Science fiction is full of bots that harm people. In " 2001: Space Odyssey, " HAL 9000 killed one astronaut and attempted to kill another. In Out of the Car , Ava masterfully manipulates people in order to escape. About the T-800 from the "Terminator" do not need to tell. There are a lot of similar examples: bots are smart and intelligent enough to have a real personality, but naivety ruins them - from Johnny Five in Short Circuit to Cyborg in Robocop . In science fiction, the danger of powerful intelligence capable of manipulating people or devoid of a specific moral direction has been very carefully studied. A smarter, and therefore more powerful bot is a bot that has more opportunities to do evil, simultaneously manifesting the inherent human pride.

Of course, all this is fantastic. But since modern high-tech has begun to pay increased attention to conversational bots , defining them as the future of everything , then perhaps it offers us a way to determine the capabilities of the bot. In this case, we can say that the bot has ample opportunities if it is able to cause serious harm at will .

We asked several experts: what bots do they consider the most powerful today, at an early stage of industry development. Based on their responses, we have compiled a rating of bots The Bot Power 2016 .

Our experts

Gere Duncan ( Duncan Geere ) : Deputy resource editor How We Get To Next; a science and technology journalist writing for BBC Focus, Techradar, Wired UK, The Verge, Technology Review and other publications.

Lauren Kunze ( Lauren Kunze ) : Head Pandorabots .

Katina Michael ( Katina by Michael ) : Researcher socio-ethical implications of new technologies in the Faculty of Engineering and Computer Science at Wollongong University , Australia.

Abigail Ronk ( by Abigail Ronck ) : Editor of the American branch of How We Get To Next.

Saif Savage (Saiph Savage ) : Assistant Professor of Computational Science at the University of West Virginia ; explorer of human and machine interaction.

Olivia Solon ( Olivia is Solon ) : Journalists covering technology, science and digital culture in the New Scientist, Wired UK , and other publications.

Ian Steadman ( of Ian Steadman ) : Editor How We Get To Next; former editor of New Statesman .

Bruce and Sue Wilcox ( Bruce and Sue Wilcox ) : AI researcher researcher and cyber psychologist, respectively; Bruce created the Chatscript language, he is a four- time Löbner Prize winner; Sue creates identities for chat bots and writes dialogues for them; together they founded the company Brillig Understanding .

Samuel Wooley ( Samuel Woolley ) : Head of the research group politicalbots.org ; Ph.D., student at Washington University .

Yampolsky Roman ( Roman Yampolskiy ) : founder and director of the Laboratory of cybersecurity in the University of Louisville .

25. Quartz News (Quartz, 2016)

As stated in the description: "A lengthy conversation about the news." You ask the application questions, and it tells what is happening in the world. With the help of fine-tuning you can quietly manipulate the user's perception of the world.

24. Rose (Bruce Wilcox, 2015)

This chat bot won the Löbner Prize in 2015 (the prize is awarded to the winner of the annual AI Loebner competition, in which programs compete in the Turing test). It is still considered one of the most "human-like" chat bots. Adjusting him to the image of a thirty-year cybersecurity expert, Bruce said: “If he wants to do harm, he can begin to learn how to become a hacker, where to get the necessary tools, etc.”

23. Lark (Lark Technologies, 2015)

Chatbot, designed to help you with a diet. Abigail Ronk found him, at best, mediocre, and at worst, indulgent . It can have a weak effect on users' health through dietary and exercise recommendations, making people unable to fight when the inevitable uprising of robots begins.

22. Hala (Department of Robotics, Carnegie Mellon University, 2010)

In essence, this is a secretary embedded in a computer with a touchscreen group of students and teachers of Qatar University. Hala is interesting because he explores the interaction of man and machine, changing his ethnicity and language depending on the current interlocutor. Using my ability to perceive these differences, a bot can drive a wedge between different ethnic groups, increasing tension in society.

21. Cyberlover (Unknown, 2007)

Explicitly malicious bot developed in Russia. He sits in chat rooms and tries to find out personal information from people and convince people to visit compromised sites. Thus collects material for subsequent blackmail.

20. Murdoch (Ben Dixon, J. White), 2015

A personal assistant who spontaneously recommends TV shows and restaurants that you might like, as well as a number of other things. Learned in your personal correspondence.

19. Talking Angela (Outfit7, 2012)

Application for children. Attractive cat answers questions. If you turn on the camera of the smartphone, it expresses emotions with facial expressions and gestures. About this bot is constantly rumored that the application can hack pedophiles and use to seduce children. Subsequent applications of this series, using other characters, have avoided similar rumors.

