Latest / CXOTalk / Generative AI in the Enterprise (with EY's Chief Technology Officer)
Transcript
- 0:00Today we are talking about generative, AI tools, such as
- 0:06chat GPT, what's the impact on the Enterprise?
- 0:12Our Guest is Nicola Marini bnz, know your CTO of ey.
- 0:17Tell us about your work. I'm responsible for all the
- 0:21technology that generates revenue for the firm.
- 0:23So I spent the last two years developing what we call the ey
- 0:26fabric, which is effectively platform data and and and with
- 0:30seven components actually station Intelligence on top of
- 0:33it, the powers our service offering.
- 0:36So we have about one and a half million clients that are using
- 0:39it every day. It's interesting.
- 0:41We don't think of a technology of a consulting firm as being a
- 0:47technology organization, but obviously you are, I don't think
- 0:50that is an organization that can be not the technology
- 0:53organization of the same time. I think technology is, and look
- 0:57at what we're going to be talking about today, right?
- 0:59The college is Raising in our lives that there is no business
- 1:02that doesn't as in users. And then, you know, if you start
- 1:05thinking about the Consulting organization, when we provide
- 1:08services to our clients that are very data-intensive technology
- 1:13becomes so Central, so fundamental to the quality of
- 1:16the service that we offer the ability of doing and many and
- 1:19across many countries, etc, etc, Nicola products like chat GPT
- 1:25Google is coming out with their own version, Microsoft.
- 1:29Has adopted this now in bing. It's so popular.
- 1:35But why as an Enterprise CTO do you care about this?
- 1:40I think what we're going to see. First of all it's an arms race
- 1:43which is great right? Because usually when the players
- 1:46are competing with each other you know we tend to benefit from
- 1:49it right as consumers in this case, or Enterprises.
- 1:53So it's very exciting over. We're going to be going next in
- 1:55terms of the capabilities of the technology, but be more specific
- 1:59about Jenna. May I and you know I'm looking
- 2:01at my own Organization, for example right?
- 2:04Which is a knowledge-intensive organization, the ability to
- 2:07summarize knowledge in the way. You know some of the language
- 2:12models are capable of doing is absolutely off or opens up an
- 2:17endless number of opportunities for organization.
- 2:20So, for example, you know, how do you systematize the knowledge
- 2:22within the organization, how you being oncologist in
- 2:24organizations? How do you give advice right to
- 2:28clients or how you have? For example, scenarios where you
- 2:31have a sort of a co-pilot, right?
- 2:33So in pretty much any, any one of the roles and the jobs that
- 2:36we have in our organization, right?
- 2:38We would definitely benefit of having an intelligence agent
- 2:42intelligent agent, sitting beside us, you know everyday not
- 2:45replacing us, but supporting our knowledge and our capabilities
- 2:49trying in everything that we do. So endless possibilities.
- 2:54And so most and also not only in the Enterprise values in our
- 2:57lives. I think that would be a In fact
- 3:00as a practical matter as you think about these generative AI
- 3:07capabilities, how do you, how do you think about a do break it
- 3:11down into areas that you think might be useful?
- 3:13Or is it still just too early of an ex exploration phase or where
- 3:19are you in your thinking? I think he's going to go very
- 3:21very fast. What already exploring?
- 3:24I mean in our specific business, right?
- 3:26Which is to give advice to clients based.
- 3:29On our knowledge or regulations that our knowledge of other
- 3:33interaction with other clients. Etc.
- 3:35You know the ability of you know tapping into the knowledge in a
- 3:39way that is multi-dimensional like you know some of these
- 3:41tools can do is absolutely incredible and and fundamental.
- 3:46So we have several projects already Eddie why that are
- 3:50tackling that. But I think, what is most
- 3:52fascinating about this? Latest development is so we had
- 3:56chatbots for a long time, really and not know exactly when
- 4:00Laughter. But you know, obviously some
- 4:01companies had their child boss for at least you know, eight to
- 4:04ten years, right? But what is different about?
- 4:07This one is the fact that you actually can have a conversation
- 4:10with it. Sometimes this job boards you
- 4:12know you are doing Custom Image as a cook as a as a consumer.
- 4:16I you go onto a website because you have one I have some more
- 4:19information about the product so you are complaining about the
- 4:22protocol something right? You get into these very
- 4:24frustrating, you know, interaction with the chat bot
- 4:28that if you don't. Exactly Hit the question, the
- 4:30way, the child bought the specs, it then you go in nowhere and
- 4:35you list is what I do. You get to a point where you
- 4:38say, please I want to speak to an agent, a real human, right?
- 4:42And this will these tools now that are coming to.
- 4:45The market are actually very different because they allow you
- 4:48to actually have a dialogue because they can maintain the
- 4:51state of that conversation, right?
- 4:52So the previous job boards, you ask a question and get an answer
- 4:55and that's it. There is no memory of that
- 4:58interaction. That the Book everything but now
- 5:01you can so you can go. It's almost like I call it but
- 5:04maybe a little bit abusing the term by sort of a dialectic of
- 5:09AI. Right?
