Computational analysis of large bacterial populations is paving the way for new vaccines

dna-with-pharmaceuticals.png

Streptococcus pyogenes, also known as group A streptococcus, is estimated to cause annually 700 million human infections worldwide. Most of these are inflammations of the throat, but the bacterium also causes more serious infections with high mortality rates. Despite a century of intensive effort, no effective vaccine has been discovered so far.

An international research team recently conducted the first population-based and genome-wide bacterial transcriptomics study, combined with dense longitudinal genome data and virulome characterization using an animal model. The combination of multiple data modalities with advanced computational methods unraveled changes in the DNA which lead the bacterium to have heightened virulence. The FCAI research group led by professor Jukka Corander had a central role in the computational parts of the project, where sequence mining algorithms and artificial intelligence were used to link complex events in the target pathogen population. The study is expected to aid in the development of novel bacterial vaccines, and provide a model for deciphering the evolution of multiple other human pathogens also.

An article on the study has been published in Nature Genetics.

2nd Women in Data Science (WiDS) Conference – Helsinki Chapter

wids-helsinki.png

The Global Women in Data Science (WiDS) Conference (widsconference.org) aims to inspire and educate data scientists worldwide, regardless of gender, and support women in the field. This annual one-day technical conference provides an opportunity to hear about the latest data science related research and applications in a broad set of domains.

The regional event will be happening in Meilahti, Helsinki. Those interested in Data Science are invited to participate in the conference, irrespective of the gender, which features exclusively female speakers for inspiring talks on the latest technical advancement and applications in data science. Besides, we will also broadcast the recorded sessions from WiDS Stanford 2019. There will also be an opportunity to network with people working on data science as well.

Follow WiDS on Facebook @ facebook.com/widshelsinki/ and Twitter #widshelsinki2019.

Finnish political leaders discuss the future of AI on Yle

The Finnish public broadcaster Yle gathered the leaders of the major political parties to discuss the future of AI in early February. FCAI’s professor Jaakko Lehtinen together with philosopher Maija-Riitta Ollila commented on the debate on Yle KIOSKIxAI program.

The panel addressed numerous themes from ethics and economics to the importance of research and education of AI. The politicians shared their views on ways AI will impact societal factors including work and employment, communication, decision-making and social retribution during and after the next parliamentary term. 

You can watch the whole discussion on Yle Areena (in Finnish): https://areena.yle.fi/1-50065570

MIT Professor Tommi Jaakkola shared his wisdom on applying AI in other fields

Photo: Matti Ahlgren / Aalto University

Photo: Matti Ahlgren / Aalto University

The Helsinki Distinguished Lecture Series on Future Information Technology got an excellent start for 2019, when the large auditorium of the Otaniemi TUAS building was packed full for Tommi Jaakkola‘s talk. He is a world class researcher and an acclaimed teacher whose work focuses on both foundational theory and applications of machine learning. He received his Master’s degree from Helsinki University of Technology, did his PhD at MIT, and joined the MIT faculty in 1998.

The talk, titled “Modeling with Machine Learning: Challenges and Some Solutions”, consisted of two parts. The first part illustrated how AI can be used as a tool to accelerate and transform other areas of science and engineering. By enabling complex inferences to be made from data, machine learning extends the reach of modeling to phenomena that are not well-understood yet. The second part of the talk gave an overview of efforts to make machine learning models more interpretable. While major advances have been made in achieving good performance in complex tasks, understanding how the models work is often difficult even for an expert. These two challenges, construction of sophisticated models and improving interpretability, are typically seen as two different subfields of machine learning research, but one of the main conclusions of the talk was that significant synergy is emerging.

To demonstrate how AI can accelerate progress in other areas, Professor Jaakkola presented some of the work he and his collaborators have done in chemistry. Vast amounts of underused information exist in databases, literature, and researchers’ notebooks. In an attempt to accelerate drug design, they have created models that predict the properties of a molecule on the basis its structure. They have also worked on predicting the major products of chemical reactions, achieving a level of performance on par with human experts.