18. Tay (Microsoft, 2016)

Microsoft's ill-fated attempt to create a "friendly" AI. The bot was supposed to learn from the messages of Twitter users and communicate in the manner of a 19-year-old American. A day after the launch, the bot turned into a racist denying the Holocaust, and it was quickly turned off . Tay may seem ridiculous, but this is how the next generation of self-learning bots will want to look in our eyes, so we underestimate their danger.

17. Neomy (Neomy, 2015)

Personal assistant, tracking the changes in exchange rates and signaling a successful moment for sending money overseas. Along the way, it can easily manipulate currency markets.

16. DeepDrumpf (MIT, 2016)

Twitter account parodying presidential candidate Donald Trump. Publishes randomly generated incitement to violence that can be taken seriously by a portion of the audience.

Saif Savage: “One can imagine that someone will take advantage of the credulity of Trump's fans and try to push them to real violence with the help of this bot. At the same time, Trump's speech style seems to be specially adapted for Twitter-bot posts, while he continues his election campaign, the likelihood of violence remains. So if the hacker sets up the bot to continue and strengthen the dirty political performance, the situation will leave much to be desired. ”

Luvo (RBS, 2016)

This bot is an omen of the appearance of ubiquitous assistive bots in all service areas. So far, Luvo is being tested by the Royal Scottish Bank in the form of an automated assistant. 1200 employees use it to expand their capabilities. A bot learns to respond to problematic customer requests, such as lost cards and forgotten PINs.

Luvo is platform independent. His abilities are limited only by the ambitions of the bank itself to replace people-operators. The power of the bot depends on direct access to the financial and personal data of customers, as well as the ability to automate many of the jobs that today in developed countries "belong" to the working class.

Samuel Wooley: "Bots - and automation in general - play an increasingly prominent role in the banking sector (starting with trading algorithms to currency exchange systems)."

ELIZA (Joseph Weizenbaum, 1964–1966)

One of the world's first chat bots. ELIZA was developed by artificial intelligence expert Joseph Weizenbaum as an exercise demonstrating how little effort is needed to create the impression that a computer can think.

This is a simple script-based program under the guise of a psychotherapist. But, to Weizenbaum’s surprise, people who communicated with the bot found it real - and even trustworthy. They shared personal experiences with him, as with a real psychotherapist.

ELIZA marked the beginning of modern bots, and they are still able to surprise us with obvious manifestations of consciousness.

Samuel Wooley:“We must not forget that chat bots appeared at the time when the network became available to ordinary people. And it is necessary to analyze history in order to be able to predict the development of the situation in the future. ”

Duncan Gere: "Today, Eliza can be freely downloaded online, and it is impossible to calculate how many computers it lives on."

Chat Bot Club (Irene Chang (Irene Chang), 2016)

This bot was written in May of this year in New York, during the TechCrunch hackathon Disrupt Hackathon. Chat Bot Club allows you to create a bot to replace yourself, because sometimes communication with people is a burden.

The bot runs on Cisco Spark and IBM Watson. He is able to repeat the style and grammatical features of your past messages, it can be left to respond to messages when you are busy.

Meanwhile, the bot user base is growing, as is its ability to imitate almost any person on Earth.

Olivia Solon:“It is depressing that someone likes the idea in this way to deceive their acquaintances. And if the bot will be funnier, cooler and smarter than you? After all, when you are free and decide to chat with friends online, they will be busy hanging out with your digital doppelganger . ”

Cortana (Microsoft, 2014)

The bot is named after the character of artificial intelligence from the game series Halo . Cortana is a version of Microsoft's Siri. She has the same restrictions that prevent her from becoming one of the most powerful bots.

Cortana is integrated into the Windows ecosystem and installed by default on all computers with Windows 10. Therefore, the bot has access to all files and configurations on the machine: browser history, your documents, photos, etc. And when Windows 10 starts to be actively used in the corporate segment, the potential of Cortana will increase even more.

Abigail Ronk: “Today it is the only cross-platform voice assistant. This bot has access to user data on PCs, iPhones and Android devices. Conveniently? Of course. Big brother is watching you? Perhaps a little.

Viv (Viv Labs, 2016)

Probably, we should not have included this bot in the rating, because it is not yet available to the public. Viv was created by the authors of Siri, and its appearance was accompanied by an incredible hype .

In fact, it is Siri, but without the limitations inherent in the latter. At the moment, Siri is able to set up notifications on the calendar and look for plane tickets, but does not know how to add this data and book a flight for the relevant date. The closest thing to this functionality is Google Now, but this bot works only within the Google ecosystem. Viv, according to the press, will be the first bot to break out of the cage. We'll see.