- 5:09So where you ask a question, you get an answer and you can
- 5:12another you ask another question and you get closer to what you
- 5:15want to get out of it as opposed to a sort of an expert system
- 5:19where you know, it's question and answer and that's pretty
- 5:22much it, right? So it's very interesting those I
- 5:25think that is the feature that to me is most exciting.
- 5:29Would it be Correct to say that you feel the defining
- 5:32characteristic is this ability to remain to to retain the state
- 5:39of the conversation? There are so many other things
- 5:43that are great, right? The ability of writing in in
- 5:47multiple languages, right? So, I'm Italian original, right?
- 5:50So, that's my first language. He writes as good in Italian as
- 5:53it is in English, right? And you can ask the questions,
- 5:56you can you can Envision like you know, Meetings where, you
- 6:02know, you have everyone speaks his own language, really think
- 6:05it would be an incredible for global companies will be an
- 6:08incredible achievement. Right today, we still have to
- 6:11there is still a lot of Lost in Translation that these tools are
- 6:14so refined and sophisticated now that even the nuances of a
- 6:18translation can be really, you know, highlighted in the right
- 6:23way, which is incredible. We have a question that came in
- 6:28from Twitter. Already, which is an important
- 6:32question from our Salon con are salons, a regular listener.
- 6:36And I've always thank Our Salon for his wonderful questions.
- 6:39And he says this, he says, when we use AI as a co-pilot, we have
- 6:45to be careful about the data. It is pulling in order to give
- 6:50us suggestions if we aren't aware of biases or stereotypes
- 6:56in the data than how what should we Do it's one of the
- 7:01fundamental questions of dealing with these types of
- 7:04Technologies. Absolutely.
- 7:06I had lots of people that inside my own company, right?
- 7:09The rich are say, I want to try GDP, I want to.
- 7:11How can we use the with clients? I said in a so it will do
- 7:14caution in. I think it's where warranted at
- 7:17this point, right? Meaning that you are the human
- 7:21in charge, you know what I mean? So it cannot delegate completely
- 7:26the work to these agents because they're not yet at that, Level
- 7:30of being able to understand, you know, how ethical code, right
- 7:34the morality. Then you know what is acceptable
- 7:37or not acceptable? Unless it has been labeled by
- 7:40another human by. So the human in the loop
- 7:43approach is absolutely essential here, right?
- 7:45So you should so I use it as a sort of a, an aid as opposed to
- 7:50a delegation of responsibility, right?
- 7:52That cannot happen. So the human has to be in charge
- 7:55and hopefully the human. I mean you can argue that even
- 7:58the human-sized a lot of Show flaws, right intensive, what
- 8:01kind of answers? You know is capable of giving
- 8:04but you know what not a pusher at the point yet where we can
- 8:08delegate a function to you know, a tool like this.
- 8:12I heard a very interesting comment yesterday on the news
- 8:17from a journalist who said, Who as a reminder that tools like
- 8:24Chad GPT or what Google is going to come out with all of these
- 8:27tools, they're really like advanced auto correct on your
- 8:32phone. They have they give the
- 8:34appearance of sentience but they're just it's just like
- 8:39autocomplete. We could go into it.
- 8:42Like a very long philosophical. Conversation was you know, what
- 8:46is sentence was sent into really means but I agree, I think I
- 8:49mean I would say the outer item is a little bit to trivialised
- 8:54right as a definition. Definitely is now.
- 8:57So if you autocorrect is on one end of the spectrum, It's
- 9:00awesome. And on the other end is
- 9:02sentences. Like it's it's in the middle,
- 9:04right? So I think we need to
- 9:06understand. So sometimes, you know, we tend
- 9:09to whenever there is a new technology especially this kind
- 9:12of technology that a little bit more difficult to understand
- 9:15because they're actually quite complex, we tend to give them
- 9:18either, you know, kind of transfer some sort of
- 9:22alienation, right? We transferred to them the
- 9:24expectations that we have from another human being, right?
- 9:27So if we do that, that is the wrong approach, right?
- 9:30This is not a human being. This is the bunch of silicon,
- 9:33you know? And a lot of data that gives you
- 9:35some kind of an answer. But if you actually think about
- 9:38it as a sort of a in a dry in taking, an aeroplane to go from
- 9:43San Francisco to New York, that is what these tools are.
- 9:46So it's a machine that helps you do things faster, right?
- 9:50But if you think about it it's not that different than at the
- 9:54beginning, truly of the mass adoption of the internet, right?
- 9:58So in the eye, I cannot do anything in the 90s from
- 10:02University. Right.
- 10:03So whenever I had to do something research paper, I had
- 10:07to go to my library in my city to find the physical book and,
- 10:11you know, taking forever to get the information out excited.
- 10:14And then the search engines came up and that was like an
- 10:18incredible because you can access the whole knowledge of
- 10:21the human race, you know, either the keystroke.
- 10:25And so it doesn't mean that we assign to that technology like
- 10:29Superhuman ability, right? It's something that helps you
- 10:32search things better and the saying I think we need to look
- 10:35at these new Tools in that same way show me what's possible.
- 10:38So show me the search criteria. Show me what the search results
- 10:42are and then I need to be what the one that makes the decision
- 10:46on what is relevant and what is not.
