Aktuality  |  Články  |  Recenze
Doporučení  |  Diskuze
Grafické karty a hry  |  Procesory
Storage a RAM
Monitory  |  Ostatní
Akumulátory, EV
Robotika, AI
Průzkum vesmíru
Digimanie  |  TV Freak  |  Svět mobilně

Nicolò Sgobba on supercomputer IBM Watson and artificial intelligence

12.8.2019, Milan Šurkala, Karel Polívka, rozhovor
Nicolò Sgobba on supercomputer IBM Watson and artificial intelligence
Aritificial inteligence is a big topic these days, either because of its potential or risks. Therefore we interviewed Nicolò Sgobba from IBM, who is a Data Analytics Consultant and a Watson Information Architect.
In our interview today we are going to talk about artificial intelligence and supercomputer IBM Watson. We have talked with Nicolò Sgobba, who is Data Analytics Consultant and Watson Information Architect at IBM. This interview is also available in Czech language (zobrazit rozhovor v češtině). First of all, he is going to introduce IBM Watson to us.
IBM Watson is an Artificial Intelligence (AI) system. AI is what gives machines the power to learn, adapt to new inputs and make better decisions. Machine learning (ML) is a subset of AI and uses computer algorithms to analyze data and make intelligent decisions based on what is learned. It's how streaming web sites recommend new music by comparing what one listener likes to others with similar tastes. Watson uses a sophisticated ML technique called Deep Learning (DL). Behind DL, there are algorithms that create Artificial Neural Networks (ANNs) that can continuously learn on the job determining whether decisions were correct, constantly improving the quality and the accuracy of results. This is what enables Watson to learn from unstructured data such as photos, videos and audio files.
Logo of IBM Watson

AI is changing the way the world works, making business faster, smarter and more secure. At IBM we are helping companies put AI to work at scale, giving them an unparalleled business advantage. With Watson, businesses are personalizing customer experiences, streamlining processes, minimizing risks, and sparking innovation.

With Watson, engineers are keeping millions of elevators moving, predicting when they will breakdown and proactively fixing them. Watson can tell an insurance company exactly how a car is been damaged in a crash by discerning the make and model before comparing images of an undamaged car to the damaged one. With Watson, banks are deploying virtual agents trained on thousands of customer inquiries helping them to provide expert service to millions of customers faster. With Watson, a range of industries from Healthcare to Automotive to Telecom to Education, are working faster and smarter.
Nicolò Sgobba
Nicolò Sgobba's own introduction: At IBM, I am currently a Data Analytics Consultant, helping and advising clients from a variety of Industries on how to start with cognitive solutions based on Watson. By leveraging clients’ data, my job is to identify, create and present to C-level clients (i.e. Chief Technology Officers, Chief Information Officers, Senior Managers, IT Directors) recommendations that have to result in services operational efficiency gains or other business opportunities. I have been a Watson Information Architect for a client leading in the Travel and Transportation Industry, and I have been designing and implementing knowledge domain models using IBM Watson technology to deliver end-users solutions through a Cognitive Assistant. On top of this, I am an IBM Senior Inventor with patents in the fields of Brain-Computer Interface, Security, and Speech-To-Text technology, and I am leading as Director a series of initiatives about Patenting, Innovation, and Research for the Technical Expert Council of Czech Republic and Slovakia.

1. What kind of development of IBM Watson is in Czech Republic?

The IBM footprint in Czech Republic spans from Prague to Brno. In Prague, we have the Watson Innovation Lab, that already realized hundreds of Proof of Concepts (POCs) across Europe for AI use cases. An example of a successful POC is around the areas of Chatbots and Virtual Assistants, for Service Desks' calls deflection and first level support, where end users can ask Watson in a first instance, and then speak with a human only if Watson cannot help. Another example is in the area of image recognition and processing for manufacturing, for cases like quality inspections. We have also seen combinations of image recognition technology with Internet of Things (IoT), where the team realized drones that can inspect areas, and notify concerned authorities in case of anomalies. Also, significant steps ahead have been done to let Watson to learn Czech language.

