that instantaneously around millions of machines that you're not even sure what machines are holding the information. You just can't do it. That's really interesting when it comes to providing trust and transparency, which is what BotChain's trying to do.
We are creating a registry of AI developers and their machines, and the things that their machines do, in a way that you can't really manipulate or hide that information. We have a community of, we call them curators, who are rewarded or incented for verifying that the developers are who he or she says they are, and that the things that they've built are working correctly. What we're trying to do is bring this quickly growing world of AI some semblance of definition in standard that everyone can benefit from.
AI needs a lot of data. Facebook, Microsoft, IBM, they have such huge advantage because they have so much data. But, there's a lot of talented developers out there that are trying to build AI companies. You might be one of them on the podcast. As you're thinking about how to build AI, you need access to data.
Blockchain projects that are out there today are building marketplaces of data, where you could go and find bot skills, or AI skills, or certain machine learning algorithms, or data sets they you need to train a machine, and you could get some of that stuff from a blockchain project, a decentralized project that's sharing information. But, would you do that if you didn't know where the data came from? Or who the algorithm was developed by? Is this trustworthy or not?
Blockchain providing a standard for identity can be used by everyone in this ecosystem. We can all know whether we're sharing data, sharing algorithms, registering our machines, providing a certificate of who we are for our consumers, like the Bank of America example. BotChain provides this really basic utility that a lot of developers projects can use to bring more transparency into the AI that they're developing. That was a lot.
BT: Lot of good stuff. You talked about developers, what are some of the things or some of the companies that people would recognize who are working on the BotChain platform? Who's already using it?
AH: Great question. There's a lot of talented developers trying to build AI, and there's a huge limitation in that you need data to train your artificially intelligent machines. Your machine learning algorithms need data. Sometimes you could use platforms to build things, there's lots of tool kits and platforms out there to help build machines or build bots.
We have a couple of examples of that at BotChain. I'll give two examples, Gupshup is a paid bot development platform, and Botkit is an open source project. Both of those companies provide platforms for developers to go in, use their toolkit, use their development tools, and processes and workflows to build consumer-facing bots on platforms like Facebook, or Slack, as examples. So, those two companies already have massive adoption, tens of thousands of developers that use their platforms to build things.
They decided to become partners with BotChain because they believe and appreciate we're working in this world that's growing so quickly, that's lacking certain standards. But also, companies and the consumers who engage with the bots of companies want to have more transparency of who they're talking to and what's being done. Which is very much what GDPR says as well. So they would like to be involved with BotChain initially, in a way that every developer that works on their platform. The machines or the bots that they build, those things will be registered.
Through using Botkit and using Gupshup, you will be able to register yourself with BotChain, and then you will be able to hash what your machine is doing onto BotChain, which enable compliance and audit, as I mentioned earlier. Then a company like Gupshup is proactively thinking about how they will develop. On the web today, Brooke, you know how you see https, and you see Verisign as sort of this trust seal of a website or of a check out experience in a shopping cart?
Gupshup is thinking about how they could bring a BotChain visible experience into all the bots that are built on their platform. So, no matter who the company is that the developer has built a bot for, one of the values Gupshup will bring the market is that any bot you build on Gupshup will be BotChain certified. That will have specific meaning in the market place around transparency, auditability, and trust.
That's a pretty cool thing. The more developers that build with these standards, or leverage the BotChain protocol -- it doesn't cost anything to do it -- we will be rewarding all early developers who register themselves and their machines, or who think that they'd like to build something on top of BotChain. It's all open source. We actually will rewarding them with the tokens from our project, it's sort of way of saying thanks in building the ecosystem around the project.
BT: Just so I understand here, you're saying that BotChain is for the developers, but if I'm a buyer, am I also buying? Am I buying something from Gupshup and BotChain?
AH: Yeah, if you are an enterprise that is implementing AI in your company, you should be asking your team or your vendor to register with BotChain. You should be asking them to hold themselves to a standard of registration, of certification of decisions, of retention of decisions, right? As we evolve as a project, we'll start thinking about things like, when you're BotChain certified, you might also be SOC 2 compliant, HIPPA compliant, GDPR compliant.
All of these things, if you're a Fortune 1000 company, you have very rigid compliance expectations and requirements. Why shouldn't your AI have the same standards? So the buyer should be pulling their AI into this expectation as well. Developers will do this because there are benefits, but the buyers should really start demanding that all of us, as AI companies. Our buyers should be asking all of us to commit to minimum standards, really, basic standards of how we operate, and how we bring transparency into the work we're doing.
BT: So much good stuff. Anthony, it sounds like BotChain is still in development, yes? When will it be in market?
AH: Our plan is in October, any developer in the public could come in directly to BotChain and register. In advance of that, we have beta, and we have early partners that are starting to help us finish development of the platform, testing registration, and some of the early work required from this registry. If you are an AI developer and have any interest at all in participating in the project, please do visit botchain.network, and go to our page where you can sign up for early participation as list a developer. I would love to have you involved in the project, but yeah October is when we expect this is more, if you will, production in some other context. Anyone could come to it and register.
BT: Awesome. So we are bumping up the end of time here, but I want to ask you one more question, which is Rob May, our CEO's favorite question. What you think about the Elon Musk versus Mark Zuckerberg debate in AI?
AH: BotChain's trying to fall right in the middle, actually. As a personal statement, I'm super excited about AI. I am more excited about the possibilities than the risks of it. I guess that's a more Zuckerberg thing. BotChain is actually trying to fall right in the middle.
Back to a comment that I said in the very beginning, we can build AI in responsible ways. Certain standards, like BotChain, which asks every developer to register, and asks every machine to be hashing what it's done. Those are certain, if you will, safe practices. If we all employ those sorts of things, we can both achieve the potential and the excitement that Mark sees in the market, and also try and hedge and protect from some of the risks that Elon sees in the market.
BT: Well Anthony, thank you so much for joining us today. We will be back next week with a new episode. In the meantime, please subscribe on iTunes, or wherever you get your podcasts, and you can learn more about AI at Work the show, and Talla, Inc., the company behind it at talla.com and thank you to Alyssa Verzino, who runs production for us.
AH: Thanks so much, Brooke.