- 10:48I think one of the challenges for folks in the Enterprise or,
- 10:52or just our personal lives, is that the answers these tools
- 10:57give comes across as being Authoritative.
- 11:01Yes. Yes.
- 11:02But it may be totally incorrect. Absolutely.
- 11:07And that is what we shouldn't. This is the thing, right?
- 11:10If we think about it, we can start talking about human
- 11:14characteristics and we transferred them to describe the
- 11:18tool we make a mistake, right? We will never call, you know, a
- 11:24car like, you know, fast legs, you know, maybe that's the wrong
- 11:27example. But you said, I mean somebody if
- 11:29we Use them and say, okay, so we need to put them in sort of the
- 11:32in the box, right? And say like you know, today you
- 11:36do some of the search or Google do a Google Search, right?
- 11:40And it comes back like with millions of pages, sometimes,
- 11:43right? So it's up to us to select.
- 11:45What is the page that has the highest level of relevance of
- 11:49what we're looking for, right? We need to look at it the same
- 11:52way as opposed to. Okay.
- 11:55That is a you know, it in a super lap.
- 11:58You superhuman intelligence Since the, you know, knows
- 12:01everything, right? So that's why I think, you know,
- 12:03when you think about that Katrina as well.
- 12:05I think if we start relying on these things too much, while we
- 12:09lose the fundamental understanding of a domain or the
- 12:13subject, right? It can become a little bit
- 12:15dangerous from that perspective. Please subscribe to our YouTube
- 12:19channel. Hit the Subscribe button at the
- 12:22top of 60 talk.com, our website and subscribe to our newsletter.
- 12:28Well, we have another Question from Twitter that's related to
- 12:33this. And this is from Chris Peterson
- 12:35who says, in business Communications, whether it's B2B
- 12:39or b2c. Do we have an ethical duty to
- 12:43call out what's written or edited by an AI assistant?
- 12:47He says, if a human reviews, it modifies it and hit, send, does
- 12:53that change this ethical duty today, when you publish a paper
- 12:59There is a process to publish a paper, right?
- 13:02So you need peer reviews, you need to, you know, cite your
- 13:05sources, you need to do all these type of things, okay?
- 13:08You cannot just say, I came up with this right without, you
- 13:12know, that that would be pleasure is MM, right?
- 13:14So I think there has to be a framework similar to that one in
- 13:20this space, right? So if I'm like doing, for
- 13:22example, like, you know, a client, for example, asks me
- 13:25about whatever some regulation somewhere, right?
- 13:28I can do some research and I can use charge apt or can use
- 13:32another tool that I can use, you know, my own traditional, you
- 13:36Google searches or something, to get the information out my own
- 13:40internal knowledge and provide that answer, right?
- 13:42So when I do that in an email, right?
- 13:45As long as I think, you know, it is fine to do it as a human
- 13:48being, right? So I think, I don't know why
- 13:50there shouldn't be any other way, but if I'm producing a
- 13:54paper just using that right relying on somebody else's work
- 13:57and I don't, you know, to That and I don't do the citations and
- 14:01I don't do this. And that then, of course, you
- 14:04know, I'm getting into plagiarism.
- 14:05I think it's the same way that a human would be doing in a
- 14:08research paper like that. So complicated, so I think that
- 14:11is the need of especially in the IP in the hole, you know, legal
- 14:16discipline of intellectual property, is probably as well.
- 14:21That there is a need of some kind of a rethinking about some
- 14:24of the fundamentals of that because this is different.
- 14:27Yeah. And clearly There are a whole
- 14:30set of ethical and legal issues. Absolutely, right.
- 14:36So and and so that's why I think we need it's funny that
- 14:39philosophy, right? Probably the day ancient most
- 14:41ancient academic discipline is still, you know, very much alive
- 14:46is because we need to ask ourselves so these questions,
- 14:48right? And we need to have Frameworks
- 14:50to to address them. Like, you know, it wasn't
- 14:53considered when the the beginning of intellectual
- 14:56property came up as a protected. You know, right that the in
- 15:02those days like when the legislation started to get
- 15:06introduced to wear similar conversations, I'm expecting,
- 15:09right? So you're not supposed to take
- 15:11somebody else's intellectual work and make it yours, right?
- 15:14And so I think we need to have that dialogue today.
- 15:17What is going to be much more difficult today?
- 15:19Is that these tools that the marginal cost of producing or
- 15:25generating, you know, this type of documentation.
- 15:28So, whatever we generate is is almost 0.
- 15:31So we will be inundated by this volume of documents papers,
- 15:38books, images and stuff and will be really hard to understand
- 15:43what is really generated by human.
- 15:45What is not excited. So, you know, maybe the
- 15:47Christmas asking, like, identifying that content be, you
- 15:51know, a really good step, first step, I guess.
- 15:55And this is our Salon. Khan comes back again and he
- 15:58says when everyone Is quote doing AI?
- 16:02What sort of governance structures should be placed?
- 16:06And I love this. Lets you know, let's let's
- 16:09right. Let's write a thesis on this.
- 16:11What sort of governance structures should be in place at
- 16:14an organizational level as a governmental and as a global
- 16:20level, what are we just simply start with organizational
- 16:24governance of a iine that's going to become important.