In Brno, where I work, we have the Client Innovation Center, with thousands of IBMers from a variety of nations that are every day supporting hundreds of clients and their IT services (Cloud, Network or Security, …). We also have teams focused on Automation, like mine, that design and develop cognitive solutions based on IBM Watson technology and services.

2. IBM Watson provides high computation power for large companies and research teams working with Artificial Intelligence (AI). Could you tell us some of the large and well known companies which are using Watson and what is it used for?

That’s correct. At IBM we design, build and run the foundational systems and services that the world relies on, literally the backbone of the world's economy. IBM Watson is the number one Artificial Intelligence (AI) for business with more than 20 thousand IBM Watson client engagements across 20 industries. We do a lot of things, we are leaders in the Hybrid Cloud market, in the Enterprise services and Security, in Blockchain and for the 26th consecutive year, we are at the very first place for patents in USA, and IBM inventors received a record of 9100 patents, including more than 3000 in AI, Cloud and Quantum Computing. We have a lot of happy clients and success stories, I suggest you go through our website, there is much more than I can tell - https://www.ibm.com/watson/ai-stories/

3. Are there some breakthrough achievements which were accomplished using IBM Watson and IBM is really proud of them?

The first time the world got to know Watson, is when Watson competed in a US TV quiz show named "Jeopardy!" and won, beating the reigning champions. This happened around 8 years ago, in 2011. At that time, Watson was a computer running some early form of machine learning and natural language processing software. In 2014, IBM brought AI for business with the introduction of our Watson platform. And early this year, IBM announced a new stage in this journey, from AI experimentation to wide-scale deployment and industry transformation: IBM has made Watson portable across any cloud, empowering them to start deploying AI wherever their data resides. That is also why, today, Watson is the world's first and most advanced Artificial Intelligence (AI) platform.

IBM Research continue to advance Watson's capabilities while building the future of AI, with the development of new software, services, applications, and next-generation infrastructure. And it's not what IBM achieved during these years, but more what our clients achieved and will continue to achieve with Watson – you can read some stories here: https://www.ibm.com/watson/with-watson/

4. How would you describe how IBM Watson operates to someone who doesn't know anything about today's AI algorithms and who thinks that it operates only as a very robust decision tree connected to very big database of relevant data?

A lot of people view and perceive AI as some magical technology that is being put to work like it is a black box, with little understanding of how it works. They might view AI as something out of this world which is relegated to experts only, and for this, AI has taken on an air of mysticism with high promises and out of the reach of the ordinary people. The truth, of course, is there is no magic to AI. The term Artificial Intelligence was first coined in 1956 and since then the technology has progressed, disappointed, and re-emerged... and this will happen again and again, that's the only sure thing.
And as it was with electricity, the path to AI breakthroughs will come with mass experimentation. While many of those experiments will fail, the successful ones will have a considerable impact. And this is exactly where we are today, and as several people already said, think about AI as "the new electricity". In addition, while becoming ubiquitous and increasingly accessible, AI is also enhancing and altering the way business is conducted around the world. It is enabling predictions with great accuracy and automating business processes and augmenting decision-making: the impact is vast and ranging from greater customer experiences, to intelligent products and more efficient services. And in the end, the result will be an economic impact for companies, countries, and society, finally all of us.
For all these and other reasons, organizations that drive mass experimentation in AI will win the next decade of market opportunities. To help demystify AI, we need to consider two key elements: the componentry and the process - in other words, identifying what is behind AI and how it can be adopted. Regarding the componentry, much like electricity was driven by basic components such as capacitors or diodes, AI is being driven by math, by models, by algorithms, by software. The software enables the process, unlocking the value of data. To conclude, there are no magic nor "decision trees", but intelligent software, capable of analyzing data and to make intelligent decisions based on what is learned.

5. How do you train and use AI algorithms in practice? How long it will take to train it and what are the biggest challenges in order to make a reliable AI?

