Tekoäly fiksun kaupungin rakentamisessa

Tekoäly on aikamme teknologian suuri lupaus. Se liittyy digitaalisuuden megatrendiin. Lähdimme VTT:llä pohtimaan tämän lupauksen suhdetta kaupungistumiseen eli toiseen keskeiseen tulevaisuutta suuntaavaan megatrendiin. Kohdistimme huomiomme siihen missä megatrendit, digitaalisuus ja kaupungistuminen kohtaavat eli fiksuun kaupunkiin.

Pidimme 17.11.2017 My Smart City -työpajan, jossa kysyimme millaisia mahdollisuuksia tekoäly ja yhdistetty todellisuus tarjoavat kaupunkien kehittämiseen? Millaista tietoa kaupungeissa on ja kuinka sitä hyödynnetään? Millaisia tulevaisuuden näkymiä on ja miten tekoäly ja yhdistetty todellisuus voivat auttaa yhteiskehittämisessä kaupungeissa? Työpajan lopuksi pureuduimme eettisiin ja psykologisiin teemoihin aiheen ympärillä.

Keskityn tässä blogissa tekoälyyn, koska se tuntuu herättävän enemmän ristiriitaisuuksia ja samalla uusia mahdollisuuksia kaupunkien kehittämisessä. Yhdistetty todellisuus on laajemmin jo hyväksytty juttu. Yhdistetyn todellisuuden avulla voidaan visualisoida kaupunkilaisille ja muille päätöksentekijöille tulevaisuuden ratkaisuja kaupunkiympäristössä. Tämä on todella hieno ja hyödyllinen juttu, ja samalla on hauskaa ja kiehtovaa siirtyä tulevaisuuteen virtuaalilasit silmillä! Tekoälyn mahdollisuudet kiehtovat nyt tässä enemmän siksi, että siinä on myös jotain vastustettavaa.

Tekoälystä pormestari?

Kaupunkien toimintojen ja ympäristön suunnittelu on täynnä päätöksentekoa, joka luo tulevaisuutta. Hyvässä päätöksentekoprosessissa kerätään tietoa, analysoidaan se ja sitten tehdään päätös. Voisiko tämän prosessin hoitaa tekoäly, ehkä paremmin kuin esimerkiksi poliitikko? Tekoälyllähän voidaan automaattisesti kerätä tietoa, ja se ymmärtää nykytilan isoon aineistoon perustuen, ja pystyy sitten ennustamaan tulevaa, ja sitten se vielä oppii pikku hiljaa. On jopa esitetty, että kaupunkien pormestari voisikin olla tekoäly! Sanottua on myös, että tekoälyn avulla pidetään ihmiset mahdollisimman kaukana päätöksenteosta mahdollisimman pitkään, ja näin saadaan aikaiseksi parempia päätöksiä, joissa kaikenlaiset valittajat eivät pääse hidastamaan ja sotkemaan päätöksentekoa.

Mutta, kuinka voisimme luottaa päätöksentekijään, joka ei erota chihuahuaa mustikkamuffinsista? Tekoälyhän ei välttämättä osaa tätä erottelua, ja vaikka se oppisikin pikku hiljaa, niin käytännössä tekoäly osaa vain verrata tietoa aiempiin kuviin, joista sille on kerrottu mitä ne esittävät. Eli nähtyään tarpeeksi monta muffinssia ja tarpeeksi monta chihuahuaa, se alkaa kohtuullisella varmuudella erottaa ne toisistaan. Tekoäly oppii siis lineaarisesti ja siinä mielessä näkee myös tulevaisuuden nykyisyyden jatkumona. Miten se osaisi käsitellä yhteiskunnassa tapahtuvia murroksia, joissa kehityskulkujen suunta muuttuu?