- 16:27Very important. Any thoughts on that?
- 16:29So, for example, I was sunk Mentioning the example of a
- 16:32knowledge company like ours, right?
- 16:34So first of all, you want to be able to store then knowledge in
- 16:39a way that is relatively secure in the sense that no secure, I
- 16:43mean, of course, from attacks from the outside, but that's one
- 16:46piece. But also secured in the sense
- 16:48that is the knowledge that the company wants to protect, right?
- 16:52So the Nola. So if you think about an
- 16:54ontology within an Enterprise, right?
- 16:56And you have the definition of some entities or concepts, you
- 16:59want to make sure That definition is shared by
- 17:04everybody. It is approved, Etc.
- 17:06Right? So I think that is the kind of a
- 17:08need of establishes. A sort of in a, I'm gonna say
- 17:11this but probably the wrong word, right?
- 17:13It's all sort of in. It is toriel board, right?
- 17:17For these type of knowledge, the guess systematized within the
- 17:21knowledge repository of the Enterprise.
- 17:23So the concept of you know, what is revenue, for example, for a
- 17:28in all the company, right? It can be different.
- 17:30Then what is used in on the street.
- 17:33And so when people search for the word Revenue, there has to
- 17:37be an approved definition of it behind the scenes.
- 17:40So to me that is highly curated content so you have to have an
- 17:44organization I believe that looks after that make sure that
- 17:48nothing you know we've asked Revenue only gives you another
- 17:51answer, that's not good. Okay.
- 17:52So that type of, you know, potential Divergence has to be
- 17:57understood as to be prevented and has to be Much.
- 18:01So that to me is the key with other areas where they kind of
- 18:05it, you know, being true to the book definition of things is not
- 18:11that critical. I think, you know, a little bit
- 18:13more dialogue could be, we welcome, right?
- 18:16And I think we need to lacks the tool itself, learn and evolve
- 18:20Etc. So too much controls and too
- 18:22much governance, and I think is going to be a good thing.
- 18:24So I guess the answer is it depends right?
- 18:27Technical Consulting answer, but it depends in the Is that, you
- 18:30know, what is the type of knowledge that you want to
- 18:33maintain govern and protect as opposed to the other ones that
- 18:36you want a little bit more free flowing and and more
- 18:39interactive. That makes a lot of sense and
- 18:44looking at other important domains inside an organization
- 18:48that require governance. Just, this is a natural kind of
- 18:53a pro to use that as a model. But then be open to the fact
- 18:57that, as you said generative a I will Change and evolve over
- 19:01time. And so we had we can't be to
- 19:04lock down about it. You cannot?
- 19:05You cannot constrain it too much.
- 19:07I think because otherwise you lose the value of it, right?
- 19:09It's just curate and knowledge. And you know, if there's a
- 19:11knowledge it's it's it's real view, more knowledge, right?
- 19:15So you don't, you don't, you're not able to anticipate what's
- 19:18coming. We have another really
- 19:20interesting question, short question, hard question from
- 19:25Anil Vos and he says, what kind of jobs?
- 19:30May become redundant by generative, AI tools.
- 19:35I don't think this technology yet, okay.
- 19:38I don't know about a next five to ten years, okay.
- 19:40That's different. But right now, I don't see the
- 19:44again, the delegation of the function completely towards
- 19:49another, to even if in the Creator's economy, right?
- 19:52Even if you create art like digital art in your life, that's
- 19:56your job. I don't think he can be
- 19:59completely created by the machine, right?
- 20:01We're seeing, you know, a I asked Etc but at the same time I
- 20:04think if you are an artist and if you are the Creator, that is
- 20:08a lot of additional value that you can add on top of what the
- 20:12machine creates by itself, right.
- 20:14And also, in the Enterprise, I see more jobs created than less
- 20:20the jobs disappearing, right? Because I see a lot of
- 20:23opportunity for example, for you know, people that can
- 20:26systematized. In a different way, right?
- 20:29Or can, you know again? Like I am very big on this
- 20:34concept of ontology is because I think if you can store the
- 20:38Enterprise knowledge, you know, they're its value that
- 20:41translates into shareholder value, right?
- 20:43And it's also like, in allows you to protect the IP and the
- 20:48capabilities of an organization, right?
- 20:49So think about our organization 360,000 people, eight hours a
- 20:54day, is 2.4 million hours of work every day issue.
- 20:57Can capture the knowledge structure it in a machine, and
- 21:01then having give access to it to our clients, at our own people,
- 21:06that makes our the, our company even more valuable as what it
- 21:09is. So I see more opportunities than
- 21:11less. If that makes sense, you have
- 21:14been alluding to and talking about this Knowledge Management
- 21:19function. How does generative a I change
- 21:24Knowledge Management relative to this?
- 21:27Torkoal tools. If I look at my company, right?
- 21:30And not going outside of our business that is to me is the
- 21:35the low-hanging fruit. Okay?
- 21:37That is so the fun day we had able to access all the knowledge
- 21:43and the fingertip in a way that is human friendly.
