S3: E3:Who owns your intelligence?

Welcome to the ALSO AI podcast! Discover the latest advancements in artificial intelligence through insightful discussions with industry experts. Each episode will bring a fresh perspective on AI and addresses their real-world challenges, opportunities & interests. you as an IT reseller might have or your customers could ask about.

Featuring experts from ALSO and leading Vandors, we explore diverse insights on AI. Welcome back to season three of the ALSO AI podcast! In episode one we discussed what AI might cost us. in Episode two We asked who benefits from it? And Who gets left behind. today I want to ask something a little bit more uncomfortable.

if businesses every day upload supposed confidential documents, whether it's a pricing strategy an unreleased website or internal knowledge customer conversations. AI doesn't just analyze that information It creates it recommends with it and sometimes even acts on its. so the big question is who owns your intelligence?

To tackle this massive question I am joined today by many exciting people from HP and ASO. Welcome all! Thank you, thank you. so my name is Praveen Matthew And I'm a Drutron HP responsible for building our AI ISV ecosystem to help deliver an end-to-end experience with our devices as well As the software or customers And I'm Jonathan Goes down from HVM, the director of technical marketing and enablement for this security GPU.

About about twenty years of security background thanks. I am Alonso from ALTZU. Uh...I am a group sales lead on cloud and AI. Super excited to be here and participate in these talks around AI and possibilities uh..around it in future. So Federico Ascoli here, excited as well to be back at the post office.

I'm responsible for our platforms here in Helsinki so very excited to engage this conversation with all of you! Thank you very much for the introduction. we're very pleased. Maybe to break the ice a little bit, I'll just ask a few questions. Would you upload this or not? HP begin to answer and then Alzo can continue Next year's pricing strategy.

would you upload it into an AI tool yes or no? No i mean if its completely public tool than...no I like that nuance. Thank You Jonathan Totally agree. Yeah, also if it's public told for sure better not. Okay so that's a no consensus there. how about this? the transcript of this podcast before its goes live?

That's me in my mind! If it saves me money yeah I would certainly do that. Come on is important content? i wouldn't. I think it goes to that extent. If its important content, you need be careful where your uploaded right? I agree with that as well. This is perfect what you're saying Daniel. and yes i would like challenge a little bit because organizations today many colleagues employees upload documents of this nature every day into AI tools might not necessarily aware the consequences or actually doing so.

The first question I would like to ask you and to discuss, where does it all go? All this data. You know i think the short answer Morgan has is goes what our company are using to get results out of? that's a short an easy answer right. there's more nuance too from a consumer standpoint for example if your asking anything in chat gpt or any other ai tools is taken up by them in whatever form to be able to process and train.

And what not, right? Now as it comes to customers I think there's a bit more of an nuance. you know i think You can broadly think off. It has two aspects. There's the training aspect In this pumping aspect. As far as training goes Typically these days you could work with many Of the AI companies To say hey...you're Not supposed use my data for Training.

And that are ways For Customers to Be Able to Secure That. Put That On Prem or on Local Devices. So there are ways to get around that prompting is a little more tricky. I think that's what nuanced. But yeah, I think they're still In an area where we tried to figure out exactly how do we control all the data?

Where it's training or prompting to kind of stay within environment. Interesting interesting thank you Praveen. Yeah agree i mean i think that businesses in general going back to the initial question i think They need To expand The questions for how their data is actually being processed, right? So what data am I really providing to these tools.

How it's going be processed and will the provider do with that? who can ultimately access those information? so there are a few questions. probably businesses should ask themselves more when it comes understanding where they did actually goes The thing. one of things this challenge The use of AI is so broad that any kind of general statement about it, Is challenging.

So to me I think and its a good discussion for the team... ...I kinda split into two broad categories. One would be i'm standing up an AI application To do certain thing Like chatbot on my website To take care of customers or burn down tickets Or something like that And build a whole pipeline behind them.

The other one is, I think what we were talking before. Is i'm just a task worker and I want to do my job as easy as possible. so I shove things into some AI tool on the machine whether it's co-pilot or anthropic or Chachi BT whatever And get some answers. then go to races which are very different sort of use case.