Tekoäly on aina ensin jonkun ihmisen alkuun laittama, jonkun ihmisen, jolla on omat arvonsa, kulttuurinsa, intressinsä ja ymmärryksensä. ja koska tekoäly on ihmisen tekemä, niin se on kulttuuriin sidottua. Niin myös sen keräämä aineisto, analyysi ja päätöksenteko kuvastavat joitain arvoja ja uskomuksia, ihan niin kuin meidän ihmistenkin. Jos haluamme, että tekoäly analysoi meille, kuinka tehokkaasti saamme aikaiseksi 25 000 ihmisen asuinalueen, jossa on 10 000 työpaikkaa, opetamme sen tekemään tällaisia analyysejä. Sen sijaan, jos haluaisimme tietoa siitä, kuinka 25 000 ihmistä olisi tyytyväinen asuinalueellaan, meidän on perustettava analyysi jotenkin muuten.

Suuri potentiaali on minun mielestäni keskustelevassa tekoälyssä. Voisimme esimerkiksi asentaa puistoihin tai leikkikentille keskustelevia penkkejä tai keinuja, joiden avulla samanaikaisesti kerättäisiin tekoälyn ja kansalaisen välisistä keskusteluista tietoa päätöksentekijöille, ja samanaikaisesti tehtäisiin kaupunkilaisille hauska keskusteluelämys. Ainakin oma 8-vuotias poikani juttelee ja pitää hauskaa mielellään puhelimen tekoälyn kanssa. Puiston penkille levähtämään istunut kulkija voisi halutessaan joko jutella penkin kanssa, tai sanoa sille, että olen pahoillani, mutta tänään minua ei keskusteluta.

Toinen mielestäni lupaava huomio tekoälyn mahdollisuuksista on yhteisäly, jossa ihmisen ja tekoälyn muodostava kokonaisuus kykenee aiempaa parempiin päätöksiin. Esimerkki tällaisesta voisi olla prosessi, jossa tekoälyn avulla kerätään tietoa ja analysoidaan se karkeasti. Tämä karkea tulos annetaan sitten ihmisten käsittelyyn, jolloin kokemus ja arkiymmärrys pääsevät vaikuttamaan analyysiin. Käytännössähän toteutamme tällaista prosessia jo esimerkiksi navigaattoria käyttämällä. Jos meillä on aiempaa kokemustietoa valittavana olevasta reitistä, saatamme jopa valita navigaattoria vastustavan reitin. Navigaattorin tekoäly ei siis pääse ohjaamaan meitä silloin, kun emme tarvitse ohjausta.

Laajoja tietoaineistoja on kuitenkin usein mahdotonta käsitellä ihmisaivoin loogisesti; tässä tekoäly voi auttaa meitä. Voidaan myös esimerkiksi tekoälytutkija Timo Honkelan ajatusten ohjaamina, uskoa, että tekoälyn avulla voisimme lisätä demokratiaa päätöksenteossa; tekoälyn avulla tuhannet tai jopa useammatkin ihmiset voisivat osallistua päätöksentekoprosessiin ihan oikeasti. Tekoäly voisi siis oikeasti olla fiksun kaupungin olennainen elementti, päätöksentekoa tukeva työkalu. Kyllä minä edelleen toivon, että kehitys vahvistuu tähän suuntaan, mutta niin, että me ihmiset pysymme kuitenkin asioiden hallinnan johdossa.

Lue lisää:  www.vttresearch.com/sustainable-and-smart-city

Nina_Wessberg
Nina Wessberg
Research Team Leader, VTT
nina.wessberg(a)vtt.fi
@NintsuW

 

 

Tämä blogi on ensimmäinen tekoälyä ja yhdistettyä todellisuutta fiksun kaupungin toimintaympäristössä tarkastelevassa blogisarjassamme, joka tullaan tästä eteenpäin julkaisemaan VTT:n blogeissa vuoden 2018 alkupuolella. Blogisarjan avulla haluamme elävöittää tekoälystä ja yhdistetystä todellisuudesta käytävää keskustelua.

Will machines take our work? – Part 1: Healthcare and unempathetic artificial intelligence

The Finnish population is getting older, which is creating pressures in the healthcare system. Robots have been envisioned as a potential means for alleviating the workload of nurses. But is artificial intelligence suited for nursing? Would you want a robot to steer your physiotherapy? Can a surgical robot perform as well as a surgeon?

It is generally thought that robots steered by artificial intelligence (AI) will replace people in increasingly complex work tasks. I personally examine the matter from the perspective of work research. On this basis, I can sense if some suggested technological development path appears difficult to implement. On the other hand, work research offers a good perspective on the development of technology, because it offers a medium for identifying the needs of employees, and the division of labour between a human and a machine.