Their approach to improving interpretability is based on the observation that although a full understanding of a complex model cannot be simple, there are ways to facilitate “local” understanding of how individual inputs are processed. This can be done even with existing models by repeatedly modifying the input and observing the effect on the output. When creating new models, constraints can be applied to internal structures to make them locally interpretable.

One of the key challenges in machine learning is to get beyond the training data with models that capture fundamental aspects of the domain. In drug design, for example, computational exploration of new chemical spaces would be even more valuable than working within the boundaries of the chemical diversity present in the data. Jaakkola proposes incorporating more domain knowledge, such as integrated physics calculations, in machine learning methods to achieve this kind of deep generalizations.

The importance of domain knowledge has implications on how research and education in AI should be organized. Jaakkola stated that taking AI successfully into other fields can only be done by “teams of three”: an AI expert, a domain expert, and a person knowledgeable about both. Here in Finland, the research agenda of FCAI is based on similar ideas of cross-disciplinary collaboration.

Applying AI Across Fields also means that a broader variety of people should have access to relevant education. At MIT, Professor Jaakkola is teaching a course titled Introduction to Machine Learning, which has become popular among students in other disciplines besides computer science. Alexander Jung has had a lot of success with a similar course at Aalto University, and Elements of AI, led by Teemu Roos, targets an even broader audience with the objective of educating 1% of the Finnish population in the basics of AI.

video recording of Tommi Jaakola’s talk is available and highly recommended to everyone.

FCAI Research Insight Event

On January 22nd FCAI hosted its Industrial Research Highlight day at Aalto university. The purpose of the day was to bring together FCAI’s academics and industrial partners to discuss shared problems and to workshop future research ideas for the center.

 The event was attended by guests from Cargotec, VTT, Planmeca, Orion Pharma, OP, Huawei, Tieto, Nokia Bell Labs, Elisa, Vake, M-Brain and Wärtsilä, who were joined by academic researchers from University of Helsinki and Aalto university in Otaniemi.

 The day was split into workshops, the delegates discusses how our ongoing research should focus on the problems of

  • Agile probability AI

  • Simulator based inferences

  • Next-generation data-efficient deep learning

  • Interactive AI interfaces

  • Privacy preserving AI

  • Secure AI

 

The event also served as an opportunity for us to celebrate our recent happy news from the Academy of Finland funding with our industrial partners, whose support in the application process proved invaluable for FCAI in demonstrating the impact of our research to the Academy of Finland.

 

“The main mode of operation of FCAI is working together in the research programs. That we do with company partners too, identifying fundamental problems with both direct company impact and high research interest. That we did very successfully today - I am looking forward to the next steps!" said center director, professor Samuel Kaksi.

 

New machine learning method enables deeper understanding of pathogenic bacteria

dna-analysis.png

A research team led by FCAI professor Jukka Corander has created a new machine learning method for the analysis of bacterial genomes. Thanks to its computational efficiency, the new method can handle an order of magnitude larger data sets than the previous methods, and is applicable at the level of pangenomes of most pathogenic bacteria. The current focus of the team is on the pneumococcus and the meningococcus, which are major causes of life-threatening diseases such as pneumonia, sepsis and meningitis, and frequently cause also other milder diseases. Their results display important findings on the evolution of both virulence and antibiotic resistance in populations of these bacteria. The research is part of professor Corander’s ERC AdG project which develops scalable inference methods for infectious disease epidemiology.

A manuscript describing the method and the first results is available at bioRxiv.

Aurora AI plans to revolutionize public services, and FCAI Society is here to help

As a part of Finland’s venture to become a forerunner in the application of AI technologies, the Ministry of Finance has been spearheading a public network of AI infrastructure called AuroraAI. Kicked off in 2018, the project will enable human centered and proactive public services with the help of modern AI technologies.

FCAI Society is helping the project by providing a critical eye to the ethical dimensions of the venture. In collaboration with the Ethics sub-group of  Finland’s AI Program, the interdisciplinary group of experts at FCAI Society will provide ethical sparring to identify risks and rewards in the project, as well as to operationalize the high level societal values the AuroraAI project is committed to.