So would cut that two up. if you're building your application for your own businesses. That'a whole difference. scene right where all kinds defense in depth, you can choose to use an open weight model. You can choose pay-to take a frontier model and stick that inside your AWS or Azure instance. I mean it's whole different thing right versus what your task workers are doing trying put guardrails on the input side of what they do.

so i just think you need make very hard consideration because its too broad. if say hey hi Yeah, it's a big distinction. Maybe I should be more precise with the vocabulary i'm using. Interesting word guard rails Daniel. any reactions? I think that distinction is super interesting because you need data to structure your business workflows for sure To tell AI agents know what do we got and that data for sure is data super important For Your Business Company.

You Know maybe data is something that you want to keep on for you. And then the other one, it's the prompting one where all of my frontier labs or all the front end interfaces are going be analyzing what type of user we have and needs as well when searching. so this part will enable these frontier labs to understand better customers.

And from other side, like if you want to build up those workflows. Yeah for sure You need to understand your work flows and you need to provide the data For the agents To be able to build these Efficient workflow. so on that scenario for sure? We need to consider what data you share. How do you share it in one type of scenario If its hybrid cloud or premise?

there are many things to consider On that area. well I agree with Jonathan. I guess the question would be... The obvious thing, and i think it was seen in what we were talking about is if you have information that's going to get you into big trouble. If it leaks out then You don't want to give it a public model right?

Like your are in Europe And you're subject to GDPR and upload a bunch of personal information. Once they do that They'll keep their document but train on data And maybe they keep your document for thirty days or whatever. It's kind of wide open, but if not I think it can be really challenging to stop people from using it except that use case.

That is my problem because it saves so much time and people are generally lazy that its losing battle. So you have risk rate like anything in security problem you're trying to solve and start by going after the most risky ones, either through the reputation or financial situation of a company. Or because of compliance?

You said it yourself Jonathan, risk maybe to Ravine and Federico. now why in your view does all this matter do think users actually understand what's going on behind-the-scenes where their risks are? if so they being informed? So I think that with many different technologies, it's not so necessary. the end-user of that technology application needs to know every detail on what is behind.

When you're driving a car and don't know anything about how your car keeps you secure but at the same time the responsibility for using this is new if anybody is hitting something. So I think there should be a system of governance behind that makes sure is that, uh... There are some guard risks as you guys were mentioning before.

That make sure and users feel safe into using the specific tool which will never probably cover one hundred percent of risk. otherwise the cost of protecting from this risk going to huge. but it's kind balancing game right? so how much risk am i willing take and how much am I enabling my end-users to know what they can do, what they shouldn't do.

And allowing that risk to be there? so... To finish it off i don't think users need to know every detail about this. They have a strong governance system behind But still trust it though Exactly Praveen any reactions? I think yeah you are still trusted many of them not completely cognizant You know, because it's hard for me as a user to discern what is going to be sucked up or what will stay local.

So there are some use and awareness but this also has to be some guardrails that you put across the enterprise of an organization in order to avoid users getting into these situations. It's such a complicated area, and it is so dynamic. And organizations frankly don't have the resources to even understand what's going on... ...and changes week by week.

I think that most important thing to do is establish an AI governance process…. …an awareness right upto top of the organization.. ..and set policies consistent with goals! Don't do anything until you've done that but do it now. There is an advantage, in fact of setting these guardrails pretty much early on.

In your view guys it seems quite obvious that technology providers, distributors and IT players have a immense role to play in helping make decisions. Any thoughts about this Daniel? Yeah we played big roles here. first of all testing the technology understanding kind of suggest to the partner, our channel how to tackle it.

At the beginning for sure we need to try out and do some internal research. that's super important now. That role that we play there is critical from companies partners and customers. For sure every end customer will try different type technologies but they're gonna rely on trusted partners, trusted vendors guardrails and reliable technology that they can build on top of.

So yeah, I think it's a huge role like this panel both companies are playing right now. absolutely thank you Daniel. maybe if i can build in that i totally agree that trust is the key thing here. so there's industries so complex and uh the governance of speed growth which is changing. I mean, it's very difficult for you know partners around the world to really have a grasp on every and once of that.