To also link with my previous answer, let me start answering to your very last question. In my opinion, the fact that people do not trust AI (yet), mostly depends on perceiving this technology as a black box, because it is not clear how it works. Also, we have been reading in the news of software and algorithms making unfair or unethical decisions - it takes 30 seconds on Google to find a variety of these examples. But again, this is a new technology we are still developing and experimenting, and with experiments, we should also expect some failures.
At present, a crucial principle and challenge for both humans and machines is to avoid bias and therefore prevent discrimination. Bias in AI systems mainly occurs in the data (bad data used to train AI can contain implicit racial, gender, or ideological biases) or in the algorithmic model. As we work to develop AI systems we can trust, it is critical we develop and train these systems with data that is unbiased and to develop algorithms that produce explainable results we all can trust.

To finally answer to the whole question, what we are constantly working on today is on mitigating human bias in AI. Last year we announced the MIT-IBM Watson AI Lab, whose efforts are drawing on recent advances in AI and computational cognitive modeling, such as contractual approaches to ethics, to describe principles that people use in decision-making and determine how human minds apply them - to make AI transparent and explainable. The goal is to build machines that apply certain human values and principles in decision-making. IBM researchers developed a methodology to reduce the bias that may be present in a training dataset, such that any AI algorithm that later learns from that dataset will perpetuate as little inequity as possible. In terms of how we train AI systems in practice, this depends on the problem, the available data, the amount of data, the technique and this also affects the needed time.

6. Is IBM Watson suitable even for smaller tasks to have really meaningful results? It looks like it is needed to be something big for which it would make sense to train it for which could take a lot of effort and time. Can you comment on this topic?

IBM clients are typically enterprises, and our services and solutions are mainly designed for them. Hardware resources can be a limitation as AI and deep learning applications at some point can hit what I refer to as "the infrastructure wall". What we offer to our clients is a full-stack cloud platform that spans public, private and hybrid environments, with hundreds of products and services covering data, serverless, containers, AI, IoT, and blockchain. At IBM we think at enterprise scale.

7. How do you see the development of autonomous cars today? When do you think the SAE Level 5 cars which can really go everywhere in all weather conditions will be reality?

That’s a very interesting, broad and debated topic because it involves several technologies and "factors". To better explain, let me start clarifying the classification from the Society of Automotive Engineers (SAE): in this classification that ranges from zero (0) to five (5), 0 represents a fully manually-operated vehicle and 5 is a completely autonomous vehicle. At present, I believe we do not have a proper infrastructure nor the necessary regulations, and that all this uncertainty makes very complicated to make some accurate estimation.
Anyway, my personal perspective here is that the main problem remains, in several ways and levels, the "human factor". Why? Imagine a system where you have fully autonomous vehicles only: this is the perfect case, everything goes just fine because all the vehicles in the system are constantly communicating and exchanging information with each other making easy for every vehicle at which speed to proceed, which path to follow and so on - and if the system is a cognitive one, this will also improve, becoming more efficient over time.
To this system, add also Smart Cities, with millions of sensors everywhere. Put also 5G technology. That's the paradise for the system, tons of data from any kind of sensors on which to leverage on and just machines. Now, let's introduce a human in this perfect machines' paradise: that's "panic mode" for machines! In this scenario, a human is an unpredictable variable and statistically, it is very likely that some accident is going to happen. To conclude and answer to your question, I believe that until we will have a hybrid ecosystem with both autonomous and manually-operated vehicles, the SAE 5 will remain just a dream. In the meanwhile, there is lot of work to do, lot of problems to be solved, and lot of fun to build and invent this entirely new world.

8. IBM is working on IBM Watson IoT platform which is about monitoring data from sensors in vehicles which are connected to it to improve safety and performance. Can you share with us some latest interesting news from this area?

Let's put it in this way: Internet of Things (IoT) delivers the data and Watson powers the insights from this data. IBM Watson IoT Platform ingests device data and transforms that data into meaningful and actionable insights. We also have solutions for vehicle connectivity, but it's not just vehicles: we can capture and explore data from a variety of devices, equipment, and machines, with the objective to discover insights that can generally drive better decision-making thanks to cognitive analytics. With this technology, we can find optimization opportunities while ensuring that data is accurate and secured and, for instance, we can capture data in real time and spot trends and act accordingly before they have any negative impact.

