Welcome to the Ko call webinar. My name is George Barza. I am an AI practice lead and a project manager at Ko call Analyst. Today. I will be speaking about artificial intelligence, disruption ahead, strengthened opportunities, limitations, and risks of current AI deployments. A short background about coping call Analyst. We are an independent and neutral Analyst company founded in 2004. We have offices around the globe with the HQ in Germany. We offer vendor neutral guidance, technical expertise, and top leadership.
We support end user organizations across many industries, system integrators and software vendors with technical and strategic advice. We are specialized in information in cybersecurity identity and access management, identity governance, risk management, and compliance, and all areas concerning the digital transformation. Our business areas provide you with high quality content. We bring communities of experts together, and we offer coaching. We do this to our research activities, which are vendor neutral and always up to date.
Furthermore, we organize topical conferences, workshops, and summits, where you can meet the experts in the above mentioned fields and benefit from the networking opportunities. We also offer advisory services where we work together with you to make your business more successful using our advisory phases, which have multiple steps. We start with evaluating your business needs. That includes support in the selection of suppliers and tools, and also coaching in subsequent deployment phases.
As mentioned above, we also organize various events among them, the European identity and cloud Congress, which is just two weeks away. Our events cover various topics such as digital finance, blockchain for enterprise, those events taking place in Frankfurt on September 18th and 19th consumer identity world tour starts in Seattle on September 25th to 27th. And the European leg is an Amsterdam October 22nd to October 24th. The cyber next summit is taking place in Washington, DC, October eight to 10th.
The cyber security leadership summit takes place in November 12th to 14th in Berlin together with the German speaking access management event, cyber access summit. Furthermore, we have an AI team event. Impact summit dates are 27 and 28. I will be speaking about some of those events in more detail in the coming slides. So a lot of exciting choices for you here now about some housekeeping, you are muted centrally. You don't have to mute or unmute yourself. You control the mute and unmute features. We will record the webinar and the podcast recording will be available tomorrow.
The Q and a session will be at the end of this webinar. You can enter the questions anytime using questions feature in, go to webinar control panel. So here's the agenda. I'm the only speaker of today's webinar. And my talk will be followed by Q and a session as already stated above.
Let us start with the historical overview of artificial intelligence as a separate subject. The term artificial intelligence was first mentioned at the conference abstract and then subsequently at the conference itself called the Dartmouth summer research project on AI. This was in 1956, it's now 2019.
So what happened between 1956 and today after the initial optimism and government funding partially fueled by the cold war, lack of significant results and use cases resulted in a sort of deep freeze. First one in late seventies, then there was a short wave of optimism, which was partially fueled by sales of AI specific hardware and software so-called expert systems. But this was followed with another stagnating period in the late eighties and early nineties.
Nowadays, there is a significant awareness about AI starting with the computers, beating players and chess and go and other games and thus having AI powered assistance in our smartphones.
This revival has several reasons. One of them is the amount of data we collected until now, which needs understanding from our side and structure. But this data can also be used to train the artificial intelligence.
Furthermore, there is a significant increase in computing power and specialized specialized hardware available for specific AI use like tansor processing and graphics processing units, but most importantly, the advancements of machine learning and its subset deep learning has made it possible to develop new speech and image recognition programs. Before we proceed any further, let's look at the AI buzzwords.
The graph I am about to show you is available in the 2017 EIC keynote presentation from Martin Kuppinger founder and principle Analyst of comic, as well as in his recent video blog called AI in a nutshell. So the broad definition of AI it's, it's a science of making computers solve tasks, which usually requires human intelligence.
We can split AI in a strong or general AI where computer has the same mind and senses as human beings. But I think we are far away from that and we move to more real applications. Weaker applied AI is a software focused on solving specific problems.
If you focus on current ongoing AI research, we see a lot of concrete cognitive solutions. That is the practical applications of AI. We have different areas in information security, self-driving cars. And so on. These practical applications are based on cognitive technologies, such as computer vision, language processing, and so on. And these technologies are built on various machine learning methods like pattern recognition, outlier detection, generic algorithms, or deep learning, underlying algorithms and methods are neural networks, cluster, and regression analysis. And so on.