So its important too kind of Have yourself some trusted partners within the channel. so make sure That You develop and deploy securely your AI Deployments. i think what partners need To do is they Need to decide What their business strategy Is? And i Think ai Opens up A lot Of opportunity. i remember i was in event In eastern europe About three months ago in front of, I think about a hundred people representing partners of HP and i asked them how many have aggressively started to actually develop AI applications on behalf of their customers.

And there was only one out of those hundreds... You're pre-empting one of my questions. great! No yeah okay so sorry if I am but no it's fine please. But the key point was they were doing some really innovative stuff. I won't like monopolize The conversation with details, but there are very specific examples of customer had a problem and because partner invested in AI They're able to actually build out solution which is not technical for people who do rock climbing or whatever And so completely changing value given to their customers.

Nobody else was right. So you got these guys and he got everybody else. And so that I think is both the challenge in The opportunity for partners going forward, and it's becoming increasingly important to have these discussions early on. It reminds me a little bit about responsibility when we prompt AI?

And it generates something that could win a million dollar deal. who ultimately Is responsible? The company behind the learning model, is it a person who makes that decision? It comes back to car example. I think Federico was explaining but what do you think? I'm the responsible party, i can have data processors behind me processing data.

It doesn't matter what they're doing and if using AI or not it doesn't change anything. The reality is that im still responsible. They have a set of responsibilities. however ultimately its me And that framework doesnt' change with AI. Yeah! Its interesting. because of this reason some these decisions can be extremely tricky right?

So in recent interview we did our customers. One other thing they mentioned was. there are couple of areas where they do NOT want AI to be, they are reluctant to use AI which is basically anything that involves for example money transactions right? Anything that involves people decision let's say hiring somebody or anything that involved or affects the health and life of a person.

So those three areas we heard from our customers if you wouldn't use AI precisely because when it gets to the crux of this question then you don't want AI influencing giving bad decisions. But at the end, I'm just giving you a little bit of different angle. At the end there relying on more precise tools even though RAEI right now and for example in healthcare are empowering these doctors to decide on a better way.

so uh... You have this contradiction. they're like ...you don't want AEI make their decision but you rely heavily on this type of tool because nowadays is so precise That you're not going to go by yourself. You know, you need these tool sets To make a decision properly. I think there is a combination There but for sure i stand forward the the responsibilities on The human.

you know decision-making. Yeah, I think that's down at the nail on ahead? I mean I say a couple of things and give some health care examples. what they do. You know how they're going to do. What they're doing right now Is it'll insert A human in the loop. so let's Say like I don't know looking At x-rays to detect a disease or something Like that Right.

And So AI now is just as good as a human most of the time, at detection or this sort of thing. So what they'll do is so that they run it through AI. however That won't be the diagnosis. there will have to be a radiologist Or whatever the term is looking at and saying yep I'm okay with that or nope i'm Okay right?

And so that's the way though They'll frankly like keep the human in The loop provide some sort of a sort of protection or you know Like Just like You'd Have A. Take a second opinion right if I was diagnosed with cancer, i might get. A second opinion from another doctor now it's basically like will be valid by the human.

thank you jonathan. and ai seems to be evolving a lot more quicker than all the regulatory policies in regulation behind that. how do u think from a vendor technology provider standpoint and from a distributor businesses should responsibly operate in that? I want to say chaos but not chaos. Navigate, how do businesses navigate all of this?".

You know the recent incidents more about open AI agents going rogue it's kind. there is no better example than that to show actually a big gap in terms of how fast this was developing versus so much we are able keep up right and your question especially technology as an industry has a responsibility be more responsible and to allow for regulations and guardrails, I would say in the past historically you're not being very good at it.

It's been always like hey let us break a bunch of things then see what happens. I think at least we should take the lesson and learn this time. This is actually a far more disruptive technology than any of its predecessors, right? We should bring that into picture as well. so you absolutely have responsibility to be able to slow down and be more accountable here.

changing at a speed that it's never been like in the past for other technologies. However, that gap is natural. so we need and our channel but all companies to find their way to limit risks out of these gaps. And this not only about what we talked about before kind putting guardrails and preventing But its also trying to having ways to understand if something went wrong.