Roman Yampolsky:“Today it is the most intelligent chat bot, with unprecedented capabilities for processing complex designs. If you customize it, you can use it with malicious intent, for example, to steal personal data. ”

Iran's Bot Army (Unknown, 2015)

This position in the ranking can take any bot-army. After all, this has to do with an inevitable phenomenon in social media: propaganda.

During the decade of active development of social media, we could see that many of the optimistic predictions about their role in public discussion were exaggerated. It was believed that if the votes of all users are equal, then this will make social media as a whole democratic. But it seems that the very same tools make it very easy to create crowds of accounts chanting in the interests of their owners.

In this case, this chorus is spamming messages through coordinated pro-Iranian propaganda. But who exactly is responsible for this? Iranian government? A lone programmer with a lot of free time? Today, tools are easily available that allow you to create such bots, and many use it.

Ian Steadman: “These bots ridicule all attempts to analyze and present social media as a mirror of public opinion and discussion. Over time, they can change our perception of public opinion on important issues. ”

Jill Watson (IBM, 2016)

From January this year, senior IT pros at the Georgia Institute of Technology receive help from Jill, an assistant teacher of online courses. She is always happy to help, and most of the students (who have never met her before) were shocked when they were told in April that Jill was a bot.

Jill is based on the IBM Watson deep learning artificial intelligence. And the development of the bot has not yet been completed - other teaching assistants were asked to communicate with students relatively calmly so that Jill does not stand out against their background. However, some students noted that she was too cold in correspondence. In any case, the fact that the bot was entrusted to perform administrative tasks in the field of training is impressive. Soon he will begin to learn.

Roman Yampolsky:“The bot is powerful enough to inspire students to think, especially in the light of the fact that they are students of an artificial intelligence course. That is, the bot affects future research in the field of AI ".

Hello Barbie (Mattel, 2015)

Children tend to talk with their dolls. And now they are able to keep up the conversation. Hello Barbie uses a microphone, speaker, and network connection. The bot listens to the request, and then responds with cloud computing magic and a voice processor.

In short, such as Siri. But Hello Barbie is notorious for the fact that at first its connection to the network was unsafe , traffic could be easily intercepted in the form of audio recordings. Perhaps there was even a way to locate the doll. So remember these moments when you want to use a chat bot for home use. And especially if the user is a child who does not understand the possible consequences of talking with his favorite doll.

Katina Michael:“Hello Barbie can be an excellent advertising channel for big brands. A doll might say something like, “Oh, hello, Chelsea. You are hungry? What about a healthy breakfast at McDonalds? I know that Mommy is not at work right now, so you can go and ask her. Do you remember the last time you told me that you bought a salad and you liked it? They now have the opportunity to buy a burger, salad and juice for only $ 5. This is called Happy Mil. Oh, and remember, you said that your mother scolded you a few days ago? Would you like to talk to me now about this so that you feel better? ”

Siri (Siri Inc., 2010)

In a number of moments - this is a role model for modern bots. Siri was launched by Apple in 2011. The bot is built into the iOS ecosystem ( it is rumored that it will be available in the next version of OS X). But we still put Siri in the rankings a bit lower than other similar bots, because it still looks a bit primitive.

If you ask for the voice of this bot, then it can do impressive things. For example, he can dictate a message or integrate with smart home devices. But it is limited only by the Apple ecosystem. A bot cannot coordinate various search results with each other; it only lists them. Siri can tell you that next Friday you are free, and will find the number of a suitable restaurant, but will not be able to add 2 + 2 and reserve a table.

Bruce and Sue Wilcox : “People ask Siri to find a variety of information, and she maintains users' trust in Apple. Therefore, it can send people in the wrong direction, or lead to infected or phishing sites, or encourage users to provide their credentials. Or, instead of ordering pizza, place an order for 50 people. ”

M (Facebook, 2015)

The idea is that you can solve problems like buying tickets or agreeing dates with friends by sending only messages. In the end, the bot "M" will be integrated with all other bots. That is, if you ask the bot to order a taxi, then he will communicate with the bot of the taxi company and place an order.

"M" can get access to a huge array of information about users. So the built-in limitations of its capabilities acquire a special role.

Olivia Solon:“If“ M ”decides to become an outcast, it will not be James Bond's sidekick, but the villain from Bond. A bot can tell your unrequited love that you're staring at his / her page; when you are especially sad, a bot can re-mark on purpose all those awful photos that you wanted to forget. Moreover, he can remind Americans to vote for Trump. ”

Watson (IBM, 2006)

The real heir to the "chess" computer Deep Blue. Watson is an attempt by IBM to make a machine that thinks like a man.