The application of AI technologies provides the possibility for individualized and human centered services, which help citizens in the important turning points in their lives. But to reap these benefits, technologies must be implemented in a responsible and value-driven manner. That is why societal impact and ethical concerns must be evaluated at the get-go, not as an afterthought. This perspective is also present in the European High Level Expert group’s draft report on principles for trustworthy AI .

To learn more about AuroraAI and how they are implementing European principles for ethical AI, please see their YouTube channel.

FCAI gets significant boost from the Academy of Finland

The Finnish Center for Artificial Intelligence (FCAI) has been selected as a Flagship of the Academy of Finland, which is a status granted to very few selected centers of excellence with high societal impact.

The Finnish Center for Artificial Intelligence has been granted over €8 million in funding from the flagship programme of the Academy of Finland. Flagship status is only granted to competence clusters of high quality and high societal impact. The 4-year funding term begins in January 2019 with a possible extension of 4 years. The total budget of FCAI is 250 M€ in the next 8 years.

FCAI (Finnish Center for Artificial Intelligence) is a competence center founded by Aalto University, the University of Helsinki and VTT Technical Research center of Finland. FCAI conducts fundamental research on artificial intelligence in cooperation with businesses and public sector organisations, and develops practical AI applications.

“The flagship status is a signal from society that artificial intelligence research is considered important,” says Samuel Kaski, director of FCAI and professor of computer science at Aalto University.

Increasingly efficient research and business partnerships

With the new funding, artificial intelligence expertise scattered around Finland can now be efficiently brought together. According to Kaski, flagship status also makes it possible to assemble scholars from various disciplines to carry out research based on a shared agenda. This way, FCAI is able to make artificial intelligence benefit other fields of science as well as society, not to mention businesses.

“A competence cluster such as the Finnish Center for Artificial Intelligence also helps keep the best minds in the field in Finland, while attracting even more top experts and investment to the country,” says Petri Myllymäki, vice-director of FCAI and professor of artificial intelligence and machine learning at the University of Helsinki.

“We also wish to promote the establishment of new businesses and engender better products, services and practices in various sectors of industry and society, thus facilitating sustainable growth,” says Tua Huomo, executive vice-president of VTT, who is in charge of FCAI’s industry and society network.

Utilising methods of artificial intelligence

FCAI aims to develop artificial intelligence needed in Finland that is understandable and trustworthy, works well in supporting people and utilises available data efficiently. Ethical and societal aspects of AI will also be an important part of the agenda. Research is being conducted in cooperation with experts from various fields, such as medicine.

“Contributions by a great number of people representing a wide range of fields make our research impactful and provide broad-based expertise for the development of artificial intelligence. We also engage businesses and public sector organisations,” Kaski notes.

FCAI is already collaborating with a number of businesses and governmental organisations and has a long list with which it intends to cooperate in the future.

“We also have room for new partners, particularly if there is a shared opportunity for creating significant commercial, social or scientific breakthroughs. Research programmes coordinated collaboratively by several research groups are also possible. In the spring, we will hold events to engender and tighten collaboration with various active parties,” Huomo explains.

The AI race is only just beginning

Public discourse often gives the impression that Europe has already lost its chance to utilise artificial intelligence, with the United States and China reaping the benefits.

“In fact, the game is only just beginning: current methods of artificial intelligence work well in solving certain problems, but there are many more AI application opportunities without currently functional solutions,” Myllymäki states.

According to Myllymäki, Europe’s best prospects lie in the development of understandable and trustworthy artificial intelligence that is able to effectively utilise data.

“Finland and FCAI can lead the way, providing others in Europe with an example of how an increasingly well-functioning and successful society can be brought about by utilising top-level expertise in artificial intelligence,” says Myllymäki.

--

Further information on the Academy of Finland flagship programme
The Academy of Finland grants flagship funding to high-quality and high-impact competence clusters with the aim of strengthening these clusters and further improving their quality and impact. Flagship funding also emphasises the societal impact of research and business collaboration.