- 21:47So today, right? If I have to say for example in
- 21:49the in our tax business, if I had to help a client you know
- 21:53understand the taxation regulation around 10 count It's
- 21:57usually I had to read like, you know, millions of pages of stuff
- 22:00Ryan and has to be an army of people doing it.
- 22:03If I have a tool that's not that, you know, with level of
- 22:06confidence summarizes. So that information I can spend
- 22:09more time talking about client with clients about their options
- 22:13than just, you know, reciting what the regulations of the
- 22:17right. So people will be happier will
- 22:19have access. We have course, I've everything
- 22:21has to be cured as I said, right?
- 22:22Especially in this type of businesses, as to be curated and
- 22:25management government class. I can The wild west, right?
- 22:28Because you don't wanna, you know, then the machine thinks
- 22:32that, you know, we're looking at the legislation of Switzerland.
- 22:34And in reality, we need to talk about Austria, right?
- 22:37So, that's, that is a potential issue, right?
- 22:41But when you have done that, I think, you know, the research
- 22:44and the work they will be doing is much more, will be more like
- 22:48the number of hours that we spend doing more valuable work
- 22:51will be. I think hiring people will be
- 22:53satisfied and the club, the quality of the work that we'll
- 22:56be doing. - 1. How will these kinds of tools
- 23:01change work inside and Enterprise in every business
- 23:07application, right? The one that we build that we
- 23:10don't have it yet, but, you know, we're going in that
- 23:12direction. I still have like I will have my
- 23:14little, you know, some window or the frame on this side of the
- 23:19main screen that allows me to ask questions on the fly in real
- 23:22time about what I'm doing. And so think about like someone
- 23:26gets out of school. I am 23, 24 years old, out of
- 23:30University to joined our company, right?
- 23:33Instead of, you know, said getting them to do like, you
- 23:35know, weeks and weeks of training and asking making sure
- 23:39that they get to the right person as a mentor excited that
- 23:42they can do that, of course because the human relationship
- 23:45is so important. But at the same time if they can
- 23:48access their knowledge, institutional knowledge,
- 23:51codified in the system, I mean there will be there will be much
- 23:54easier for them to work and to grow, right?
- 23:57So I think think that to me, that's what the copilot concept,
- 24:00right? It's not like it's the function
- 24:03will be automated or replaced but I have this agent that I can
- 24:08talk to and ask questions, we fantastic.
- 24:10I wish I had that beginning of my career.
- 24:13I to a fair amount of writing and I use chat GPT and I just
- 24:20subscribed or was put on the, the user list for Microsoft's
- 24:26Bing and I can't wait. For Google to come out with
- 24:29their product. And I actually subscribe to
- 24:32several of these products. And as a co-pilot I have to say
- 24:37it's a phenomenal, I mean, the body of knowledge, the
- 24:41efficiency that it helps. And just coming up with new
- 24:44ideas, it's very fast much faster than than I could do on
- 24:49my own absolutely and the new ideas I mean they're not really
- 24:53the machine, this is the thing right?
- 24:54We shouldn't think about. So Alan a thing.
- 24:57Us in the sense that we give these to, you know, too much
- 25:03responsibility in terms of, you know, the level of intelligence
- 25:07because ultimately there is always like a today.
- 25:10At least, there is always a human behind the scenes that has
- 25:13created the content. I structured in a certain way as
- 25:17trained the model in a certain way so it's still human
- 25:20knowledge, right? But it's but the nice thing
- 25:23about it is that you can have access to it immediate.
- 25:25So you know, I was making the Earlier of the the old school,
- 25:30you know, Heart, You Know, cover tackle library, right?
- 25:34You have to find the book, you know in the library to be able
- 25:38to talk about the specific domain here.
- 25:40You can we can have access to an incredible amount of data like
- 25:43very very quickly. We have another really
- 25:46interesting question from Anil Vos.
- 25:50And he says, if the data set that trains, the generative, I
- 25:56say I have As an intrinsic bias, how can we get rid of this bias
- 26:01or overcome this by us? So that it is neutral.
- 26:06And I think he's talking in two things.
- 26:09One from the point of view of the software developer, and
- 26:12number two, from the point of view, as us as consumers as
- 26:17users, the people that manage did, they send that drains the
- 26:20moment, right? That's what fundamentally comes
- 26:23down to, right? So if you feed it in the model
- 26:26racist, You know type of Concepts, write the model will
- 26:30Jason that so that is 0 bias, you know, of any kind.
- 26:34So it's a fundamentally it's a human responsibility, right?
- 26:38So in Enterprise has, as I said earlier, we need to govern that
- 26:43we need to have, you know, a sort of a, first of all we need
- 26:47to have values, right? The Enterprise level values, the
- 26:50things that you wouldn't do and not do because they will help
- 26:54guide. You know what is the you know
- 26:56what is the Onset of bias and how people should should should
- 27:00should manage that part, right? And then at the same time, I
- 27:03think, as I said earlier, during the has to be in an authorial
- 27:06Board of some sort that oversees and takes the responsibility or
- 27:13food for what goes into the machine ultimately, right?
- 27:17So, I think it wouldn't be there is risk for sure, right?
- 27:20And it's something that we need to take very seriously.