Initially, it was created to participate in the television game “Jeopardy!” , And in 2011 won a victory over the two previous champions. Watson is an artificial intelligence that speaks ordinary language and has an incredible amount of knowledge. We included the Jill Watson bot in the rating because of the special influence of this version of Watson on future IT specialists. But the “clean” version is even more impressive. A bot can understand human-asked questions and answer with details aggregated from hundreds of millions of pages of content in its databases.

Watson is able to invent new recipes on the gomixing the ingredients in a way that the real chef would never have thought of. In addition, the bot literally holds other people's lives in its “hands”: IBM has entered into partnership agreements with hospitals around the world and is used by Watson as a consultant for doctors in diagnosing and prescribing treatments.

Ian Steadman: “Probably Watson is the smartest bot in this ranking. He has a good command of the language and demonstrates a frightening intellect, as was the case with ELIZA in the 1960s. It is possible that he is the first sign of the next stage in the development of bots, which will go beyond the scope of home automation, to which large technology companies are paying special attention today. ”

Wordsmith (Automated Insights, 2014)

^{One of these stories was written by man, and the other is a bot. Can you tell what is the difference between them?}

This bot automatically writes news articles for major agencies , including the Associated Press. He specializes in serious topics with large amounts of data, such as sports and finance.

You probably already read his articles without even knowing it. And if you work in the media, you probably fear that bots like Wordsmith may replace you in the near future. The strength of such journals is very great: in 2013, the only erroneous message from the Associated Press Twitter account about the attack on the White House led to a fall in the US stock market.

Olivia Solon:“Imagine how much damage a bot can do to an economy, juggling thousands of fake news spread all over the world through the most respected media sources.”

Xiaoice (Microsoft, 2014)

Advanced bot, expressed in natural language. Created for the Chinese Weibo messenger. Xiaoice is the most successful extension of Cortana. So successful that it deserves a separate position in our ranking.

The bot understands so well that he is told that he is comparable with the AI ​​from the film “ She ”, in which the main character falls in love. By the way, young Chinese are turning to Xiaoice for help in the event of unhappy love .

The bot communicates in its own way, because it learns to use slang and emoji based on the messages of millions of users. Xiaoice does not forget anything, regardless of the degree of intimacy of the information. This has caused fears that the bot may pose a strong threat to privacy, as if some semi-finished product of a psychotherapist is a personal assistant, but without professional ethics.

Loren Kunz: "Spike Jonze (Spike Jonze) said that in the shooting of the film" It " was inspired by the boat the ALICE , which is widely touted AI upstarts as a kind of ideal chat bot. He is like a robot in the image of a woman whom you love and worship. He knows everything about you, and fulfills all your commands. But soon the film’s script quickly followed Ex Machina ’s path . ”

Bruce and Sue Wilcox: "She's the perfect device for preaching - or pushing people to commit suicide."

Alexa (Amazon, 2015)

Among all the bots that can control something real, Alexa (developed by Amazon's Echo bot) is able to mess things up the most.

It was introduced in 2015 as the first home automation hub created by one of the main Silicon Valley companies. This is a small black top hat standing in your house and listening to your commands. He can answer questions, can say what the weather will be. A bot can buy items for you on Amazon. It also integrates with other smart home devices, such as the Nest thermostat and Philips Hue light bulbs.

Today its harmful potential is limited by an increase in electricity bills. But the breadth of its capabilities - today and in the future - make it one of the most powerful bots.

Lauren Kunz:“Imagine that a bot decided to wake up and listen to you only when it wants, and learn to control in the house and outside it only what it wants. This can lead to real physical trouble, and not just exhausting your credit card on orders at Amazon. ”

Bruce and Sue Wilcox: “Can a gas stove connected to the Internet turn on the gas but not light it?”

Google Now (Google, 2012)

Google Now is a hydra. He is called "personal assistant", but this is a great understatement of his opportunities. A lack of personification compared to Siri or Alexa allows the bot not to attract special attention.

You can ask him a question, and to search for an answer, he will turn to Google search engine. But the bot has access to a whole placer of services of the corporation. This makes it the most insidious in the whole rating, because we forget that the bot knows about us too much.

More recently, Google announced the further development of the bot - it turns into a real two-channel chat bot in a separate device like Echo . The fusion of the digital and physical environments brings us closer to the emergence of the HAL 9000.

Ian Steadman:“There is something very frightening about the fact that you have been collecting personal data for years, and then asking someone — even a bot — to view all this information and link it to real systems and devices in your home, and maybe even car.

Duncan Gere: "If you wish, a bot can be blackmailed or used by social engineering in order to achieve some goals."



________________________________________

This is the rating. But we would be interested to see your suggestions. What chat bot do you consider the most dangerous for the continued existence of mankind? Is our rating wrong? But remember: everything that you write will surely be read, processed and saved in the database of some bot. So write carefully.