We have a couple of areas which are built in a complex way, but it's not machine learning, but just the regular statistical analysis, even though it seems quite complex and it is complex.
Everything based on predetermined rules is not machine learning until machine is learning and is trained. Then it's not machine learning. Let's talk about the strengths and current, let's talk about the strengths of current AI deployment. We can start with the significant benefits for your business.
For example, the robotic process automation could improve the customer service chats by being smarter than regular dial chat boxes, improve the experience for the customer and also be beneficial for the business owner. They will be no more training and supervision necessary for such a service and AI can be leveraged to reduce costs, grow revenue, and offer generally a better service. It can also help businesses in small scale or just the whole countries to gain competitive advantages.
Many governments are fostering AI development and enhancing various industries where AI can help them to catch up with the leading technological powers.
And AI's already being used in finance and trading by leading trading firms. It has been mentioned that by using those techniques, the companies better understand their data and can focus on collecting only necessary information they require to improve their business. Let's move to the next point, which is increasing productivity and innovation.
For example, retailers are starting to use AI techniques to offer smart recommendations to the customer and logistical companies can use it also to adjust their delivery routes, save time that way. And in general, work in a more efficient way by taking over mundane tasks, AI can leave time for more human innovation. AI itself can augment human creativity by generating novel patterns and approaches to a specific problem. This brings us to the next point, which is AI and humans benefiting and learning from each other.
A good example is the ping pong game, which I recently saw on YouTube between AI powered robot and a decent player. The human player, the machine was learning in real time from a player and adjusting to the player's style while the human player had to change his tactics and approach the game in a new and innovative way.
This example is of course, transferable in some ways to day to day business in many areas, as I said, AI is being applied in various industries, which I mentioned and will mention on next slide speed in cybersecurity, marketing, logistics, healthcare, and many more.
But let's first talk about the limitations of current AI. As I mentioned, two slides before we are currently seeing the deployment of narrow AI solutions. So the software can only tackle specific problem. Maybe it can talk, tackle it very well, but the software will not be transferable to other problems. So there is no one size, one size fits all solution obtaining high quality data sets that are sufficiently large and comprehensive to be used for training is also quite a big challenge. This data can be compromised.
And a good example for is for this is a notorious AI Twitter bot from a famous company, which was instantly attacked by trolls on internet. And it had to be deactivated after just 18 hours of launch, furthermore creating or obtaining sufficient number of sensitive data with medical data or from individuals who are under 18 can be an issue. And it has to be handed handled with extreme care.
Yeah. When people hear about AI, they often think about killer robots or in the best case, they imagine Jarvis from Ironman movies.
And they're either disappointed with the real AI examples in their life or fear that AI will outsmart humans and try to rule the world. Well, I really hope the latter will not happen, but the complexity of AI techniques makes it sometimes difficult to show which factors led to a specific decision or prediction and how this decision was achieved. This is particularly important in applications where the result have some social implications, for example, be it in court cases or lending. And so on.
People are also scared of losing jobs to AI, which is true, but it can be looked in another way that AI is creating jobs either way. The process always is accompanied with public UNE. The last point is concerning the lack of skilled workforce two years ago, roughly two years ago, New York times estimated that the number of machine learning and deep learning experts was just around 10,000 in the world.
Maybe not every company needs an AI genius, but at least understanding the basics.
And the way that technology works will give people a chance to modify it if necessary, to adjust it to specific business needs. But this still remains a limitation. Let's move to a very important aspect, which is AI ethics or ethics in AI to partially avoid public skepticism and fear which I described on last slide. It would be an important aspect to of an AI system. AI system should be transparent, which means that all aspects should be well documented. It should be auditable and testable.
It should have embedded specific societal values, which reflect fair and democratic society and data used for training should not be biased because the biased data may produce undesirable and biased results and prompt customer and social dissatisfaction, wrong results might, might be discriminatory, sexist, or racist, and this will affect society in a, also your business in a negative way. Even if it's not intentional from time to time, we hear about many AI so-called hiccups on the news.
This can be avoided and the world economic forum has recently released the report on how to prevent discriminatory outcomes in machine learning.