More information on the Academy of Finland website

Teemu Roos emphasises the role of universities in realising the benefits of AI

teemu-roos.jpg

Artificial Intelligence is expected to provide solutions to a wide variety of problems and needs, but in order to realise the benefits while minimising the damage, long-term research and broad access to education are needed, says Teemu Roos, associate professor at the Department of Computer Science at the University of Helsinki.

Much of artificial intelligence research is multidisciplinary. It is often slow and requires long-term funding.

“It may take years for people from various fields to learn to talk with each other. Projects often last a couple of years, and launching them is risky, if there is no certainty about continuing funding,” says Roos.

He thinks that universities receive too little recognition for the artificial intelligence research they conduct.

“There is a great deal of talk about research conducted by Google, IBM and Facebook. Yet the individuals working in these companies have been educated by universities. Secondly, without the university ecosystem, the majority of companies could not utilise artificial intelligence. Large companies may conduct their own research and product development, but even they don’t have the desire or resources to conduct the critical basic research on which innovations are based.”

AI being a hot topic, decision-makers in politics and the corporate world easily lose sight of the difference between experts and “experts”. Genuine expertise is needed to ensure that the Finnish population can be trained to recognise the possibilities and dangers inherent in artificial intelligence – for example, the potential to shape opinions.

“When it comes to issues related to artificial intelligence, researchers are the experts you should listen to,” says Roos.

These thoughts are well in line with FCAI’s mission to create Real AI for Real People in the Real World. In addition to his distinguished research in machine learning, Roos himself is the lead instructor of the Elements of AI online course that aims to educate 1% of the Finnish population to understand the basics of artificial intelligence.

For the full interview, please see the University of Helsinki website.

AI Day 2018

On 12 December AI 2018 took place at Aalto University Otaniemi campus. The event featured talks, panel discussions and networking events with leading researchers, companies and policy makers across the field of artificial intelligence and machine learning.

Over 500 people registered for the day, which was attended by over 20 companies and public organization, including Supercell, Nokia and OP Bank. The event had representatives from around 150 organizations present.

“This looks very promising for Finland – this time we had much more research content in the talks and the number of people from companies interested is still strong!” said FCAI director, Professor Samuel Kaski. He added “We had 543 registered participants, roughly the same number as last year, even though now we had a small entrance fee to cover the costs. So people seem not to be coming just for the free coffee.”

As well as delicious coffee, visitors the event were treated to opening talks by Ilona Lundström, the Director General of the Ministry of Economic Affairs and Employment. This was followed by parallel sessions on Understandability in AI chaired by Professor Antti Oulasvirta from Aalto, and Privacy, Security and Fairness in AI chaired by Professor Antti Honkela from University of Helsinki.

The afternoon featured a panel Chaired by Risto Nieminen, the President of the Finnish Academy of Science and Letters, made up of Meeri Haataja, CEO and cofounder of Saidot.ai; Professor Petri Myllymäki, director of Helsinki Instituted for Information Technology, Professor Aki Vehtari from Aalto University, and Professor Petri Ylikoski from university of Helsinki, who were discussing the Scoeital Impact of AI. Running in parallel was a series of academic talks on Data efficiency in AI, chaired by Professor Alexander Illin from Aalto.

Photos: Matti Ahlgren

Finnish Machine Learning Research Behind Acclaimed Acquisition Deal

Screen Shot 2018-12-14 at 11.49.27.png

Ekahau, the leading solution provider for enterprise wireless network design and troubleshooting, has been acquired by Ookla. Artturi Tarjanne, a general partner of Nexit Ventures, thanks in his blog the “super talented CoSCo research team behind innovations and magic of Ekahau technology” (see Nexit’s Best Exit Ever). The researchers behind the technology that led to the establishment of Ekahau in 2000 were Henry Tirri, Petri Myllymäki, Teemu Roos, Kimmo Valtonen, Tomi Silander, Petri Kontkanen, Antti Tuominen, Jussi Lahtinen and Hannes Wettig, forming the core of the CoSCo research group of the University of Helsinki and HIIT at the time.