- 27:23This is again from our Salon, Khan who says, when it comes to
- 27:26life and death Situations. And the AI is wrong, who should
- 27:32we Sue AI? Vendor the data owners, the
- 27:38algorithm creators, who do we Sue?
- 27:41When there's a, when something really goes wrong, I think we
- 27:44should not get into position when it's a question of life and
- 27:48death left to a machine. That to me, is the number one.
- 27:53So if we had to make, if you have to get to a position where
- 27:57the machine has to make a decision and it has, you know,
- 28:00the power of, you know, driving an event of this level.
- 28:06I think there is something wrong in the way of approaching it.
- 28:08These Technologies are not at that level yet, right?
- 28:12So we do and we don't want that, I think, yet until it can be
- 28:16proven in a different way. So if let's say it's a medical
- 28:20Your diagnosis that has to be provided.
- 28:24The expectation is that there is always a physician in charge
- 28:27with the help of the tools, right?
- 28:30But fundamentally it's down to the human.
- 28:33So I don't I mean everybody really even you can think about,
- 28:38you know, just stretching out a little bit from just the Genera
- 28:43to the eyepiece of it even for we're not ready yet to do full,
- 28:48you know, full a full self-driving cars.
- 28:51Right foot for the number of reasons, but one of the reasons
- 28:54is not just, the college is about their kind of a who's
- 28:56liable. Who's accountable?
- 28:58What kind of ethical decisions they need to be taken excited?
- 29:01So if you're not already there which is in a way, the more
- 29:04mechanical way of doing it, we shouldn't be ready to add to
- 29:08delegate. Again, you know, the
- 29:09responsibility for these type of decisions to adjust the machine.
- 29:13Yet we have another great question from this time.
- 29:17From Elizabeth Shaw on Twitter and she says So why do tools
- 29:23such as Google barred or chat GPT hallucinate even when it's
- 29:30using correct Source data and what are the implications there
- 29:36for for using these kinds of tools for Knowledge Management
- 29:41inside an organization? Like see why it is better
- 29:44described because at the end you know it's reinforcement learning
- 29:47you learn and you know that the answer that gives You know, the
- 29:51best results are the ones that the win, right?
- 29:55So there is a sort of a, you know, a selection mechanism of
- 30:00the answer. So if the answers that are
- 30:03awarded, are those that are going, you know, on a tangent.
- 30:06Probably, you can get in this scenario.
- 30:08This is at the end of the day, it's a statistical tool, right?
- 30:11So I mean, we call it in artificial intelligence,
- 30:14whatever, but effectively, it's pathetic, supplication of
- 30:17Statistics, right? To a very complex problem.
- 30:21Violent. But distinct statistic.
- 30:22And so there are up absolutely situations where statistics gets
- 30:26you into the outlier. Right.
- 30:28Absolutely. So that's possible.
- 30:30Even I don't know, I'm not a mathematician but I'm sure that
- 30:32a lot of wasted that this can be described mathematically in
- 30:37terms of what needs to happen for the Enterprise.
- 30:39As I said earlier, right? We should that should not be
- 30:42allowed, right? Because you don't want to have
- 30:45you know, a tool speculating on what tax legislation might be.
- 30:49If I look at our own business All right, so we need to give
- 30:52specific answers to clients. That's why this kind of a
- 30:55curated approach to it, you know, for the time being is
- 30:59absolutely critical. So will be, it's an investment
- 31:02that every company needs to make both from the bias, the ethical
- 31:05side, but also, for the content of the, of the answers and the
- 31:09questions, Etc. So there has to be some sort of
- 31:12the labeling that happens on the tool.
- 31:15When you give an answer in your train, it right to make sure
- 31:17that this is the right, that the answer that is proper, right?
- 31:21Can I ask to what extent are you thinking through these kinds of
- 31:26processes at a why we are experimenting?
- 31:29So we're trying to understand, you know, how much work needs to
- 31:33be done. Etc.
- 31:34So we are in the process of defying, all of that.
- 31:38I think that's why, you know, the previous questions was, you
- 31:41know, are we gonna replace people?
- 31:44I don't say that. I think, you know, the more we
- 31:46use this to us, the more we need to add it up.
- 31:48Human expertise to it. So it's going to be The opposite
- 31:51but I think if we put at least for the next you know few years
- 31:55until these tools mature and there is more knowledge about
- 31:58it, you know there is definitely the need for for investing in
- 32:03the Enterprise in human capabilities around that Chris
- 32:06Peterson on Twitter comes back. And he's asking if you have any
- 32:11thoughts on cybersecurity issues, created a need for side
- 32:15for the Cyber insurance industry and do such firms and These have
- 32:21a role in shaping and governing the use of AI in the Enterprise
- 32:26regarding liability. I think in the future probably.
- 32:30Yes, I don't again, I don't see any Enterprise function,
- 32:36completely delegated to Ai. And so if it's always a sort of
- 32:41a center model where there is the human and the Machine
- 32:45together doing something on behalf of the Enterprise, I
- 32:49think fundamentally it's the same level of liability or
- 32:52responsibility that you have today in the future, okay?