In terms of the trend in the IoT world, I believe that - and this is my personal view - if you speak with any Technologist, they will tell you that there is an ongoing shift toward Internet of Everything. In this sense, what I truly believe will be an enabler and a game changer, is the 5G technology. With 5G technology, all smartphone users worldwide will be able to download things up to 20 times faster. But potentially far more significant from this, is the capacity for 5G to unlock the full power of the billions of IoT devices and sensors which will be connected to the network. Indeed, all the investments made by the telecommunications industry for 5G must look for growth using software as the differentiator in a broader set of use cases across industries.
This includes apps, analytics and software that address specific user needs; for example, AI-enabled manufacturing, edge-specific decision making, and augmented reality that can assist with everything from utilities repair to construction trade work. In many cases, AI-capable chips will enable edge applications. Chipset providers will likely address both hardware and software for IoT devices, enabling decisions to be made closer to the device instead of sending it to the cloud. What I mean is that more things are expected to be processed locally, at the device level, just as your phone face recognition process happens on the phone itself and not in the cloud. To go further with this “envision the future” game, as the society is becoming more connected, people need more energy to power all this.
Connected cities, energy and utilities, agriculture and transportation are all 5G applications for Massive IoT (MIoT), which increases the scale and speed that edge computing and machine-to-machine applications can deliver while using less power. MIoT use cases increase machine intelligence across areas ranging from the tractors harvesting smart agriculture, to trains delivering connected transport to smart cities. Energy and utility monitoring with MIoT will operate across broader physical infrastructure and smart grid technology responding to changing conditions in real time. Sounds interesting, isn't it? But that’s what you can do when you put together an infrastructure, hardware, software and data. That’s the future, be ready.

9. Even IBM Watson will some day need hardware upgrade to provide more performance if it will be used for increasing number of projects. Do you have some estimate when it could be needed, or is actual hardware still providing enough performance with a lot of reserve for actual usage, so there are no plans for upgrade right now? Is IBM Watson still running on cluster of 90 IBM Power 750 servers with 3,5 GHz eight-core processors with 4 threads per core in each one of them (which means 2880 processor threads in total) with 16 TB of RAM?

I have to admit here: I have no idea! This is something which is transparent to me because over the years, IBM changed the way how Watson and all our cognitive services are accessible and offered to our clients. We have offering, products, solutions and services available through the cloud. For example, you can right now sign up an IBMid and create your IBM Cloud account and by doing so you can start developing at no cost (with cap based Lite plan services), and you can start on your projects right away and play with Watson to see how powerful it is!

10. What is IBM planning to improve IBM Watson in the near future? What are the areas in which the current version is struggling and what needs to be improved or is there some completely new project on horizon which could supersede it?

Watson is powered by the latest innovations in machine learning and is an open, multicloud platform that allows our clients to automate the AI lifecycle. The future is already ongoing: IBM recently announced that Watson is now available Anywhere, so that our clients can apply AI to their data wherever it is stored, helping businesses accelerate their transformation through AI.
We announced new Watson microservices that are based on open source technologies and scalable across cloud environments. Regarding challenges, anyone in this industry could expect that the number of biased AI systems and algorithms will increase. But we as IBM will deal with them accordingly, coming up with new solutions to control bias in AI and champion AI systems free of it. This is an ongoing and identified problem also for any competitor in the industry, and we are already on it.
IBM Watson in Jeopardy
zdroj: Atomic Taco [CC BY-SA 2.0]

IBM also recently revealed Project Debater, created by IBM Research scientists. Project Debater is IBM's next big milestone for AI, following previous breakthroughs like Deep Blue (1996/1997) and Watson on “Jeopardy!” (2011). Project Debater digests massive texts, constructs a well-structured speech on a given topic, delivers it with clarity and purpose, and confutes its opponent. The goal is that the technology that powers Project Debater will help people reason by providing compelling, evidence-based arguments and limiting the influence of emotion, bias, or ambiguity. Why is this so important? The world is awash with information, misinformation, and superficial thinking. Project Debater pushes the frontiers of AI to facilitate intelligent debate so we can build well-informed arguments and make better decisions.
Another exciting thing is IBM Q, which is an industry first initiative to build universal quantum computers for business, engineering, and science. This effort includes advancing the entire quantum computing technology stack and exploring applications to make quantum broadly usable and accessible. A quantum computer is a special kind of computer that calculates things in a completely different way than how a common computer does. This technology will allow us to solve very complex problems which are simply not resolvable with common computers, like mimic the nature. To what pertains AI, using quantum systems to train and run machine learning algorithms could allow us to solve complex problems more quickly, potentially improving applications like disease diagnosis, fraud detection, and efficient energy management. There are a lot of things on which IBM is working on and I cannot be more excited about it!