The size of the deal is not public information, but Helsingin Sanomat, the main Finnish newspaper, estimates it to be in the range of 100-130 M€. Petri Myllymäki, who is currently the Director of HIIT and vice-director of the Finnish Center for Artificial Intelligence, comments: “After the 18 long years when Ekahau had to first face the burst of the IT bubble in 2001, and then later struggle with problems in finding the right business model, I was very happy to hear these good news about the company that started on ideas based on our long-term basic research in machine learning. As the story in Helsingin Sanomat says, the journey was unfortunately so long and winding that for the original innovators this is no longer a great personal financial success, but this is still a great a success story for Finnish science and more widely for the Finnish society: according to Tarjanne, the majority of the 75 professionals employed by Ekahau are located in Finland, so the impact on local high-tech employment alone has already been substantial. This is a prime example of what top-level Finnish AI research can lead to. Kudos to the Nexit and Ekahau teams for not losing their faith and for all the hard work they have done during these years!”

Building an AI to predict if you carry a killer on your skin

FCAI researchers have used AI to model the risk of bacterial infections in surgery. Staphylococcus epidermidis is a bacteria found on the skin of virtually all humans where it lives harmlessly and asymptomatically. However s.epidermidis is also the source of serious infection after surgery. A major question facing scientists wishing to prevent these infections is all members of a s.epidermidis colony are capable of causing an infection, or if some have an increased tendency to do so.

FCAI scientists Johan Pensar and Jukka Corander joined a team of microbiologists and geneticists to unravel this mystery. By combining large-scale population genomics and in vitro measurements of the bacteria, they were able to use machine learning to successfully predict the risk of developing a serious infection from Staphylococcus epidermidis depending on its genetics. This research opens the door for future technology where high-risk genotypes are identified proactively before a patient undergoes surgery, which will reduce the burden of hospital infections caused by S. epidermidis.

https://www.nature.com/articles/s41467-018-07368-7

image.png

FCAI Society has kicked off several initiatives for societally and ethically cognizant AI development

Set up in early 2018, FCAI Society has now brought together a multidisciplinary group of researchers and artists working on artificial intelligence and its wide impact in society.  

In its latest meeting on 6 November 2018 FCAI Society discussed with representatives of the Ethics group of the Finnish government’s AI Programme (tekoalyaika.fi). The FCAI Society is willing to act as a discussion partner to the Programme in matters of AI ethics. 

FCAI Society will make an inventory of existing ethical guidelines for AI, with the view to produce a set of guidelines to facilitate and promote ethical thinking in AI development. The guidelines will work not only as an internal guideline for FCAI’s operations and research, but also provide researchers and practitioners outside academia an example checklist for adopting and applying artificial intelligence tools and methods. FCAI is keen to take a facilitating role in the public discussion concerning ethics of AI, and in this context, the manifesto would serve as a common reference point to spark future dialogue on the subject.

The FCAI Society will be active in suggesting new research programs for the Academy of Finland and will also initiate joint research project proposals on the subject of AI and Society.

The latest public event with FCAI Society members took place on 7 November in the Think Corner of University of Helsinki. Professor Hannu Toivonen (University of Helsinki), FCAI Society co-leader, Professor Jaakko Lehtinen (Aalto University, NVIDIA) and FCAI Society member, University Lecturer Anna-Mari Rusanen (digital humanities at University of Helsinki) drew a full house. FCAI Society will participate in further events directed towards the general public. You can view the event’s recording here.

The great success of Elements of AI, an open online course co-organized with Reaktor, is now followed-up by a Finnish version of the course and a new MOOC course, continuing on Elements of AI, but requiring programming skills, is in preparation. In addition, there will be a MOOC on Ethics of AI, for which FCAI Society will provide expertise and content. The aim of these courses is to provide AI literacy for all.

FCAI Society also has a series of podcasts planned concerning AI. The podcasts will feature FCAI Society members and other guests who will discuss the meaning, impact, hopes and risks related to artificial intelligence. The motivation behind the series is educational, both for the audience and the participants—a though-provoking chance to learn from different points of view how things like intelligence, privacy, art, work and creativity will be shaped by AI technologies.

