Tag Archives: AI

04.09.21
by

AI is good – or is it?

A computer only does what it is told. Provided we are diligent in our programming and keep track of our code before the algorithms run wild, articifical intelligence and machine learning liberate us and make our lives more efficient. When we coexist, decisions made and carried out based on algorithms are as liberating as the invention of the washing machine.

In his book Människor och AI (Humans and AI) written together with co-author Professor Jonas Stier, Swedish technologist Daniel Akenine looks at a multitude of scenarios where using AI can benefit and enhance human existence. And a multitude of scenarios where it cannot – or should not. A quick review of his discourse:

AssumptionPositive PotentialNegative Potential
AI is good for humansAI’s maximize freedom and give us more time to do the things we want to do.AI’s maximize profit and productivity whilst ignoring the risk of creating chasms in society or human integrity.
AI is fairAI’s do not discriminate based on gender, sexual preferences or ethniticity.AI’s mirror or reenforce inequality in our society by using inappropriate source of data.
AI is secureAI-based systems in warfare are subject to limitation. Civilian AI-systems are built with well tested security standards and robust security protocols.Cheap AI-based systems maximize casualties when these systems are used in conflict. Civilian AI-based technology can harm people due to bad objectives or faulty programming.
AI is transparentWe understand how algorithms make decisions and when they have difficulties making the right decision.AI can be compared to a black box where the results can neither be explained nor questioned.
AI is responsibleEverybody has a clear responsiblity for the results achieved from the algorithms they build. From developer to user.Nobody takes responsibility for the damage an AI-system may cause.
Daniel Akenine is a frequent lecturer on all aspects of AI

In the book, subject matter experts consider these points from their field of expertise. Daniel Akenine’s purpose, as he clearly states, is to create a more nuanced view of what an AI and its smart algorithms can and cannot do and how this will impact human existence in the near future. Topics like AI-supported judicial systems, the future of urban development, taxation, conflicts and warfare, human integrity.

Let me pick just one for a further deep dive

How do we prepare for the risks?

Contributor Åsa Schwarz, Sweden Security Profile of the Year 2017, thinks as I do that algorithms as such cannot be trusted blindly: Open your eyes – your AI is biased. On the other hand, there is not yet reason to fear a Terminator/Skynet or Matrix scenario where machines take over and human kind is in danger of being eliminated.

But if we circumvent the human influence which hopefully includes an ethical starting point (AI is responsible in Daniel Akenine’s chart), there are bad decisions being made today using AI which lead to severe consequences for people and nations. If you feel inspired, the recent Netflix release Coded Bias helps you understand what I mean with consequences.

Just consider something simple like using AI to make selection of job candidates more efficient. By ticking all the boxes that were pre-determined and programmed, the system will not see the potential of candidates who do not fit but may add unique experiences and creativity which the recruiting company would need to grow and survive. Uniformity and stagnation led to the downfall of the Roman Empire. You need barbarian forces who are not following the rule of “this is the way” to continue to evolve.

Åsa Schwarz, Head of Business Development at Knowit Cybersecurity & Law. 

An AI can evolve based on the construction of its algorithms and make decisions that impact its actions. And there is both the malicious and the accidental consequence of AI to consider. Accidental negative consequences may be prevented but it requires the imagination of the programmer to provide the AI with the potential risk assessment: The AI has no imagination.

If you program an AI say to remove rubbish from a predefined surface (Åsa Schwarz uses the example of the Stockholm Central Train Station), it will do just that. Remove what the coding identifies as rubbish = not belonging there from the area. Now imagine you have designed a self-programming AI to help it continue to become increasingly efficient. And it may take that further instruction to remove the cause of the rubbish – i.e. remove the humans. Boom.

Malicious intent is everywhere, eg. in the physical sense through the use of drone warfare. But much worse – as an ever growing threat to us all – in Cybersecurity. Tom Leighton, Founder and CEO of Akamai Technologies, mentions during TechBBQ (2019) how a concerted attack on critical online systems can paralyze entire nations: “You can disconnect most countries now from the rest of the internet now through a coordinated Denial-of-Service attack…”, he says in the video interview.

If humans want to safely reap the obvious benefits of integrating AI even further into every aspect of society, we must understand who owns the system. We do.

So we have to act responsibly, and focus on developing concepts for the safe and secure architecture and design of AI driven processes with built-in failsafes and barriers. Microsoft as one example provides a comprehensive framework called the Security Development Lifecycle . Another important aspect is the sophisticated support for incident handling. And my favourite topic: Complete documentation so that you are able to trace and track the problem to fix it.