In this article, in addition to healthcare, I touch upon the meaning of empathy in practical work. People have their own experience of what it means to be a human being, so they are capable of placing themselves in some other people’s position, in other words, considering a matter with his or her situation, perspective and feelings in mind. Almost all work tasks require some sense of empathy, since work almost always serves other people’s needs. In most professions, you work for either a customer or an employer, and generally in collaboration with others.

Artificial intelligence is good at playing games

How people act at work is largely based on the expectations of other people and their more or less emphatic understanding of other people’s points of view. Robot, on the other hand, executes rules programmed in it. In addition, a machine can be programmed to edit its own rules, or to learn. Provided with a massive amount of data, a learning AI may develop quite extensive skills within operating environments with clearly definable rules, causal connections and goals. An AI may be good at playing games, and it can be used, for example, for increasing the productivity of a social media marketing campaign.

Even though there are certain rules in healthcare, nursing is not strictly steered by rules, since the work requires situation-specific flexibility. Every patient has a different body, mind and precise clinical status. Therefore, also the way the patient is medicated, washed, operated on, massaged or nursed in general varies. It is not advisable to treat patients with a formal routine. Even though the work may appear as being routine, research shows that nursing includes continuous and discreet micro-level decision-making and adaptation of working methods. Besides on medical training, this decision-making is based on intuition and – as I at least assume – the personal experience of an empathetic care worker of what it means to be a human being.

In a greater degree than most other sectors, healthcare is characterised by uncertainty, since, at an individual level, the outcomes are difficult to predict. Patient’s recovery from, say, a surgical procedure always involves an element of chance, regardless of how well the operation went.

Empathy enhances the quality of treatment

Even though it is not advisable for a care worker to start feeling what the patient is feeling, a good employee acknowledges that the patient’s feelings are of importance with a view to recovery and the quality of treatment. Without empathy, it would be difficult to calm down a patient verbally or by touching.

As a rule, being incapable of empathy, AI is not well suited to replace humans in care work, since care work is difficult to model in the language of mathematics due to the complexity of the phenomena involved. Feelings, the bodily and verbal interaction between people, and intuitive patient-specific decision-making at micro level are phenomena that are difficult to measure and control.

The existing care robots perform simple tasks in hospital logistics, provide entertainment and activate patients. Toylike robots with slight resemblance to humans may act as physical trainers for patients. Research results seem to indicate that patients find robots more motivating and pleasant than exercise videos. Some robot-like intelligence may be programmed in physiotherapeutic devices, i.e., the exact form and challenge level of therapy can be automatically adapted to the patient’s performance. However, the ability of AI to provide personal advice or guidance to patients is very limited or non-existent. In other words, robots will not replace physiotherapists. Devices are not capable of providing hands-on guidance or making comprehensive analyses of patients.

The surgeon’s responsibility and the patient’s destiny

I have personally studied surgical work in particular. In surgical procedures, good experiences have been gained from automation performing certain limited and precisely defined parts of an operation. This takes place under the surgeon’s close control, and the robot has no responsibility whatsoever for the overall performance of the operation. Surgical robots are devices surgeons steer with their hands and feet. The operation is performed through tubes, and the surgical robot also provides a 3D view inside the patient’s body. Viewed through a robot’s eyes the world looks strange and peculiar; object recognition is a real challenge for a surgeon in such an environment. Further development of sensor technology could be of assistance in this matter.

If I personally end up under a surgeon’s robot one day, I would not want the operation to be performed by a superficially skilful surgeon who is in a way playing a computer game with my body. A surgeon follows a pre-agreed surgical plan, but changes are made as the situation requires, because the view of the patient’s clinical status may become more accurate or even change during surgery. Furthermore, a surgical procedure always involves a difficult conflict: the goal is to treat the patient, but surgery always causes some damage as well. For example, in a prostate surgery, the aim is usually to remove the prostate gland containing cancer cells from the body. If you remove too much tissue, the nerves and muscles important for erection and continence may suffer, but if you do not remove enough, cancer treatment does not work. A comprehensive understanding of the patient helps in the decision-making, when the surgeon needs to solve this conflict during the surgery – sometimes the patient may even be so old that it does not matter much whether he maintains his erectile function or not. It is difficult to draw a scientifically established direct connection between sense of empathy and surgical decision-making, but I would personally hope that the surgeon performing the operation would understand at a personal level how important sex and continence are.