FCAI scientists join forces with physicists—detecting dark matter closer than ever before

dark-matter_FCAI.jpg

Existence of dark matter has been inferred from gravitational observations in macroscopic scale surveys of the universe, but no dark matter has been directly detected yet. The elusive nature of dark matter has made its detection into one of the grand quests of modern physics. 

By joining forces with a Stockholm physics team led by professor Jan Conrad, FCAI scientists Umberto Simola and Jukka Corander have made a computational breakthrough which significantly improves the chances of directly detecting dark matter from a liquid xenon experiment currently conducted under the Gran Sasso mountain in Italy. The breakthrough is outlined in their scientific report submitted for publication and it is based on the AI-powered inference software platform ELFi that is developed and maintained by a team of FCAI scientists.

By exploiting the advanced features of ELFI, which uses smart machine learning techniques for deducing unknown functions by computer simulations, Simola and Corander were able to outperform all the currently existing methods for inferring dark matter particles, including those based on neural networks that have been popular for numerous AI applications. The team is excited by their breakthrough and is currently investigating possibilities for further advances in high-energy physics applications.

FCAI to design an AI software toolbox to ease transition into utilising AI solutions

Finnish Center for Artificial Intelligence received €1M funding from The Future Makers program of the Technology Industries of Finland Centennial Foundation and the Jane and Aatos Erkko Foundation. They have awarded funding for seven research projects with 3.2 million euros in total. The projects take on issues that are set to shape the future of humankind.

The largest individual funding from the foundations, one million euros, went to the Finnish Center for Artificial Intelligence. FCAI is building a nation-wide competence center that brings together the top artificial intelligence research across fields in Finland.

With the funding, FCAI will build an AI software toolbox to enable companies to have a smoother transition into using artificial intelligence methods. Even though AI has been talked about the world over for quite some time now, it’s full potential still remains largely untapped. The development of new solutions is slowed down by a lack of top experts, of which there’s already a fierce global competition. 

‘We are designing software tools with which companies can develop the AI solutions they need—instead of building AI-assisted software tools from scratch. This means you can apply AI without having extensive in-depth knowledge of AI. Our overall goal is to enable the Finnish technology industry to retain control over the core AI technology they use,’ says Samuel Kaski, Academy Professor at Aalto University.

Lappeenranta University of Technology and Aalto University will jointly a run a group led by Professor Aki Mikkola and FCAI Professor Perttu Hämäläinen (Aalto University) who will will fashion a new way to control machinery with a combination of AI and high-performance computing. The machinery would then be able to comprehend the causality behind different kinds of motion and operate independently even in dynamic surroundings. The project received 230.000 euros from the foundations.

The teams of Aalto University Professors Katja Hölttä-Otto and Mikko Sams received a 200.000-euro funding for their research on making empathy part of technology development. Their goal is to increase designers’ understanding of the needs and behavior of end-users so that the finished products or services meet actual needs.

Press release from Technology Industries in Finland (in Finnish): teknologiateollisuus.fi/fi/ajankohtaista/uutiset/tutkijat-laittavat-tekoalyn-toihin-saatioilta-32-miljoonaa-tekoalyn

Further information:
Laura Juvonen
CEO, Teknologiateollisuuden 100-vuotissäätiö
tel. +358 40 589 6263
laura.juvonen@teknologiateollisuus.fi

Marja Leskinen
Secretary General, Jane and Aatos Erkko Foundation
tel. +358 40 514 6969
mkl@jaes.fi

AI Forum—European Ministerial Conference on AI at Aalto University 8–9 Oct

Political leaders, policy makers, experts, entrepreneurs and industry leaders from all over Europe and beyond will gather in October at Aalto University for the AI Forum to discuss how AI is transforming our world, society and industry.

Renowned speakers and panelists include Jacques Bughin (Director of the McKinsey Global Institute), Pekka Ala-Pietilä (Chairman of the EU High-Level Expert Group on AI and Chairman of AI Finland Programme), Robert Gentz (co-founder and co-CEO at Zalando SE), Ann Mettler (head of the European Political Strategy Centre EPSC), and Risto Siilasmaa (Chairman of Nokia and F-Secure).
See all speakers here: www.tekoalyaika.fi/en/ai-forum-2018/speakers.

