Up next in 10
The Cloud Show with Magnus Mårtensson ft. Sakari Nahi - Ep. 10
3K views
Oct 30, 2023
Join this live session with Magnus Mårtensson ft. Sakari Nahi for the next episode of The Cloud Show with Magnus Mårtensson on August 23, at 01:05 PM (EST). The show is about cloud leadership and all the important questions relating to cloud projects. Certainly, many matters when a company is going to and want to be successful in the cloud, are about technology. However, there are many additional matters, adjacent to technology, that we also need to tend to regarding business strategy, human resources, organizational change, planning for a technical cloud approach, and many more questions. These conversations are critical for a healthy cloud and for a swift and accurate cloud approach. GUEST SPEAKER Sakari is responsible for the vision, brand, and quality of Zure. He is a Microsoft Regional Director and an Azure MVP. He has over 20 years of experience of creating custom software and 10 years of experience with Azure. He is the founder of Finland Azure User Group (FAUG) and a co-host at the Finnish podcast Ikkunastudio. # Corner - Community of Software and Data Developers https://www.c-sharpcorner.com #CSharpCorner #TheCloudShow #CSharpTV
View Video Transcript
0:00
Hello everyone, hello friends. Welcome back to the Cloud Show
0:06
So currently it seems like everything has to have AI. It's like whatever application you had
0:11
and sprinkle some AI on top of it. AI is everywhere, AI is everything, that's what we hear
0:17
and that's what we are going to discuss today on the Cloud Show
0:30
Hello, Sakari, my friend. How are you? Hello, Magnus. It's lovely to be here. I'm very good. How are you
0:39
I'm good. I'm very good. Thank you. So who the heck are you and why are you running this wonderful company called Zür
0:46
Oh, hey everybody. Nice to see you. I'm Sakharic from Finland and I'm a CEO of Jure
0:54
This is a company we started 2011 with a friend of mine and I'm a devil
1:00
by trade, been coding dot net since 2001. And we've been very lucky to be in this boat of Azure
1:07
that's been growing all the time. Right now, Jure is in Finland and Denmark and Belgium
1:14
And what we do is we design, build and maintain digital products and solutions on Azure
1:20
That's basically it. Fantastic. That's it. And then you started this company in Finland, you are from Finland, you are Finnish
1:29
That is, that is exactly. Do you are correct? I am correct
1:34
Good. All right. So let's get down to the topic of today
1:38
AI. AI is everything. AI is all the things, as we have heard from all the wonderful announcements now
1:45
The first question, honestly, becomes, because I wouldn't say I'm a skeptic, but when somebody
1:51
says, oh, this is going to be the next rage, I go like, okay, so why is this thing so hot
1:56
now all of a sudden? What happened? Yeah, that's a good question. I think many of us who've been in the business for a longer period of time in the IT, they kind of hear of this new silver bullet. And we kind of get a little bit skeptical, you know, is it something that's really true? I think it started with chat GPT back in November last year. And it became the fastest growing service in the sense that it got like a million users five days or something. I think Instagram was the next one at 75 days or something. You lost
2:28
also in the graph. I think the hotness is about the natural language processing. It's about the fact
2:36
that you can talk with the AI like a human being. You get a feeling that it's listening. My son came
2:42
home from a football match right now. He might parched in. Sorry about that audience. But there's this
2:48
new kind of feeling. I was listening during my summer holiday, just a month back or something
2:54
I was listening to this podcast and it had the guest of You Well Know a Harari
2:59
this Israeli philosopher historian. And he was talking about AI. And it's like, why is this philosopher historian talking about AI
3:07
And he said that AI is the first tool in history that makes decisions by itself
3:15
So he was comparing it to a shovel and a knife and a fire, things that empower people into Microsoft
3:22
But AI is the first thing that he thinks could be making decisions for us
3:29
And he sees that as a potential disruption in human society, good and bad
3:36
And that's kind of a wake-up call in the sense that a historian starts talking about this thing that to us might feel only technical when it's not
3:46
It is not, absolutely. So open AI, chat GDP, we consider those now extremely disruptive technologies
3:55
And is this because they are, because you can reason with them or, you know
4:03
and you can follow up questions with them because a search on the internet was a shallow thing, right
4:08
We search in a little box and we get a result. And we have to figure out what to search for next
4:13
This thing actually suggests the next step for us, right? Yeah. I think the NLP natural language processing, my take is that that's a big thing
4:22
And people like Harari, who are not IT, people seem to be of a mind that this, when computer cracked the language, which is, of course
4:34
how we create contracts between people, how society is built on language. That is a meaningful thing
4:41
The other thing I think is that GPT that came from OpenAI, the GPT4, the latest one that came out in
4:50
March is also a so foundation model meaning that it generally usable by all companies in the world in the sense that if you want to build something on GPD or OpenAI traditionally custom development or building a tool would require a lot of work
5:11
Now, GBT being a foundation model means that it's generally usable and useful instantly
5:20
And that's something that's pretty amazing. I don't know, you probably run into the same situation, Magnus, you speak with your customers and companies
5:33
and they say that, hey, this GPT, we need to get to understand this because our board of directors or senior leadership team told us that as a company
5:44
we need to get to the bottom of this technology. And during my career, I can't remember many technologies where the impetus for learning about