OLYMPUS DIGITAL CAMERA
Mikael Wahlström
Senior Scientist, PhD (Soc Psych)
mikael.wahlstrom(a)vtt.fi

 

The author has participated in various automation projects and analysed security-critical work, for example, as a head of the Academy of Finland project called WOBLE that studied the work of surgeons who use robotic systems. WOBLE was funded by the Finnish Work Environment Fund and it was implemented in collaboration with the Finnish Institute of Occupational Health and Tampere University Hospital. The final report of the WOBLE project in Finnish can be found here.

The next two parts of the three-part series of articles will focus on logistics and on modelling machines after humans.

Will robots make us better persons?

How will our society change as artificial intelligence and robotics develop? To what kind of new humanity will robotics liberate us? Will the added value brought by robots be available for the well-to-do only? We must start discussing ethics of new technology before this technology becomes an integral part of our everyday lives. The essential thing is to keep posing questions, even if there were no answers to be found right away.

Often it is even more essential to keep posing questions than finding answers to them. This is true at least when faced with ethical questions. Ethical questions related to the use and instilling of technology require asking questions, consideration and debate from versatile perspectives and at different levels. This is particularly important, when we are in the process of pushing people whose voice is not always very clearly heard in our society to use technology: those living in the margins of our society. They need us others, who might be able to give them a voice through our own deliberations.

Very seldom ethical consideration of matters reaches a fruitful level if they are discussed between own team members or colleagues only; Establishing a wider perspective requires views from various angles and stakeholders.

Ethical technology and many-voiced approach

I ran a workshop that VTT organised in collaboration with RoboBisnes operators, or the North Karelia Municipal Education and Training Consortium and Karelia University of Applied Sciences. With a group of 40 people, we spent a whole afternoon delving deep into the opportunities offered and the concerns raised by robotics and artificial intelligence. We sought perspectives, for example, from the everyday lives of older people and the mentally disabled.

Ari Tarkiainen, Project Manager at Karelia University of Applied Sciences, made an important observation at the session: “In a way, ethics is kind of an inherent part of new technology, since new applications and opportunities produce a lot of situations of which we have no previous experience. It is also descriptive of the current situation that such new situations have not been taken into account in legislation and no practices have been established for managing them. Therefore, ethical questions should be strongly highlighted all the time. VTT has been acting as a key expert and developer in this collaboration between ethics and technology”.

Technology is not black and white

Even though universal ethical values guide us to consider what is good and bad, or right and wrong, technology is never black and white. When we listen to each other – and also really hear what is being said – the border between black and white begins to waver and we begin to see bright colours and different shades of grey. In the hum of voices (and North Karelians are known for being eager to talk!), the values shared by most of us find a fairly comprehensible form within the Finnish framework.

After a while, that clarity fades away, when we keep on examining these values from a multicultural perspective:

  • Which values can take us forward?
  • What creates trust in society?
  • What kind of fringe areas does digitisation create, and who live in these fringes?
  • To what kind of new humanity would technology going beyond our thinking capacity liberate us?
  • Why does technology sometimes raise issues of insecurity and vulnerability regardless of the fact that it also opens up new enchanting paths in our everyday lives.
  • What is the ethical thread that will last until the end?

When discussing these questions, and going forwards and backwards, we came up with some positive visions of robotics. “Robots enable easy-to-use user interfaces and increase digital inclusion. Maybe, with the help of ugly robots, we also learn to accept the different appearances of people. It is great that robots do not know how to have tantrums! I could quite easily trust them with all cleaning duties.” Some female participants were also of the opinion that, luckily, robots are quite advanced, unlike human males, which are still being developed in the right direction in many households. Tears were running down people’s cheeks with laughter, and there was room for all kinds of opinions at the session venue, near the Joensuu market place.