Lawmakers can only do so much – the real power of artificial intelligence lies in the hands of the developers of the systems.

AI’s have no ethical subroutine – you do.

03.04.18
by

Open your eyes – your AI is biased

Computations have no ethical subroutine. And understanding bias in AI is an important eye opener. Building an AI-facilitated future without properly understanding the algorithms behind the conclusions and actions is leading us to into unexpected pitfalls.

We are all very excited about machine learning and AIs. We see them as the ultimate way of automating daily life from driverless cars to personal health and medical diagnostics. But garbage in = garbage out. And to eliminate the garbage we need to be able to identify it. Long after our little helper has started working.

The main reason we need to watch out is that AI algorithms are not necessarily retraceable and retrackable. Not even the programmer understands it fully once the machine starts accumulating and filtering data. Despite its ability to learn it can only conclude based on the original assumptions built into the underlying algorithms.

vlcsnap-2011-01-17-16h21m27s224-1024x422

Looking for the ultimate answer

Whenever we rely on algorithms to make decisions – or at least recommendations – it is because we seek a simple answer to a complex question.

If the collection of big data for data driven decision making is used to create simple answers to complex questions, the complexity is solved through algorithms that in effect filter and collate based on what the human programmer considered applicable. And it concerns us more than you would expect. I recently read that the AI concept is being used within the US judicial system: Judges rely on the AI’s suggestion on whether an inmate should be granted parole based on assumptions of future behavior of set individual after release. In isolation this would seem like a statistically viable method, as there will be vast amount of available data to substantiate the conclusion.

But if the original algorithms input by a human were in fact influenced by bias such as race, name, gender, age etc., are the conclusions any better than the answer 42?

When Douglas Adams in his science-fiction Hitchiker’s Guide to the Galaxy series introduced Deep Thought, the biggest  computer ever built in the history of men and mice, the builders asked for the answer – and added that they wanted it to be nice and simple. So after millions of years Deep Thought concluded that the answer to life the universe and everything was 42. But by now, this insight was useless because nobody really understood the question.

If we see AI and machine learning as the ultimate answer to complex scenarios, then we must be able to go back to the original question in order to be able to process the answer. Not just to understand but to analyse and apply what the computer is missing – the ethical subroutine.

What will the AI choose in a no-win scenario?

One of the hot topics in the current discussion around self driving cars is whether the AI would make proper ethical decisions in a no-win scenario. Should it risk the life of the passenger by veering off the street and over a cliff to avoid running over another individual in in the street? The decision would be entirely based on the original algorithms which overtime have become inscrutable even for the engineers themselves.

Of course, this is a simplified example. An AI, as opposed to a human behind the wheel, would be able to process more details regarding the potential outcome of either option. What would the statistical probability of successfully avoiding hitting the person on the street be when taking into account elements such as speed, space available without going over the cliff, the chance of the person acknowledging the danger and moving out-of-the-way in time before the collision etc.

self-driving-car-drive-ix-625-u

(Image from Nvidia Marketing Material)

But the self preservation instinct of a human being behind the wheel would most likely lead to the obvious conclusion: Hitting the person is preferable to dying by plunging over the cliff! Would the original programmer not have input exactly this type of bias?

What I believe Douglas Adams was getting at with the magic number 42 was that there is no simple answer to complex questions. If as indicated above the AI is victim to its own programming when making complex decisions or recommendations, then as a tool we must make it as transparent and thereby manageable as any tool developed by humans since the invention of the wheel.

MIT Technology Review addressed this in detail in the article published by Will Knight in April 2017  The Dark Secret at the Heart of AI 

“No one really knows how the most advanced algorithms do what they do. That could be a problem.”

willKnightHe goes on to explain that while mathematical models are being used to make life changing decisions such as who gets parole, who gets a loan in the bank, or who gets hired for a job, it remains possible to understand the reasoning. But when it comes to what Knight calls Deep Learning or machine learning, the complexity increases and the continuosly evolving program eventually becomes impossible to backtrack even for the engineer who built it.

Despite the inscrutable nature of the mechanisms that lead to the decisions made by the AI, we are all too happy to plunge in with our eyes closed.

Later the same year another MIT Technology Review article explores the results of a study of the algorithms behind COMPAS (Inspecting Algorithms for Bias ) COMPAS is a risk assessment software which is being used to forecast which criminals are most likely to reoffend.