So I think another important point Exactly or anticipate. So I think as businesses, we should be also able to kind of create audit trades around the yeah and being analyzing. To make sure that we know what's going on. We can always trace back on What went wrong? And learn from that. so audit trails you're saying.

there was a very interesting story actually That i heard From an actual organization like A big company who is doing this. and so This Is some The Audit people this Like the director Of the Audit person. They're already using AI to look at the artifacts, produce a set of findings. So some findings come up and the humans looked one finding in particular And they said oh this is actually serious finding.

This needs to be escalated. change our control activities to mitigate the finding, but AI said no. The risk is not that large for business. you don't have do anything. and so humans spent a week arguing with AI. try convince AI. this was actually real risks. of course they were stacking by trying send data at AI to convince them their position usual confirmation bias And they actually failed.

So by the end of the week, They gave up and accepted The recommendation Of the AI. I thought that was really interesting because That's just a taste of things to come Because you've got people now who are going through that process A year or two from Now... They might argue with it for about an hour.

If that! Then they're gonna accept It and move on. We talked About ownership responsibility Trust & control which means it's now time to see how much we actually agree on with the myth or reality rounds. I will ask you all two, three questions and answer by myth or realty in providing a short ten-fifteen.

second explanation are you guys ready? Ready! Super. So am i... Myth Or Reality and I won't accept it depends. So if company data never leaves the device, its safe myth or reality? let's start with our lovely colleagues from HP. It is a myth. I can say it's a myth as well. location helps but not full story around security.

you know what i mean? You get hacked then they get in tight Like the transcript of this podcast, right? Very valuable transcript. We'll see. we'll see. Myth or reality. business should be able to explain every important decision made with AI that start without so myth or reality. I would say a reality.

Important decisions need to be owned by people and organizations. Every hundred percent with erico. That's the way to go. It's a reality as well. I would just add another dimension to it, which is like there always has to be of course as the company said companies have to own but also has been audit trail that anybody can come in an audit and Be able to show what they did.

Okay? I'm gonna be out the outlier. i think its a myth. And my counter argument with being yeah. so Like I was reading A book like a cold war book nonfiction right where There's these double H & spies on all this. And one person thinks this. other person is a double agent, but she can't explain why. She just has like feeling.

so humans cannot explain A lot of their decisions right? AI doesn't have gut feelings Yeah So and the lot of solutions don't either? So for example if I use a product to do Prioritization of alerts that come into my sim and i have a list Right nobody goes in there and asks The vendor of the SIM explain why this alert is above that alert.

I mean, it just doesn't happen like in theory. you'd think you care but very quickly You become beholden to the tool and i predict That what will happen? Is unless It's something incredibly serious This idea of Like i have To audit? it sounds great until you get to The practical reality Of it And then i bet it won't Happen Very much.

Maybe this next myth or Reality question is Particularly Directed to you jonathan. ai can make a fairer decision than a human myth or reality. Yeah, well knowing me and my poor decision-making could certainly make up more fair decisions that I will so... And let's leave it this way like in San Francisco.

here if you're riding your bicycle in San Fransisco which i sometimes do when my partner drags me into it much rather be surrounded by the self driving cars we have That's for sure, because I trust their decision-making on the road. So yeah... It is a very fair point! Well i'm going to go in opposite direction Morgan with Jonathan who says it's more of a myth only reason I said mythos.

there are situations where true like healthcare examples that we mentioned were AI can make better decisions but There Are so many cases today where they hallucinate right? Maybe at some point later it will, right? But till then I'll take all of them. Okay Thank you Paravin Daniel and Federico miss the reality about fair decisions with AI.

i would say It's a reality but that not automatic. And especially when AI doesn't have That much of a context of real world where probably things are not always perfect Then the best decision might not be so easy for AI. So its a reality ,but is not so automatic. I know, i shouldn't say it depends but its really depends on the context.

You said that it depends? Yeah! But it depends for sure because if we take Jonathan's example... The car driving itself and has some sensors they can sense audio better in a way than their driver. They will make decisions before the driver can take them. So that would be an example of more context information for AI to make a decision, but in other cases for sure human decisions are the ones who made this decision right?