Talking and laughing together, sharing our common experiences, does good to us people. That is something a robot is unlikely to be capable of any time soon. But robots can liberate us of from many dull tasks to having the kind of ‘quality time’ together as described above.

We will organise more similar workshops in the future. Consideration of the borders and framework conditions of humanity is important, and right now, in the middle of major changes, it is particularly important.

Read more at www.vttresearch.com/services/digital-society

jaana_leikas

Jaana Leikas
Principal Scientist, VTT
jaana.leikas(a)vtt.fi
Cell: +358 407 500 211

Artificial intelligence is already here – and the expectations are high

Why is artificial intelligence (AI) popping up in every direction nowadays? The expectations for the almost miraculous power of AI are high. In the following, Antti Vasara, VTTs President & CEO, reveals whether there is any reality behind these expectations.

Antti Vasara VTT

Artificial intelligence (AI) and the opportunities provided by it can generate a totally unprecedented, almost exponential productivity leap, and raise the human race to totally new spheres of prosperity. The ground is fertile for any positive news about AI, even though ever since the 1950s every decade has witnessed an enormous hype about something that has ended in disappointment. However, now we have justified reasons to expect more. At least technologies have taken major leaps forward, thus making the application of AI much more realistic today than in the years before.

Modelled after humans

What exactly is AI? A few years back, American researchers Russell and Norvig defined it in the following manner: “Artificial intelligence is… when a machine mimics ‘cognitive’ functions that humans intuitively associate with human minds, such as ‘learning’ and ‘problem solving’.” Another definition starts with the same idea, but expresses it a little bit differently: “Artificial intelligence is what computers cannot be programmed to do yet.” Here, the idea is also that the computer learns to perform desired, clearly limited tasks without being given precise instructions on how it should complete them.

Technically speaking, AI requires three important matters:

  1. Algorithms
    • Algorithms have been mainly developed over the past decades, but they have been collected into serviceable platforms only recently: deep neural networks in particular (deep learning). The platforms are generally controlled by technology giants.
  2. Data for teaching algorithms
    • The digital data available for teaching purposes and analysis is increasing exponentially.
  3. Computation capacity
    • Computation capacity available at low rates is increasing.

We have begun to see how practical, real solutions can be achieved both in consumer and industrial applications. Consumer applications are visibly displayed: examples Google Now, Amazon Alexa, Apple Siri, Microsoft Cortana and many computer games.

Work tasks may change, but work does not disappear

The industry also applies AI, with industrial and process automation, autonomic transport on land, sea and in the air, and health care leading the way. Even though these industrial applications are aimed at producing better outcomes with the help of AI, probably the biggest impact will be seen in the automation of routine tasks: productivity will grow, work tasks will disappear – and new ones will emerge.

Academic researchers and consultancy firms are almost unanimous about the productivity growth. Precise figures vary, but according to McKinsey&Company’s estimate existing technologies would allow automation of up to 50% of work tasks. However, in many cases this is not economically viable. On the other hand, very few professions can be fully automated. The most susceptible duties to the impacts of AI are tasks involving routine processing and management of data.

Transition is an opportunity

In history, many similar transitions have taken place. Almost without an exception, this has lead to improvements in the standard of living and quality of life. So, let us be optimistic this time as well and make efforts to ensure that AI becomes yet another positive phase in societal development.

VTT is a pioneer in the application of AI. Our precise role is to produce the kind of AI applications that produce well-being for society and growth for companies.

Read more on the subject in our policy brief publication (in Finnish): “Tuottoa ja tehokkuutta Suomeen tekoälyllä” (Enhancing Finland’s productivity and efficiency with the help of artificial intelligence).

Antti Vasara
President & CEO
Twitter: @ahavasara

Finland will be a winner in the Artificial Intelligence disruption

Artificial Intelligence (AI) is now being hyped in earnest. The media is full of stories about its achievements: magical algorithms beating humans on their own turf in Jeopardy, the Chinese go-game and Texas hold’ m, not to mention the stock traders and radiologists soon to be displaced by machines. IBM was very clever to name its AI software suite Watson. A bit of mystique is always intriguing.

