AI will add USD 13 trillion to the world economy by 2030 according to McKinsey’s recent reports. Asia will play a key role when it comes to building solutions that are built to scale. Let’s see where Europe stands and how the two continents can collaborate to advance innovation in AI. Here are edited excerpts from our talk at AsiaBerlin Summit:
Innovation in AI: Europe playing catch-up
If we look at the history of AI, European nations provided the algorithm frameworks and language technology that other nations have built upon. Prof. Dr. Hans Uszkoreit, Scientific Director – DFKI and GIANCE says,
“ Many of the algorithms in AI, from neural networks to autonomous driving originally came from Europe. However, the U.S. has cashed in and been at the forefront of commercial success as in the case of the internet and GPS.”
The US and China are frontrunners when it comes to innovation in AI. European nations are receiving government support but are not necessarily at par with global standards. Why?
How U.S and China stay ahead of the tech curve
There are two ways that the U.S. continues to stay ahead when it comes to innovation in AI. The first is that they do not stop funding when basic research is over. Dr. Hans adds that agencies like DARPA continue to fund research till the technology is usable. This makes it possible for startups to apply research outside labs. This ecosystem was missing in Europe until recently.
“At present, Europe is trying to overturn AI innovation without trying to play catch-up. Europe is well placed to be successful in enterprise AI as China and the U.S. are focused on consumer electronics, e-commerce and hyperscalers,” he says.
SAP: Building unique enterprise AI solutions
AI can help every company upgrade their products be it software, hardware, activity tracking, or creating new solutions. Dr. Fei Yu Xu, Global Head of AI at SAP Cloud Platform and AsiaBerlin Ambassador says,
“With Machine Learning, AI solutions will improve processes like supply chain management using logistic network optimization and customer relationship management with sentiment analysis.”
Prior to heading AI at SAP, Dr. Fei was a principal researcher at DFKI for 19 years and led the AI Lab at Lenovo as Vice President.
SAP’s mission is to use intelligent solutions to empower all the companies in their ecosystem, make them intelligent and make processes more efficient. So far, general purpose or narrow AI which entails object and voice recognition is based on perception. “If we want to make good enterprise solutions, Europe is well placed to be a leader as European companies are exceptional at tracking complex processes, whether it be on the shop floor or in supply chains”, says Dr. Fei.
AI adoption opportunities for SMEs
AI is a buzzword today and many companies are using domains like .io or .ai to gain traction. Mr. Ludwig Graf Westarp, Managing Director – BVMW says, “only 8% of companies are using AI in their organizations.” Ludwig has lived in Vietnam, Shanghai, and Germany. His understanding of enterprise and facility management makes him an expert in smart city development and AI adoption.
There is a lot of potential for companies to adopt AI especially when it comes to SMEs. Instead of force fitting AI for the sake of it, he says that it is important for companies to identify their problems clearly and then ask themselves, “What is the problem I have and where do I want to use it for?” so that they are well prepared for future contingencies.
The crowded Indian AI startup ecosystem
There are more than 2000 Indian AI startups and globally a lot of investment is coming into the country. In India alone, AI startups have the opportunity to tap $1.35 trillion. A hardcore AI start-up needs talent, capital, reliable research and time.
Sidhartha Mohanty, Founder – Powerly AI says, “In terms of starting an AI company and being successful is dependent on one factor- data and without it prediction algorithms don’t work. There are a lot of research institutes that have come up but they have a long way to go.” Powerly AI is a Mumbai-based sustainable HR technology platform that helps companies bring in digital transformation through NLP and voice technologies.
AI-based startups are focused on fashion and e-commerce technology; where companies want to predict consumer buying behaviour. The healthcare sector is exploring early detection of cancer and automotive centres are solving for predictive machine maintenance.
Can AI solve global challenges like a recession or pandemic?
Dr. Hans admits, “ I don’t think that AI itself can solve global problems themselves but it is true, AI will be in every aspect of our lives, our work for sure, and will be used for solving them. Sadly, societies are too reluctant to gather data. As a result, we are unable to use machine learning extensively to understand the epidemiology of the virus.”
Currently, there are two stumbling blocks, one is within AI and the other is societal. Within AI, Dr. Hans says, “we need to move from narrow to broad AI which means not just an end-to-end learning output but we have to combine this powerful machine learning with knowledge graphs.” However, the biggest stumbling block is the inertia of society. He says, “in comparison to eastern societies, western societies are very polarised. Some nations do not want to give up cash! They are scared of giving up control over their data.”
Nav Qirti, Principal – ideactio and AsiaBerlin Ambassador quips, “In technology, the U.S. innovates, Europe regulates and China executes. In this lopsided equation, all three countries are in three different eras regarding AI.” Nav runs Ideactio, a design consultancy based out of Singapore.