By being human centered, the goal of AI should be of benefit to humans. AI system should not be led to run without absolutely any supervision. If this is the case, the results may be completely unpredictable and affect the users in a very negative way. AI system should also be robust and able to withstand the cyber attacks and not cause cause intentional or unintentional harm.
Several weeks ago, the high level expert group on AI presented its ethics guidelines for so-called trustworthy AI, the AI, it put forward several key requirements for AI systems and will also be kicking off a piloting process to gather the practical feedback. I think this is launching in summer of this year. Some companies already are setting up ethics sports for this reason. And I think we definitely need to address this problem before it's not too late. You can also read about AI ethics in a blog of Mike Small senior Analyst at copy our call. Let's move to legal implications of AI.
There is a call for a specific loss and regulations, which will be designed with the AI systems in mind. Until now we saw several calls for ethical AI from companies themselves or from international organizations, NGOs, but we will probably start seeing AI regulation in coming years. Us Congress already raised such a concern. Last October
Development
Of AI standards is, will also be paramount, different countries, approach to development in a different way. Be it for example, USA, China or European union countries.
This could lead to negative experience and the damaging of relations between the countries and the businesses who operate cross borders, regulators and policy makers will be embracing the technology and measure. And we already see models proposed for AI governance. And I think we will definitely see more in this area. One important aspect rising in the legal debate about artificial intelligence is the question of accountability.
If there's an accident because of AI who is responsible, the machine, the programmer, the company, this goes back to transparency and explainability of AI, which I briefly mentioned on the last slide. Another interesting point is the interaction between the current regulations, for example, GDPR and AI. In what ways will GDPR affect the AI development? Probably the answer is not so simple and most probably it most probably means that decision making process used by AI will need to be transparent. As I already mentioned, and data used in AI training will need to have appropriate consent.
If the data is sensitive, AI can also be used to detect GDPR violations, which is a very interesting point in itself.
So is the governmental regulations, the only way or AI should be regulated by collaborative entity between NGOs international organizations and governments, should the tech companies be involved or how, what role will the public play? I think we are looking at say exciting discussions coming our way in the next years.
One thing we can really expect is that rapid development of AI algorithms, we might soon make us face with a new wave of fake materials on the internet, which is at the moment limited to fake AI generated cat pictures. Let us move to some specific interesting applications of artificial intelligence. First of all, is virtual assistance. We all know them. We all use them on our phone and Google unveil the product sometime ago, which can make calls and appointments for you. Similar solutions are offered by other companies as well.
This is very handy and could change the way businesses operate and make, make employees increase their efficiency as well. Facial recognition is another interesting application of AI use cases are endless for this in marketing law enforcement, border checks and crossings also in finance, in banking. But of course this raises number of ethical issues, which I already mentioned before the driverless cars, we all have seen an increase in this area fueled by Tesla developments, but also experimented by other companies.
I think the primary usage in the near future probably will be supply chain and logistics branch, where the technology will be applied in the cargo transportation
In the healthcare branch, AI is being used for cancer detection and also preemptive treatment of another medical conditions. It's very important to know that AI is not perfect and cannot be fully trusted in this area, but neither are humans combination of AI and human doctors is what makes a winning formula for the patients.
There has also been a number of emotional support apps, which use facial recognition techniques to determine if a person is experiencing emotional distress. This can have a life saving effect if used while driving or in general to alert others. When there is a medical emergency, AI can definitely be a POS positive driving force for the globally accessible education, where the classes can be translated in a fast and efficient way, or the students with impaired hearing or vision can benefit from it as well.
AI can provide personalized learning schedule for each student and assist teachers in various administrative tasks. There are numerous apps on the market, which are doing this already. AI is also having an effect on farming industry. There's something called precision farming, which uses AI to get better understanding of nutrients, for example, or water management and optimal times for harvesting
And artificial intelligence, even enter the arts scene where the AI or to be precise, generative adversarial networks made a painting, which was sold for interesting sum of 430,000 us dollars.