And that is not all: venture capital is pouring into AI start-ups – 15 billion US dollars since 2012! The number of investment deals increased by a factor of 4.6 between 2012 and 2016. The prices big boys pay for AI companies are amazing – for example Intel paid $15 billion for Israeli company Mobileye, which develops smart vision for future autonomous cars.

But there is also a darker side to this. Many renowned researchers predict that 25–40% of tasks will be automated in around ten years’ time due to robots, autonomous cars and the like, powered by AI. Martin Ford, futurist and author, expressed his fears about this in the book Rise of the Robots: Technology and the Threat of a Jobless Future. Basic income may be a solution, he says.

No need to panic

We should be neither hyped nor alarmed. Instead, we should carefully examine the opportunities AI and automation are bringing to Finland. We should begin by recognising that this change will not be sudden and short lived, but long-term and profound.

I prefer to liken the AI disruption to the change in farming and agriculture driven by mechanisation after the Second World War. Finland was very much an agricultural society, with around half of its people working in farming and forestry. They endured hard work and a poor standard of living. Over the course of five decades, the number of people employed in agriculture and forestry diminished to 2.5% while production rose. This means that productivity expressed as output per working hour has increased by a factor of 20! Most of the people leaving the farms found better paying jobs in industry and services, some of them in Sweden, but most in Finland.

We can expect something similar in the decades to come due to AI and automation.

Productivity will increase – and that is a good thing! This will be needed to compensate for the fall in the number of working age people due to an aging population. Of course, increased productivity could lead to high unemployment and stagnant GDP. But that does not need to be the case. The example of agriculture shows that increasing productivity leads to growing GDP, new jobs (which we cannot yet identify) and a higher quality of life for most people.

Finland a Winner – why?

Why do I believe that Finland will be a winner in the AI disruption? There are a number of reasons:

  1. Finns have a comparatively high level of education and skills and we are ready to adopt new technology when we see the benefits.
  2. Cooperation between companies, universities and VTT is close, which means that new ideas flow in both directions.
  3. Regulations, public servants and government allow and even encourage trials and exploiting the opportunities provided by AI.
  4. We have a great tradition in applied and basic AI research, due to the legacy of Teuvo Kohonen and Erkki Oja.

Consultancy firm Accenture has also recognised Finland as a top country – alongside the US – in benefiting from AI. According to a recent study by the firm, Finland could increase its annual GDP growth by 2 percentage points in the long run through the widespread adoption of AI technology. Since this is cumulative growth, it would have a huge impact on our lives in the 2030s and beyond. The same study predicts an increase in cumulative productivity of 36% compared to the baseline scenario. By this measure, Finland would be the second best country, losing out to Sweden by one percentage point.

VTT’s AI recommendations

What should we do to achieve this positive situation? VTT recommends the following actions:

  1. Let’s prepare people for the coming changes and hear the opinions and expectations of citizens in an open discussion.
  2. We should focus research and development efforts on combining AI technologies with our traditional strengths. These include the process industry, energy, forest harvesters and mining machinery, ships, communication technology, medical and wellness technology, as well as forestry. AI is often introduced in a supportive role as part of a service business.
  3. When providing public support for company-driven initiatives, the level of ambition should be high and funding should be in relation to the results. This would lead to the genuine growth and creation of competitive ecosystems.
  4. Ownership of successful AI companies should be retained in Finland. As the AI business is fundamentally a software business, it is by nature scalable. Companies can grow fast and achieve huge revenues. Such gems have been sold abroad too often in the past. Insurance companies should consider their role here.
  5. Regulation and legislation should allow and incentivise the novel application of AI – as long as the safety and rights of citizens are ensured.
  6. Education is essential. We must “learn to learn” from elementary school onwards. A specific AI curriculum should be established in two or three universities and other curricula should provide a basic understanding of AI technology and its potential.

There are a number of other important aspects to AI besides those discussed here. Of course, AI technologies and techniques are a fascinating topic and ethical issues are becoming interesting due to the power and transformational nature of AI. We will address these issues in forthcoming blogs.

You can download the VTT’s Policy Brief (in Finnish) here.

Heikki Ailisto VTT

Heikki Ailisto
Research Professor
Twitter: @HeikkiAilisto