5:54
new technology has come from the upper management. It's kind of a rare phenomenon
6:01
Right. So effectively, you're saying, it sounds like you're saying that the market is
6:06
is all of it, the market for AI. As in everyone is now looking and kind of expecting
6:15
whatever we had before, now plus AI. So how do we put AI on our thing? It's like
6:22
icing on the kick. Yeah, and Microsoft is, you know, in our market
6:28
we work in the Microsoft bubble or sphere or community. Microsoft is crazy about
6:34
this and, you know, they're coming up with many, many different co-pilots
6:40
and they invested over $10 billion in open AI company and that's visible in the fact that
6:48
they're the only cloud provider that can provide GPD through Azure Open AI service from the cloud in a secure and governable manner
6:58
And that's one sign. Of course, Microsoft could be wrong, but I don't think they would make those kind of investments
7:05
if they wouldn't be receiving a favorable market indication. Yeah, yeah, I agree
7:12
And that's where it kind of comes back to that thing where you want to think about this
7:18
and maybe be still a little cautious. I'm sure AI is going to be great
7:24
And how great is, we haven't seen that thing yet, how great it's going to be
7:30
Maybe it's going to be entirely disruptive for everything we do. Yeah, there's, it is, it is interesting
7:37
So with the AI, there are risks that aren't easily mitigated. And companies should be careful with it
7:48
Because the large language models have been trained by the data in Internet
7:55
and they've been trained by humans in reinforcement training. It means that these models have biases
8:03
Exactly. Whatever bias is in the data is going to be in the result
8:07
Exactly. And they try to train those biases out, but they can't
8:10
If you ask for a story about an engineer, you will probably get a male engineer in the story
8:17
And that's a bias. that's hard to beat. And then there's the well-known risk of hallucination
8:23
Sometimes these models hallucinate answers because they try to generate an answer for you
8:29
And these are risks, but they are manageable. They are. In many ways, they are manageable
8:42
The issue is that market doesn't have the understanding for that yet, and I don't have the understanding for that yet
8:47
and there's no, you know, AI safety tooling is underway and so forth
8:52
But it is a train that's moving. There's many companies that are looking at using this phenomenon in production already
9:04
So it's easy to do and it's easy to set up. You have some data that you need to train it with
9:11
You need to give the AI some data, some information. And then it will very, it will very
9:17
simply for you because you don't have to do anything. It will process your stuff. And so tell me
9:23
about that. How does, we're a regular company. Okay, I, I, let's say I'm a manager of a regular
9:28
company. We have some product, product has some data. Now we're going to AI all the things
9:33
We had a search box before, but that seems like last millennia. We don't want a search box
9:37
That's, that's shallow. We need AI in this, obviously. How do we approach this? What do we do
9:43
So, first of, it would be easy to imagine you being a manager of a very regular company
9:47
I would love to see that regular company and you being the head of that one
9:51
I just made that up entirely. So first of to kind of linger on what you said The first global product we did that we went to production with at the client was in June
10:07
And we didn't train that product with the customer's data at all
10:13
So it was for a certain division for them, for people who were globally with customers to help them in their work
10:24
And going to production took up. about a month of calendar time
10:29
And that is the promise of this foundation model. It's kind of disruptive because it's hard to see a custom project
10:36
that goes to production from start to finish in a month. And the approach of regular companies
10:43
what they do, what I've seen is early adopters might have something in production, like I said
10:49
Most companies are looking to understand what can these kind of models do
10:56
So, of course, open AI isn't the only one. Meta has Lama 2 now out, and that's very interesting
11:04
because meta has this open source-like approach with their LLM, which is interesting, and Google is
11:11
coming out with Palm 2. So there are options, but regular companies that you refer to, their
11:18
leadership teams are not up to speed on what can be done with these models. They use chat GPT. It
11:26
feels magical, it gives me answers, and they want to somehow benefit from that, but they don't
11:32
know what to do in their company. And their IT might, you know, they are filled with work, operational
11:39
and they might not know either. So what I've seen during the last five months, what we've done
11:46
is we've done a lot of educational stuff. You know, one day workshop with demonstrations
11:52
half a day workshops with training. They need to wrap. their head around this. Yes
11:58
And what I've seen happen next is that the company needs to have a brainstorm between the
12:05
business process people and the IT people to kind of understand where could this generative
12:11
AI be useful. And I can give you some examples of that later if you're interested, but that's the first step
12:20
And after that, some project can be started. it's not difficult, but you kind of need to devote time to that in addition to your regular
12:30
business, and you know how hard that is finding time for people is, that's difficult
12:36
It is difficult. I agree. You need to have the ability to set aside a bit of R&D budget there
12:44
but you also need to involve the right people to understand what would we do with it
12:50
So do you have any examples without maybe disclosing brands and so forth
12:54
Do you have any examples of what are companies doing with it
12:58
Yeah. Certainly, audience, feel free to Google Microsoft reference case and all of those things