So very interesting last but not least is the area of cybersecurity where AI is being utilized as well. And let's look at it more closely. There is a growing complexity of, for the companies, I think infrastructures, where they have massive amounts of sensitive data spread across the multiple clouds and increasing shortage of skilled people to deal with those problems in this data. Even the large businesses with strong security teams cannot keep up with the latest cybersecurity risks.
Currently AI is being used to support as a support function to security experts of a company and is being used to utilize to detect malware. For example, numerous malware variants are being created daily. So to adjust to this evolution machine learning is being used by many companies already. Same for the threat hunting where machine learning helps in proactively and iteratively searching through networks to detect complex advanced threats.
Yeah, multiple AI ML solutions are already available, but more development in this area is definitely necessary. Important question is here where we want to apply these advanced techniques. The technology should be applied in a smart way to tackle specific problems like in other areas using AI can reduce time spent on mundane tasks and free up stuff to work on more likely threads.
The existing challenges are the quality of, and the source of the data on which the models are trained. This data is very frequently containing massive amounts of sensitive information, intellectual property, PII.
Otherwise machine learning systems can be attacked by adversarial machine learning. Data can be poisoned and the models can be easily tricked. We should also remember that these advanced technologies do not replace the it staff or do not reduce the security skills gap. So what's next. I think over the coming months and years, we will hear a lot about AI taking the jobs away from humans. This is partially true. As I mentioned, AI will replace some jobs, but also create jobs in various other ways. There's definitely a skills gap in the field of AI. And this has to change in coming years.
If you are trying to use AI in your company, you need to think first, if you really need it, don't follow the hype, but also don't be left behind. You should apply solid risk and project management
Where
Advancements are to be expected in AI. Technology is the edge AI, for example, edge AI uses the algorithms which are processed locally on the hardware device itself without requiring any connection. It uses the data from the device in processes it to give very fast and realtime insights in less than couple of milliseconds.
This will definitely reduce the chance of the data being tampered during the data transfer, or will also allow for better customer experience and with lower latency. Yeah, the general AI is a research topic of various groups in the world.
But yeah, I think the it's still far away from us. There are skeptics, like for example, Gary Marcus, from university of New York, he published a critical paper about current AI hype pointed out that deep learning might be reaching its limit already. So among technical questions, which she discussed key points out that the field of AI could get trapped in the local minimum dwelling too heavily in the wrong parts of the intellectual space. And that researchers might concentrate on capturing low hanging fruits.
And this would ultimately re lead to neglect of riskier, excursions, and limit AI development. In some ways.
Now let's talk about upcoming coping and call events and let's focus on European identity and cloud conference. There will be numerous talks dedicated to AI at this event focused on identity and access management and cybersecurity use cases, as well as AI, ethics and regulation. We will also organize AI innovation night will where speakers will present short 10 minute slim style talks.
They will speak about AI in identity and access management, security, marketing, and finance as well on our AI themed event in November in Munich, we offer great selection of thought leaders and speakers and the industry. And of course, Cooper our call. Casey talk at our headquarters in Reba in June will give you a great chance to participate in exciting and motivating discussion in more intimate setting with an AI experts and among them professor Christ fun Matthias book, who is one of the AI pioneers, not only in Germany, but in the world.
We have interesting research published recently on AI and there's still more coming. So check our website for the AI teamed research and our events. So now let's go back to questions if they are any. So what can we expect in AI development for the next five years?
Yeah, this is short term question, always a relevant question. I think we'll definitely see a move to more transparent AI. We should expect probably numbers of regulations already being put in place.
And most, probably many companies will try to use AI in some form successfully or unsuccessfully. And we will probably see better AI assistance, better chat bots, which will hopefully free our time to concentrate on more and more creative tasks. So the next question is can blockchain technology be combined with AI and what applications would it have?
Yeah, those are probably two hype words of the last couple of years, AI and blockchain. Well, blockchain probably can play a role and an important role in security of the sensitive data, which can later be used to train the AI.
So yes, I think those two technologies are a good match, but it of course depends on the real use cases and how it's implemented in all the other technical details. But I think there is a, there's a future in that as well. So I don't see any more questions at the moment. So if that's it, thank you for listening and I hope to see you at our events or at our other webinars in the future. Thank you.