For innovation in AI, the real issue is data collection not data access. Dr. Hans says, “more than access to data, getting data is the challenge. EU legislation has made progress but we have a long way to go. China has really embraced AI- from primary schools to equipping public administration.”
Man vs. Machine: Impact of AI
Humans are driven by survival, sustenance, and imagination. With survival, we associate fear; with sustenance, love; and with imagination, meaning and purpose. Humans are driven by emotions, whereas AI is task-driven. Nav says, “So far technology has helped us augment physical strength. However, AI today has helped us reach an inflection point where we are trying to augment intelligence.”
So far, people were the scarce resource. Therefore, when we worked more, we were more productive and we got paid more. “In times to come, jobs will become scarce and that’s a fundamentally different problem to solve. When we don’t have anything to do- our lives become devoid of meaning. When we move away from the lack of jobs the implications will be great, perhaps leading to digital dictatorships or healthcare problems,” says Nav. If we start to see things from this perspective, we need to ask ourselves the right questions.
Nav believes that Europe’s caution with respect to AI is the right approach. However, he also worries that getting stuck in issues of ethics and regulations might result in Europe lagging behind with regard to AI innovation. He says, “on one side is a utopia and on another, dystopia. However, the next challenge in the next 30-40 years would be jobs in mental health with respect to the impact of AI in our daily lives.”
On ground AI implementation
The European Commission has identified innovation in AI committed to allocate Euro 700 million in this area. These are encouraging signs and will lead to positive outcomes in the long term. Although Dr. Hans commends these efforts he expresses his concern over two issues- ethical superiority and technical superiority. He says, “there is a tendency to think that we do things better in the ethical sense without doing well in the technical sense. I think we have to do well in the technical sense and build powerful AI. In order to give the world more transparent, reliable, responsible AI we need to be as good in performance since the world will choose the powerful AI, not the ethically better AI.” According to him, premature legislation is hindering progress. He suggests having feedback loops from the field rather than multiple committees debating legislation on software and capabilities that does not exist.
Collaboration necessities: Diversity, international exchange and a healthy appetite for risk
Dr. Feiyu mentions that she established the first AI lab at Lenovo where 80% of her staff were chinese students. She says, “The majority of students in chinese universities are Chinese. However, when you look at European and North American universities they are very diverse. I think diversity and international exchange will help innovation.”
China is advanced because it tapped the open-source community. They were able to innovate on open source software which was mostly developed by western countries. She says, “western countries can learn from the excellence in execution exhibited by China”. She believes that fear of risks might debilitate us from keeping up with general trends of human society.
Calculating ROI for investors
Talking about the enterprise market, Dr. Feiyu says, “We can classify the enterprise AI market in three categories- AI in existing applications, hyperscale platforms and virtualization platforms. We make processes more efficient, improve user satisfaction rates, drive traffic and bring cost efficiency.”
Moving away from a technical calculation of ROI, Dr. Hans gives us the example of the translation industry. He says, “machine translation translates more than all human translators combined by several orders of magnitude. Still, the translating business thrives and grows by 11% a year. Why is that? There are certain types of human translation that can not be fully automated. Another example is chatbots, they will have interruptions that you will not be able to measure.” In terms of the individual AI techniques, Dr. Feiyu’s method can be applied, but we need to be careful. The most interruptive innovations by AI we will not be able to calculate beforehand.
For Sidhartha, impact is the most important metric to measure ROI for investors and customers.
He says, “Investors and customers want to know the accuracy of your predictions. At Powerly. AI we use NLP and speech recognition to assess if a participant is a good fit for an organisation or not.”
Another metric is time saved and automation of mundane tasks. He says, “playing around with data is very important as customers want to save time and money and automate certain things, which might be repetitive.” In 2019, Amazon discarded their recruitment model as it was biased towards selecting a disproportionate percentage of white applicants.The problem however, was in the data set- it only had data of white people. Data sets have to be sustainable and diverse so that it can be applied to a model that can predict with higher accuracy.
Sidhartha talks about the value his start-up adds to recruiters. He says, “earlier, people made a lot of phone calls to talk to talent during the recruitment process. We have a voice chatbot where the chatbot saves a lot of time- in terms of communication and candidate engagement. It collects answers and provides voice recordings in advance. We analyze these voice recordings and select the most suitable candidates for the position.” According to him, ROI can be measured by the quality of data present in the model and customer satisfaction.
Responsible AI and its inherent biases
Dr. Hans says, “there are no ethical or unethical algorithms, only what you use them for. The bias is usually not in the algorithm but in the data.” There are two problems namely intransparency and biases.
Efforts are being made to combine neural networks so that they are more transparent and accountable. There are also proven methods to remove biases from data. For instance, data can be weighed to discontinue bias in current datasets. In conclusion, he exclaims, “don’t blame AI for being irresponsible, it is the people who are to be blamed!” He urges the need to build trust systems so that data sharing and AI assessment can become a reality between the EU and Asian nations.