Without going into detail – I highly recommend you read the article – the conclusion was that there was a clear bias towards blacks. The conclusions later turned out to be incorrect assumptions: Blacks were expected to more frequently reoffend, but in reality did not. And vice versa for the white released prisoners.

The author of the article, German journalist Matthias Spielkamp, is one of the founders of the non-profit AlgorithmWatch which has taken up the mission to watch and explain the effects of algorithmic decision making processes on human behaviour and to point out ethical conflicts.

Spielkamp

Mattias Spielkamp, Founder of AlgorithmWatch

The proverbial top of the iceberg

Even strong advocates of applying artifical intelligence/cognitive intelligence and machine-learning (deep learning) to everyday life applications, such as IBM with its Watson project, are aware of this threat and use strong words such as mitigation to explain how this potential outcome of widespread use of the technology can be handled better.

In a very recent article published February 2018 entitled  Mitigating Bias in AI models , Ruchir Purri, Chief Architect and IBM Fellow, IBM Watson and Cloud Platform stresses that “AI systems are only as effective as the data they are trained on. Bad training data can lead to higher error rates and biased decision making, even when the underlying model is sound… Continually striving to identify and mitigate bias is absolutely essential to building trust and ensuring that these transformative technologies will have a net positive impact on society.”

IBM is undertaking a long range of measures to minimize bias but this is only addressing the top of the iceberg. The real challenge is that we are increasingly dehumanizing complex decisions by relying on algorithms that are too clever for their own good.

Actually – all of this isn’t exactly news.

More than 20 years ago, human bias was already identifed as an important aspect of computer programming

“As early as 1996, Batya Friedman and Helen Nissenbaum developed a typology of bias in computer systems that described the various ways human bias can be built into machine processes: “Bias can enter a [computer] system either through the explicit and conscious efforts of individuals or institutions, or implicitly and unconsciously, even in spite of the best of intentions”.  (Source:  Ethics and Algorithmic Processes for decision making and decision support )

11.22.16
by

Beating the Kobayashi Maru – or the human vs machine experiment with Watson

“I don’t believe in no-win scenarios.” (James T. Kirk, Starship Captain)

When you are a strong believer in datadriven decision making, building strategies on real insights, and always sticking to facts rather than fiction – it’s a hard blow when one of the world’s leading artificial intelligence systems tells you, that you are not a nice person. It’s based on data – so it’s a fact.

Many industry leaders have evangelists who are excellent presenters and subject matter experts. It’s always a privilege when you get a chance to interview an evangelist. I met IBM’s Rashik Parmar, Watson evangelist, at IPExpo Nordic a few weeks ago.

WatsonDescription.png

Source: IBM

There is so much potential for big data analysis and the learnings and insights we gain, from combining the many available sources of accssible data to draw new conclusions and find answers. That’s basically what Watson does. And then makes the logical connections. Simply put.

IBM developed a small demo engine that would analyse your Twitter personality and generate those awesome charts we all love; and few of us know how to interpret. It was reassuring to see what a nice guy President Obama is on Twitter. And my friend, Rashik, had a similar profile – so all good.

Potus.png

Source: IBM’s demo app TweetMeWatson

Lucky for me, we couldn’t make it work for my profile until I got home. When I ran it, I found out I was

“Inconsiderate and a bit shrewd”

I will spare you the rest. Either I am very delusional about how I express myself, or there was something fishy going on here. But it’s based on data! It has to be true!

Before digging a hole in the garden to hide from the world – or the equivalent: deleting my Twitter account – I decided to think it through. What was Watson picking up on, what were the actual parameters used?

The Big Five (FFM) Personality Traits

Watson is grouping our personalities according to the Five Factor Model (FFM) Wikipedia explains:

The Big Five personality traits, also known as the five factor model (FFM), is a model based on common language descriptors of personality (lexical hypothesis). These descriptors are grouped together using a statistical technique called factor analysis (i.e. this model is not based on experiments).

This widely examined theory suggests five broad dimensions used by some psychologists to describe the human personality and psyche.[1][2] The five factors have been defined as openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism, often listed under the acronyms OCEAN or CANOE. Beneath each proposed global factor, a number of correlated and more specific primary factors are claimed. For example, extraversion is said to include such related qualities as gregariousness, assertiveness, excitement seeking, warmth, activity, and positive emotions.

220px-francis_galton_1850s It all sounds very reassuring, the term “Lexical Hypothesis” makes sense –  it was analysing words. This is a principle which was developed by British and German psychologists to identify a personality characteristic. It was used to determine risk of mental illness or criminal behaviour. Invented in 1884, by the way, by Sir Fancis Galton – a stern looking fellow.