- 32:57So for example, let's say, you know, credit decision or other
- 33:03type of Thanksgiving without going into the most dramatic,
- 33:06you know, healthcare related decisions.
- 33:10I say, I don't see why not. That could be some level of
- 33:14assurance but I have it in my understanding so far.
- 33:17Okay and it's for sure, not complete.
- 33:21I haven't seen anybody yet. That is planning to completely
- 33:25you know run the piece of a business like that.
- 33:28You know in a schematic fashion what are some of the challenges
- 33:31that you see to adoption of these kinds of tools in the
- 33:35Enterprise one is over hope. And then, you know, for most of
- 33:39my career, AI starting from school, basically, and I've seen
- 33:44these herbs and flows of excitement, right?
- 33:47So I started, I started really 90s. 7 and he was like, you
- 33:51know, the winter of the I nothing really happened for
- 33:53years, right? Even if there was a lot of
- 33:55potential in those days, it was all about expert systems.
- 33:59You know, that kind of knowledge because we didn't really have
- 34:01the technology in terms of the processing power.
- 34:04You know, the memory and all those kind of things then he
- 34:08started getting into, you know, some games, you know they were
- 34:11win like you know the Chess World Championships, blah blah
- 34:14blah. So I'm not going to make the
- 34:15whole history but it was but you know every single time that
- 34:18there was a major breakthrough like One the big blue that I DM
- 34:22DT 90s you know beating Castro it was a huge you know you know
- 34:26the interest and everything then it went down again right from
- 34:30few years then deep learning came up and we started dealing
- 34:34much more with you know image recognition which I think is a
- 34:38great application of artificial intelligence.
- 34:40So then another big you know Spike of interest that actually
- 34:46is leading to this new language models right there are
- 34:48similarities and so Sometimes, you know, we tend to again to
- 34:53humanize these Technologies to the point that we the our
- 34:57expectations of what they can do becomes the same level of
- 35:01expectation. We will have from a human but
- 35:03it's not that. So I think that if I was an
- 35:06Enterprise outside of mine because I will spend time trying
- 35:11to demystify in a way what these things can do.
- 35:15But at the same time, showing the value that they can provide
- 35:19because there is a The value of their it's just unfortunately
- 35:24what happens is that people get overhyped and then they get all
- 35:27will be disappointed and so you know in to me in a year from now
- 35:32and we're going to say the tragedy was a great failure.
- 35:34Even if is actually an incredible Innovation, I don't
- 35:37think so, right? But it could happen, it could
- 35:40happen because people, you know, think that these things can do
- 35:44things that are, you know, comparable to what a human being
- 35:47can do. They can't right now.
- 35:49What It is vice. Do you have for folks in the
- 35:52Enterprise, who are Business, Leaders, and Technology leaders,
- 35:56who are looking at these kinds of Technologies and trying to
- 36:03figure out what to do the bottom line is that this is going to go
- 36:06really fast again, right? I was that saying, in the
- 36:08beginning, this is going to, it's an arms race between super
- 36:14powerful companies and they have a lot of money to invest, right?
- 36:18Then you start getting into to, you know, government except so,
- 36:22I think this is going to be exponential.
- 36:23The speed of these Innovation will be exponential.
- 36:27I think we're just saying the beginning of it and so I think
- 36:33that two things to think about so one is There is no right now,
- 36:40you need to be knowing what it is and being able to understand
- 36:45how to apply to your business. So that is no postponing it a
- 36:49say, oh, you know, I'm not going to be an early adopter.
- 36:52I'm going to be very fast. Follower very first, follower is
- 36:56Superfast, follow it very fast. Photo could be six months.
- 36:59So I think for large Enterprises that can afford to do that.
- 37:03I think that is the need of experimenting.
- 37:06Hiring the right Talent or training the right people to get
- 37:10to a good handle on where this industry is going, right?
- 37:13This is its you have to be keeping an eye on this.
- 37:17That's, that's the bottom line. And then the other thing is
- 37:19that, I think, oh, lots of questions.
- 37:21Very, very good questions about the ethical implications, the
- 37:24bias and all of that. It's really important because it
- 37:27has a massive brand impact, right?
- 37:30So there is also a risk management side of around these
- 37:34technologies that has to be understood very well sir.
- 37:36If I, you know, I would be a CEO of an organization.
- 37:42I would push my technology team to have the right skill sets to
- 37:45understand what it is and where it is going.
- 37:47And at the same time, I would put in place in a sound risk
- 37:50management, thinking around it as well to what extent do you
- 37:54think that the developers of these tools and the folks were
- 37:59gathering the data sets, need to have that ethical understanding.
- 38:04And the reason I ask is because Historically technology was kind
- 38:09of separate from the ethical, or the application of that
- 38:12technology. From an ethical perspective is
- 38:14absolutely, do you think room, right?
- 38:16And you know, think about what's happening, you know we social
- 38:18media right? So it's how important it is the
- 38:21content moderation in social media.
- 38:24So you know, this is not exactly the same of course.
- 38:26But you know if you translate it into this space in there is
- 38:30opportunity for abusing these tools everywhere, right?
- 38:34And they as I said, the brand impact, It is massive.
- 38:38So there is the need of, you know, setting guidelines.