11. I have been active in image processing research few years ago and I could see that almost all the famous algorithms tend to be replaced by AI processing. Do you think that there is still place for traditional image processing methods and algorithms or it is over and it does not make sense to create anything else than AI algorithm?

To be honest, I am not an expert in this field but generally speaking, any Technologist will tell you that the future is always about something faster, smaller, cheaper... and let me add smarter. We should expect AI technology to be applied anywhere. And if AI can benefit also in this field by reducing, for instance, computational time and cost, and possibly learn thereby improving the accuracy over time, then there might be a point of no return to previous methods - but I guess this is true in any field.

12. Can we see the results of AI in weather prediction? Which weather prediction services are using IBM Watson in order to predict weather or other climate related situations, for example nature disasters? Is there any example where IBM Watson was successfully used to help to save people’s lives?

The Weather Company is a weather forecasting and IT company, and is an IBM Business. The Weather Company delivers personalized, actionable insights to consumers and businesses across the globe by combining the world's most accurate weather data with industry-leading AI, IoT and analytics technologies. These solutions provide pilots, energy traders, insurance agents, state employees, retail managers and more with insight into weather's impact on their businesses, helping them make smarter decisions to improve safety and reduce costs. You can read a success story here: https://www.ibm.com/case-studies/energie-nb-power-outage-prediction-weather-company

Another thing which is absolutely worthy to be mentioned that also pertains to weather is "Call for Code". Call for Code is a multi-year global initiative, and IBM is a Founding Partner, that inspires developers to solve pressing global problems with sustainable software solutions. The Call for Code Challenge theme of this year is natural disaster preparedness and relief in the context of community health and well-being. This is because mitigating natural disasters is one of the world's greatest challenges. The past decade has been one of the worst periods for natural disasters and while they may be inevitable, they don't have to be catastrophic. We can do so much with technology, also change the world.

13. History has shown that today's AI has its limits. For example, it was shown in the summer 2018 that AI which was using the IBM Watson was incorrectly curing cancer. Where do you see the limits of AI and do you think they can be solved? Are there some new approaches in AI which seems to be promising?

Presently there are several limitations both on the technology - and this is true for the whole industry - and the human and ethical aspects. For example, machine learning is limited by data, by the model complexity; computation is limited by the laws of thermodynamics, and more importantly, AI does not have consciousness. In the 1940s, IBM partnered with Columbia University to create the new discipline of "Computer Science". Today IBM is expanding training and education for the growing number of "new collar" jobs for today's era of data and AI.
There are also a lot of concerns about privacy because with technologies like AI, the need for data is greater than ever and consumers are concerned. I have already talked about the problem of biased AI systems. Identifying and mitigating bias in AI systems is essential to building trust between humans and machines that learn. As AI systems find, understand, and point out human inconsistencies in decision making, they could also reveal ways in which we are partial and cognitively biased, leading us to adopt more impartial or egalitarian views. In the process of recognizing our bias and teaching machines about our common values, we may improve more than AI. We might just improve ourselves.

To conclude, there is a variety of problems and limitations, and the good news is that the entire industry is aware of them and everyone is doing is best to address this. As also previously mentioned, as it was with electricity, the path to AI will come with mass experimentation and while many experiments will fail, the successful ones will have a considerable impact.

14. AI is used almost everywhere nowadays. But there are not only positive benefits as there are threats as well. AI can replace almost all human profession and there is a huge question mark over the high unemployment. Many writers raised concern about the future and some of them were threateningly accurate (we can mention George Orwell and his Big Brother). There are also concerns about AI used in armies and weapons. People are afraid that something like in the famous movie Terminator can really happen. What is your opinion about the future of AI? Do you think it can go wrong and get completely out of our hands?