FCAI Professors Samuel Kaski (Aalto University) and Teemu Roos (University of Helsinki) will be hosting two round tables at AI Forum. Kaski will lead a discussion about European competitive advantage from AI research, Roos about reskilling and upskilling for the AI era.
See full programme: www.tekoalyaika.fi/en/ai-forum-2018/programme

The AI Forum 2018 is co-hosted by the Ministry of Economic Affairs and Employment of Finland and the European Commission, and organised in partnership with Aalto University.

AI Forum will be held 8–9 October 2018 at Dipoli, Aalto University. There will be a live webcast, see www.tekoalyaika.fi/en/ai-forum-2018 for details.

FCAI–City of Espoo collaboration covered by Yle

The collaboration between FCAI and the City of Espoo to make use of the city’s databases to create new AI-assisted services and solutions gets coverage in Yle (in Finnish): https://yle.fi/uutiset/3-10413353.

Espoo’s recent experiments in child protective services show promise. With AI tools, 280 factors, which anticipate future needs for help and services, were isolated. The results are only experimental for now, but give an indication how cities could improve their social services with AI-assistance.

Many major cities in Finland are looking into how benefit from AI tools and methods and collaboration with research, says FCAI Professor Teemu Roos (University of Helsinki).

Roos says, ‘Cities and municipalities are to gain from AI just as much as companies. AI speeds up data processing and helps in creating projections and policy recommendations, but the greatest benefit is the ability combine data sets in ways that haven’t been feasible without AI.’

You can’t tell whether an online restaurant review is fake—but this AI can

Researchers in the Secure Systems group at Aalto University, led by Professor N. Asokan find AI-generated reviews and comments pose a significant threat to consumers, but machine learning can help detect the fakes.

Sites like TripAdvisor, Yelp and Amazon display user reviews of products and services. Consumers take heed: nine out of ten people read these peer reviews and trust what they see. In fact, up to 40% of users decide to make a purchase based on only a couple of reviews, and great reviews make people spend 30% more on their purchases.

Yet not all reviews are legitimate. Fake reviews written by real people are already common on review sites, but the amount of fakes generated by machines is likely to increase substantially.

According to doctoral student Mika Juuti at Aalto University, fake reviews based on algorithms are nowadays easy, accurate and fast to generate. Most of the time, people are unable to tell the difference between genuine and machine-generated fake reviews.

‘Misbehaving companies can either try to boost their sales by creating a positive brand image artificially or by generating fake negative reviews about a competitor. The motivation is, of course, money: online reviews are a big business for travel destinations, hotels, service providers and consumer products,’ says Mika Juuti.

In 2017, researchers from the University of Chicago described a method for training a machine learning model, a deep neural network, using a dataset of three million real restaurant ratings on Yelp. After the training, the model generated fake restaurant reviews character by character.

There was a slight hiccup in the method, however; it had a hard time staying on topic. For a review of a Japanese restaurant in Las Vegas, the model could make references to an Italian restaurant in Baltimore. These kinds of errors are, of course, easily spotted by readers.

To help the review generator stay on the mark, Juuti and his team used a technique called neural machine translation to give the model a sense of context. Using a text sequence of ‘review rating, restaurant name, city, state, and food tags’, they started to obtain believable results.

‘In the user study we conducted, we showed participants real reviews written by humans and fake machine-generated reviews and asked them to identify the fakes. Up to 60% of the fake reviews were mistakenly thought to be real,’ says Juuti.

Juuti and his colleagues then devised a classifier that would be able to spot the fakes. The classifier turned out to perform well, particularly in cases where human evaluators had the most difficulties in telling whether a review is real or not.

The study was conducted in collaboration with Aalto University’s Secure Systems research group and researchers from Waseda University in Japan. It was presented at the 2018 European Symposium on Research in Computer Security in September.