Absolutely and we can trust our guts! Can't we?! Well I think it's interesting thing if you want to run this one down. so If You Play This While It is very interesting discussion For Sure Is Let's Say The Paradigm The organization always needs to be able to explain the decision. Okay, so you try that with humans right.

next time your boss makes a decision and You asked them how exactly did you arrive at? And see what happens I'll leave it at. he won't be able To produce an audit trail. It's all right. Yeah Maybe one final one. pick Your brains and this One might sound A little absolute Myth or reality, the biggest barrier to trustworthy AI isn't technology.

It's us. Everybody is thinking pretty deeply about this one I know! You said you don't accept depends. so then i'll pick a reality because it's more closer to reality. Because ultimately yes as we thought be able trust and accept And in some cases at least right? AI makes better decisions than we do.

Yeah, I think a lot of it is us. I mean AI's also fed by data information and guardrails in governance as you mentioned that we feed it so We are also responsible to make AI that is trustable. Feed it fruit and vegetables right? Yeah, I mean to trust. It's like in a relationship right where you build trust with the other person over time You don't just show up and trust each other because I sure don't anyway.

So giving an example again from healthcare but real-world example at scale which was large hospital is trying To take the process of calling humans And getting them to show for their tests Right? To come in and The nurses were doing that it was a complete pain in the butt because people don't answer the phone, it goes to voicemail and they don't call you back.

And then when I called them make excuses is very time-consuming right? So idea was have AI calls for people get him show up that nurses were skeptical not like afraid of their job so much although a big problem in healthcare cause hated doing this job. just didn't think that AI was smart enough that could trust to work with the humans, and get them in.

The solution that the organization came up with I thought was very clever. they created basically a test instance. so what they were doing is training AI on real data which are stuff from human. That had two big benefits. first one of course it improved model because the model now listens all these garbage people can throw at.

but your question from the nurse's perspective, because they saw how the model got better dealing with the stuff that they had to deal with. And that gave them the trust they needed to now move this new way of doing things and then they were able use their time for something better.". That is a fantastic example!

I think it is perfect. closing on our podcast. This concludes season three episode three. Thank you so much for joining us, Praveen Jonathan Federico and Daniel. It's been an absolute pleasure on my behalf to engage in these discussions with you. Have fun! Could keep going? Thanks very much. Maybe if he would like to leave our audience With one key takeaway from this discussion.

We'll start with the HPN then move onto Alzo. I will go first Then. So i think The Key Thing To Keep In Mind Is Just a couple of things. One is be very clear on your goals, have an AI governance policy in place and put dedicated resources on it long term because its so dynamic And there's so much pressure especially on IT organizations that if you don't have dedicated people on it they'll just move back into having to deal with operational firefighting.

Fantastic takeaway, thank you Jonathan. I would add Morgan saying that like one of the key things is there's obviously still issues or concerns about trust but as Jonathan mentioned i think it's important we move fast and we don't lose the benefits of amazing technology thats coming while at the same time taking the time to be able do in a responsible way.

Amazing. Thank You Praveen! resonates with me. the most around our conversation is trust, right? We mentioned that a lot and that's between humans and AI such as the nurses example from Jonathan. And the trust on the security around AI but also the trust within channel partners to help out navigating this complex ecosystem that AI is generating and changing so fast.

I think that's a key word for everyone, and it's conversations like this. And its best practice is sharing understanding what we want to achieve... ...and the tools we need to achieve those goals are quite important. so yes! That was really nice to participate in on this episode as well. with these interesting conversation my key takeaway Is your trusted vendors go ahead explore deploying these type of technologies and address your different work streams that can be implemented with AEI.

I think it's time to move ahead, relying on trusted partners... ...to understand how to execute and deploy this technology in their businesses. That will be my key takeaway after all the angles we have discussed. Thank you again to all of our guest speakers today for joining Season Three, Episode three.

Stay tuned with us for season four. maybe we'll engage in further discussions around AI. thank The conversation continues beyond this episode. Share your thoughts and ask a question! For more resources, links & product details discussed today don't forget to check out our show notes and visit our podcast website.

S3: E3:Who owns your intelligence?

Kapitelklicken springt direkt hin

Shownotes

Transkriptvom Podcast bereitgestellt · Zeitmarken springen hin