- 38:41So for example, internally that is why we have specific
- 38:44guidelines on how to use the tool what we can do and not do
- 38:46with it, right for now maybe they're a little bit restricted
- 38:50but we want to be on the on the safe side of things and then
- 38:53we're going to open them up a little bit more as time goes by.
- 38:57But you know, there has to be, but fundament.
- 39:00Everything starts with the true understanding of what is
- 39:03Available impossible, right? And if you don't understand that
- 39:07in detail, it's difficult to put, you know, the boundaries.
- 39:10And so today we're really at this phase where we're
- 39:13understanding that. What's reasonable, what's
- 39:16practical? What can we do?
- 39:18Yes, and this shouldn't be underestimated in terms of the
- 39:22ability of driving Innovation, but it shouldn't be either,
- 39:26overestimated intense of being a sentient being or anything like
- 39:30that because that, you know, I think it's worth far.
- 39:33From that, you know, I have to say the interesting thing about
- 39:36this question to me, is the fact that people have this
- 39:40conversation because at times these chats million human
- 39:46conversations, so what? Yeah, it's incredible.
- 39:49I think what the heck done? It's absolutely one of the, in
- 39:52my mind, one of the most incredible inventions of the
- 39:56last definitely. The century, right?
- 39:58And there is the being the iPhone 5 in mobile phone, social
- 40:02media, all that kind of stuff. Great.
- 40:04I think is super important but this one, it goes towards
- 40:07another level of innovation because it impacts our, you
- 40:11know, it's really like, you know, it's almost like looks
- 40:16like reasoning right over human, right?
- 40:18So it's absolutely incredible. So I'm, you know, I'm super
- 40:22excited about it. And why don't we finish up with
- 40:25another question from our Salon? Khan who says on Twitter, what
- 40:29do you hope future iterations of a I can do.
- 40:33I think the future of Aviation, I think it has to be more
- 40:36transparent. Right?
- 40:37So that you guys have asked fantastic questions about by us,
- 40:42ethics and all of that. So that is a problem, right?
- 40:44So if we go deep this, the count, this technology are
- 40:46powerful. If they go in the wrong hands,
- 40:48in terms of, you know, being able to generate content as
- 40:53chests Behavior. So public, you know,
- 40:56understanding of things, it's really dangerous because we're
- 40:59basically adding another layer of intermediation between And
- 41:04the sources of knowledge, right? So if that is done with, not the
- 41:09right ethical, I guess, you know, it can influence politics,
- 41:12it can influence the election. We can do a lot of different
- 41:15things that we don't want, so I think that transparency and
- 41:18ability to monitor what it does. And often strained, I think is
- 41:22really important, but then I cannot wait for the future
- 41:26because you know, like there is so much more in think about the
- 41:29breakthroughs that we can have, when When the sum of the human
- 41:35knowledge is so accessible, right?
- 41:38And in such a smart way. I think, you know, if we take
- 41:42here so I saw something that you know Monday we're gonna you
- 41:45know, really Implement nuclear fusion, that's not the thing.
- 41:48But in terms of stimulating a different level of thinking and
- 41:52summarized in this knowledge, I mean it's incredible.
- 41:55I'm so excited about it. We have one last question that
- 41:59snuck in under the wire and so how about if we finish out with
- 42:02something from Melvin aguiar who says, how comparable is our
- 42:08products like chat GPT to low code?
- 42:12No code. Will they coexist or eliminate
- 42:16one? Eliminate the other.
- 42:19And I think he's referring to the fact that these tools are
- 42:23really optimized to help write code.
- 42:25Yes. I think right now, it's not that
- 42:28big yet, but at definitely that is Direction.
- 42:31So you would ask, you know how much code would Do you have to
- 42:33write in in five years? When you can actually talk to a
- 42:38tool and say, can you build me this application with these
- 42:42characteristics? These inputs outputs data, Etc.
- 42:46So I think the direction, I think the what I say like you
- 42:50know the writing is on the wall. In terms of, you know, software
- 42:53engineering will have to be changed completely.
- 42:56It doesn't mean that they're not going to be, there's not going
- 42:59to be software Engineers. Absolutely not.
- 43:01But I think the role of the software You would have to
- 43:03change a little bit more and get closer to the business because
- 43:06you still need a human. They can summarize.
- 43:09Those needs and requirements into and ask the machine.
- 43:12What do you need? You know, armies of people to
- 43:15code you know routines and different lengths, programming
- 43:18languages, maybe not. And with that unfortunately
- 43:22we're out of time. It's been a fascinating
- 43:25conversation. I want to say a huge thank you
- 43:28to Nicola Marini beings. You know Nicola thank you for
- 43:32being here with us. Us today.
- 43:33Thank you for having me Michael and everybody.
- 43:36Thank you for watching. Especially those folks who asked
- 43:39such great questions. I always encourage you to watch
- 43:42and ask questions. You guys are an amazing
- 43:44audience, but before you go, please subscribe to our YouTube
- 43:49channel. Hit the Subscribe button at the
- 43:51top of 60 talk.com, our website, and subscribe to our newsletter
- 43:58and tune in for our live shows and we will see you again next
- 44:01time. Have a great day everybody.