Over the decades, AI has garnered a negative reputation as a force that will eventually destroy humanity. But this could not be farther from the truth. Only general artificial intelligence could theoretically have the capability to match human intelligence, and by most estimates, general AI is at least several decades away. Current AI technologies and applications have produced impressive results, and they in several cases outperform humans in specific tasks such as classifying images or transcribing speech.
The point here is that AI technologies are available to augment human intelligence. It is important to understand that the goal is that machines do machine jobs and humans do human jobs, and that the combination of human and machine creates something that none of them can do alone. Humans' strengths come into play as they perform crucial roles such as training machines to perform certain tasks, explaining the outcomes of those tasks, and sustaining the responsible use of these machines. The problem today is upskilling workers for the data and AI age because the whole point is not that we are going to need "less people", but we will need people with different skillsets. Just a few days ago IBM's CEO Ginni Rometty, in an interview on the CNBC, said that AI will change 100 percent of jobs within the next 5 to 10 years, and she could have not explained this concept better. Here is the link: cnbc.com

15. Which professions do you think that AI cannot replace and why?

As just said, I personally believe that over the next years AI will have an impact on each and every one of us and on every profession. But despite all these technologies, all these cool technological toys, and the cognitive assistants that I envision will be more and more present and pervasive while performing certain tasks, I also believe that our "profession of being humans", our way of being humans will not change much. We will still fall in love, lose the loved ones and have kids. We will keep making discoveries, we will keep traveling, creating and writing stories, and all the time that the technology will be able to save for us on our repetitive tasks will be more time that we will have for us to experience all the unique moments of being humans. Technology will have to serve us ethically, empower us, will have to augment humans’ capabilities. That's the job and the shared responsibility for all the people in this industry to make this happen.

16. Why have you decided to participate in this project? What is so attractive for you?

For my personal case, I believe this is not a straightforward answer. I have no clue if I decided to go with technology, or if technology decided to go with me, but as long as I can remember, I have always been fascinated by technology and all started with a simple computer, just like a musician starts because there is a guitar or piano around. My professional career in IT started from hardware. When I was at high school, during summer breaks, I used to work in a small repair lab, where, depending on the day, I destroyed or resurrected hundreds of parts of computers, circuit boards, chips, routers, and printers.
At University, I have studied Computer Science and that's when I focused on software and started to be passionate about the Human-Computer Interaction. As a parallel stream, from the high school period till the end of my university studies, I have also studied acting, singing, communication, and body language (overall 7+ years). I believe this is worthy to be mentioned because it's something that affected my personal and professional growth, also allowing me to focus on a broader set of tasks and interest since the very beginning of my IBM career, which started in 2013. I joined IBM as SysAdmin, and one year later I was leading a Global Virtual Team of Technical Engineers. I have been a Service Manager, working with Clients in the Automotive and Travel & Transportation Industries, and spent almost one year in the Middle East to work on an important project.
Presently, as also mentioned before, I am a Watson Information Architect and a Data Analytics Consultant, advising on IBM Cognitive solutions. I am also an IBM Inventor with patents in the fields of Brain-Computer Interface, Security, and Speech-To-Text technology and, especially in the last 3 years, I have been spending the majority of my time on whatever pertains patents and innovation. The excuse has always been that I like to study and learn new things - which is true, but it is not the whole story.
The reason for all this time spent on innovation initiatives and projects is that in a place like IBM, any idea can materialize and have an impact on all of us, and eventually change the world. At IBM you have the chance to solve real-world problems, and I really feel empowered by this opportunity and responsibility of being somehow able to influence and decide also the direction and the future of the technology. That is why this is so enticing to me, and that is why I am a proud IBMer.
We thank Nick for his answers and we wish him a lot of interesting projects in the future.

Autor: Milan Šurkala
Vystudoval doktorský program v oboru informatiky a programování se zaměřením na počítačovou grafiku. Nepřehlédněte jeho seriál Fotíme s Koalou o základech fotografování.
Autor: Karel Polívka