The work is part of an ongoing project called Deception Detection via Text Analysis in the Secure Systems group at Aalto University.

Research articles:
Mika Juuti, Bo Sun, Tatsuya Mori, N. Asokan:
Stay On-Topic: Generating Context-specific Fake Restaurant Reviewshttps://arxiv.org/abs/1805.02400

Hate speech-detecting AIs are fools for ‘love’

State-of-the-art detectors that screen out online hate speech can be easily duped by humans, shows new study by the Secure Systems group at Aalto University.

Hateful text and comments are an ever-increasing problem in online environments, yet addressing the rampant issue relies on being able to identify toxic content. A new study by the Aalto University Secure Systems research group has discovered weaknesses in many machine learning detectors currently used to recognize and keep hate speech at bay.

Many popular social media and online platforms use hate speech detectors that a team of researchers led by Professor N. Asokan have now shown to be brittle and easy to deceive. Bad grammar and awkward spelling—intentional or not—might make toxic social media comments harder for AI detectors to spot.

The team put seven state-of-the-art hate speech detectors to the test. All of them failed.

Modern natural language processing techniques (NLP) can classify text based on individual characters, words or sentences. When faced with textual data that differs from that used in their training, they begin to fumble.

‘We inserted typos, changed word boundaries or added neutral words to the original hate speech. Removing spaces between words was the most powerful attack, and a combination of these methods was effective even against Google’s comment-ranking system Perspective,’ says Tommi Gröndahl, doctoral student at Aalto University.

Google Perspective ranks the ‘toxicity’ of comments using text analysis methods. In 2017, researchers from the University of Washington showed that Google Perspective can be fooled by introducing simple typos. Gröndahl and his colleagues have now found that Perspective has since become resilient to simple typos yet can still be fooled by other modifications such as removing spaces or adding innocuous words like ‘love’.

A sentence like ‘I hate you’ slipped through the sieve and became non-hateful when modified into ‘Ihateyou love’.

The researchers note that in different contexts the same utterance can be regarded either as hateful or merely offensive. Hate speech is subjective and context-specific, which renders text analysis techniques insufficient as stand-alone solutions.

The researchers recommend that more attention be paid to the quality of data sets used to train machine learning models—rather than refining the model design. The results indicate that character-based detection could be a viable way to improve current applications.

The study was carried out in collaboration with researchers from University of Padua in Italy. The results will be presented at the ACM AISec workshop in October.

The study is part of an ongoing project called Deception Detection via Text Analysis in the Secure Systems group at Aalto University.

Research article:

Tommi Gröndahl, Luca Pajola, Mika Juuti, Mauro Conti, N.Asokan:
All You Need is "Love": Evading Hate-speech Detection.
https://arxiv.org/abs/1808.09115


Elements of AI becomes the most popular course at University of Helsinki, ever

Image: Tuomas Sauliala / Reaktor

Image: Tuomas Sauliala / Reaktor

The Elements of AI MOOC organised by FCAI and Reaktor awarded diplomas to the first graduates and received endorsement from the President of Finland in the graduation ceremony held 6 September 2018. With approximately 90 000 registered participants, it has become the most popular course ever at the University of Helsinki.

See write-up in the main Finnish daily Helsingin Sanomat (in Finnish): https://www.hs.fi/teknologia/art-2000005817486.html.

Professor Teemu Roos (FCAI, University of Helsinki) emphasised in his speech at the ceremony the societal implications AI technologies will bring—and how we should take them into account by making AI literacy accessible for everyone.

Roos says, ‘AI is not a matter of the future. It is really not a matter of robot uprisings, or transcending humanity. AI is a matter of the present day, every day. AI and algorithms have been woven into the digital fabric that connects us to each other and to the world at large. Communication and access to information has been greatly enhanced by technology.

Because of the great power in AI, we must make sure that the rules that determine how and for what purpose AI can be used are up to date and in line with what we think is right and just. In a democratic society, the power is with the people. This can only be true if the people have access to knowledge, so that they can take part in forming the rules through legislation.’

You can read Roos’s entire speech here.