13:05
Those are interesting. What I usually speak about with my clients are some things I've seen in Finland and Denmark and Belgium so that they feel closer
13:13
And what I've seen is just stuff like, let's take a company with 20,000 employees and they have a
13:22
they have a employee feedback gathering event. And these employees write free text, you know, what's going right and what's going wrong
13:32
And this OpenAI, instead of some people, HR division, reading 20,000 free text responses
13:40
Open AI might summarize all of that, categorize, contextualize, and find not the sentiment
13:47
which is something we've had with machine learning for long time. A sentiment score, yeah, yeah
13:51
Yeah, but find what. What's wrong? Find summarize and find the story
13:56
What do we need to fix? Very easy problem for very many companies
14:03
Yeah, I mean, you could put humans on doing that, working through that and coming up with this solution
14:09
But when you're saying is that you would just give all of that data to the AI and it says, all right, here's the summary
14:17
Exactly. Other other example is this, what we call. enterprise search, which is something Microsoft is coming up with business chat feature
14:28
So there's a chat you can ask questions and it gathers data from, you know, Office 365
14:33
and Graph API and stuff like that. And it gives you a response. Now, what if you want to know
14:39
about your products that are in your German flyers and you put the question in, let's say
14:46
India? Now, Open AI works, even in Finnish, that is the first
14:52
global service that works Finnish that understands and speaks Finnish fluently So it kind of amazing And what is possible with that is that if that company data from product flyers and sales brochures is added on top of OpenAI you can use it as an
15:16
enterprise search that is not possible with the Microsoft business chat because not all data
15:21
is in Office 365. So that's something manufacturing company. especially are very interested in, you know, they are selling an item in Chile
15:33
And someone from Australia wants to understand, you know, you sold this there
15:37
How did you sell it? What was the brochure and what did you say? And you can just search for it in kind of free text
15:44
That actually makes a lot of sense. Actually, I just sideways here kind of something you said
15:51
I remembered like a former disruption in data. management or having a large set of data that we can just use at our fingertips
16:04
Do you remember when Google came out with Google Earth? Yeah. Yeah, right
16:09
That felt, not at this level, but that felt also pretty disruptive
16:13
or at least it was a shock to people that we realized that every location, every street
16:20
everything, every place on the entire globe was a data set that was manageable
16:27
by an application. And since that, that was amazing, you know, all this data and I can just turn the earth around
16:35
and zoom down and I can see everything. That was huge. But now we're kind of at, could we be at the place where now we can do that or it's obtainable
16:47
for kind of anyone to do it with any data set of any kind
16:53
There's restrictions. There's a few ways you can train. GPT and open AI
16:59
models, let's not go there. No. Depends on the project. What's the right
17:03
way to train them? There's faster ways and shorter ways and so forth
17:10
What's happening is that, for example, if there's people in the audience who do search engine
17:15
optimization and they work with Google and Google Ads and these kind of things. And
17:21
now what's happening is when someone now asked from chat GPT some questions about, hey
17:27
what's sailing and what are the rules of football and stuff like that for work and whatever
17:33
in the future and there's this feature out already called calling function and it means that
17:42
chat GPT you can add as a company you can add a component in Azure OpenAIA service and say hey
17:49
if someone asks about this kind of product pull out product information from an API
17:55
Now this means that the future is, Magnus, next time your tooth aches or you need to go to a doctor or something
18:04
In a few years, it might be a chat. And you chat, you ask, hey, what's the nearest free doctor in 500 meter radius
18:12
And it fetches data from public APIs of your medical companies in Sweden
18:17
And it presents you these. And then you click one and it makes the reservation
18:22
And that would mean that medical companies wouldn't do. have, you know, their calendar on the front page that they now have, I think, globally
18:30
It would bypass Google search. It would all be this chat experience
18:35
And now when Lama 2 from Meta, for example, is coming to a size where it's almost
18:42
possible to put it on a phone and it can just reach out to APIs, public APIs
18:49
to get the data that is real time instead of old data
18:53
like it was like six months back yeah I don't know man it's it's it's it's it's
18:59
it's probably a big disruption for for all of us in the sense where we want to go
19:03
what do we want to train what do we want to learn and what are what are we going to do next and it's it's easy to understand why all of these the well the major giants
19:11
that we know about right the the meta and Google and Microsoft it's easy to
19:16
understand why they feel that they definitely need to be on this boat they have to
19:20
be on this boat they have to yeah Absolutely. So listen, this is actually coming to a close now. It's time to wrap up this
19:28
I know already. We could talk about this for hours, possibly days, right? But we have to finish it
19:36
It's been a real pleasure having you here today to lay out the land of what the heck is AI and why are we doing that
19:44
So thank you for being on the Cloud Show. And I hope to see you back sometime in the future
19:50
Thank you, Magnus. It was a pleasure to be here. Thank you. Bye, everyone
#Intelligent Personal Assistants
#Machine Learning & Artificial Intelligence
#Management
#Software
#Text & Instant Messaging
#Voice & Video Chat