But something as elusive and intangible as the human mind is so very hard to classify and illustrate in data points and charts. By creating a lexicon of words and adjectives that at the time were considered to be indicators for certain behaviours, they provided a tool to build profiles – and categorise people based by their choice of words.

Note that the method has also received a lot of criticism – many of them quite reassuring when you are on the receiving end of this exercise. Read more here. 

Phew – that means I can still be a nice person, just not when I tweet. Or speak.

It seemed safe to climb back out of the hole in the garden and meet the world face on. But knowing now what triggered my unpleasant profile, I decided to challenge Watson to a duel.

A duelling experiment

kobayashi-maru-02

@echrexperiment is the experimental Twitter profile where tweets were worded more carefully, where people and followers were thanked and nothing bad was happening in the world. No politics, no injustice, no gender inequality, no discrimination. And lots of cats.

exhrexperiment.png

 

After three weeks, I was a much nicer person. The traits that I seem to be exploiting negatively in my original profile are now contributing to a positive image.

Suddenly, uncompromising was a good thing.

Spock.png

“In academic vernacular, you cheated”

Like Captain James T. Kirk in Star Trek challenging Mr. Spock’s designed program, I cheated to win.

Most of my tweets were carefully drafted using positive semantics but remaining true to my usual topics of interest. I was not trying to be someone else, just focusing on being nice. Here’s a list of the parameters I introduced to make Watson love me more:

  • Following back – anyone who followed me, unless an obvious business account or egghead, was followed back as soon as I spotted them.
  • #FF – sometimes I used the FollowFriday hashtag to thank select people. It generated some nice interactions even between those mentioned, so I grouped them into categories – e.g. Danes, analysts, etc.
  • I thanked, and loved, and “awesome’d” and “great’ed” a lot.
  • Sharing – giving credit, not taking it. I always mentioned the source or the account where I had picked up a link.
  • Sharing the love – retweets were focused on positive news, positive sentiments and uplifting current events. I also checked the wording of the original tweet before RT’ing to avoid contamination of my positivity.
  • Getting personal – my personality and emotions were conveyed more by sharing private interests such as books, cats, travel and science fiction.
  • Language Disclaimer – all of the above choices were based on my non-native perception of the English language, and may have been different from Webster’s Dictionary which is the basic semantic interpreter used in lexical hypothesis.

What I didn’t do

Humour doesn’t travel well, so any jokes, irony, satire and cartoons were not part of echrexperiment. I may have gotten carried away occasionally, but consciously tried to avoid it.

Politics are a powerful emotional trigger, so I avoided RT’ing or engaging in conversations with political statements. That wasn’t the mission.

Automation is a powerful tool to increase the quantity of your social media posts, but with automation things like timing and engagement suffer. Sometimes, due to other news, automation may even lead to displaying insensitivity.

Automatic response is a convenient way to further promote your services and invite people to connect. But it just isn’t personal. Despite all these lovely people addressing me by name. I did not send messages to thank people for the follow, but I checked their profile and retweeted where I could to show my appreciation.

What Watson had to say about @echrexperiment

The app itself produces a lot of detail as you can see from above. Below I grouped the result into more familiar charts to share some highlights. To make sure I picked a really nice person as control, I chose President Obama’s Twitter @potus. But please remember – it’s probably mostly his staff tweeting. And they seem to have done an excellent job.

watson1

Obama – it appears – is very agreeable on Twitter, and my experimental lovely/nicey/catsy account matches this impression very nicely. We are both very open, although I am lagging on conscientousness, but hey –  I am not the President.

Digging deeper into selected parameters, revealed some interesting characteristics related to being a President or just trying to be a nice person.

We can all agree that values should be an important parameter if you are President of the United States. Strangely enough Obama wasn’t all that keen on change, and more inclined to be conservative. For self enhancement … we have identified the villain – the one parameter that makes my original Twitter account so repugnant. I leave the graph to stand on its own.

Meanwhile, President Obama scored a resounding Zero on self-enhancement – but he made it to the top already.

watson2

President Obama’s most distinguishing need is the need for structure. Love – it seems – he gets a plenty.

On the other hand, my original self seems to have enough structure in her life.

Watson3.png

But here’s the greatest insight from this entire exercise – other than confirming that it is possible to change who you are, or rather how you are perceived:

When it comes to curiosity, all you need to do is be a positive tweeter and include lots of cats.

 

IMG_0042.JPG