Artificial Ethics

24. October, 2017

While watching this video, I wondered: We’re using machine learning to earn money on the stock market and to make computers understand speech. Why not ethics?

Around @26:00 , Isaac talks about ways to develop AIs to control androids. He would like to use the safe approach of manual programming all the specific details to create an AI.

The “manual programming” path has been tried since 1960 and it’s now deemed a failure. The task is simply too complex. It’s like manually writing down all possible chess positions: Even if you tried, you’d run out of time. Machine learning is the way to go.

Which means we have to solve a “simple” problem: Encode the rules of ethics. That is, a machine learning algorithm must check itself (or be checked by another algorithm) against a basic set of ethical rules to determine whether “a solution” to “a problem” is “ethical” (quotes mean: “We still have to figure out exactly what that means and how to put it into code”).

Just like intelligence, ethics is somewhat of a soft and moving target. Fortunately, we have a huge body of texts (religious, laws, philosophy, polls) which a machine learning algorithm could be trained on. To test this machine, we could present it with artificial and real life incidents and see how it rules. Note: The output of this algorithm would be a number between 0 (very unethical) and 1 (very ethical). It would not spit out solutions on how to solve an ethical dilemma. It could just judge an existing solution.

It’s important that the output (judgement) is checked and the result (how good the output was) is fed back into the algorithm so it can tune itself. Both output and feedback needs to be checked for the usual problems (racism, prejudice, etc.).

Based on that, another machine learning algorithm (MLA) could then try many different solutions, present those to the ethics one, and pick the best ones. At the beginning, humans would supervise this process as well (feedback as above). Eventually, the MLA would figure out the hidden rules of good solutions.

That would eventually lead to ethical machines. Which would cause new problems: There will eventually be a machine, very impartial, that “is” more ethical than almost all humans. Whatever “is” might mean, then.

Related articles:


Your Phone Should be Your Butler

18. October, 2017

A lot of people share private details with the world without being aware of it. For example, they take nude pictures with their phones (NSA keeps a copy, just in case) or they sell the phone without wiping it properly, allowing the next owner to get a good idea who you are, or they install apps like the one from Facebook which ask “can I do whatever I want with anything I find on your phone?” and people happily click the “Yeah, whatever” button (a.k.a “Accept”).

When people use modern technology, they have a mental model. That model tells them what to expect when they do something (“press here and the screen will turn on”). It also contains other expectations that are rooted in social behavior. Like “I take good care of my phone and it will take good care of me (and my data)”.

That, when you think about it, is nonsense.

A phone is not a butler. In essence, a phone is a personal data collecting device with additional communication capabilities. But the main goal is to learn about you and then manipulate you to buy stuff. It’s about money. Companies want it, you have it, they want you to give it to them. Anything else only exists to facilitate this process. If pain would increase revenue, we’d be living in hell.

Case in point: Speech based input. When you click on a page, that doesn’t tell much about you. When you use your finger, the phone can at least feel when you’re trembling. Are you angry or enthusiastic? We’re getting there. But your voice is rich with detail about your emotional state. More data to milk to make you perfect offers which you simply don’t want to refuse.

A butler, on the other hand, has your interests in mind. They keep private information private instead of selling it to the highest bidder. They look out for you.

The source of the difference? You pay a butler. (S)he is literally working for you. On the phone, a lot of people expect the same service to happen magically and for free. Wrong planet, pals.

Wouldn’t it be great if phones were like butlers? Trustworthy, discreet and helpful instead of just trying to be helpful?

I hope we’ll see more technology like the app Nude (which hides sensitive photos on your phone).

Related:


Risks of Artificial Intelligence

10. November, 2016

There is a growing group of people arguing how AIs will one day kill us, either by loving or hating us to death. I find their arguments interesting but lacking an important factor: AI is created by (a few) humans.

That means AIs will inherit features from their creators:

  1. Humans make mistakes, so parts of the AI won’t do what they should.
  2. Each human defines “good” in a different way at a different time.
  3. The road to hell is paved with good intentions.

My addition to the discussion is thus: Even if we do everything “as right as possible”, the result will still be “surprising.”

Mistakes

Mistakes happen at all levels of software development. They can be made during the requirements phase, when the goals are set. Requirements often are vague, incomplete, missing or outright wrong.

Software developers then make mistakes, too. They misunderstand the requirements, they struggle with the programming language, their brain simply isn’t at the top of its abilities 100% of the time.

When it comes to AI, the picture gets even more muddled. Nobody knows what “AI” really is. If two people work on the same “AI” problem, their starting set of assumptions is very different.

In many cases, we use neural networks. Nobody really understands neural networks which is the key factor: They “learn” by themselves, even if we don’t know what exactly. So they come up with “solutions” without a lot of effort on the human side which is great. It “just works”. Many such projects failed because the neural networks tracks a spurious correlation – something that happens to us humans every day.

Good

What is “good“? Is it good when you add a feature to the software? When you’re really uneasy about it? When it’s immoral? Illegal? If it means keeping your job?

Is the success of a project good? What is “success”? It’s completed within time? Within budge? It’s somewhat completed at all? When the result is a rogue AI because too many corners were cut?

Unintentional Side Effects

The book “Avogadro Corp” tells the story of an AI which is created on purpose. The creator failed to take into account that he’s not alone. Soon, the AI acquired resources which it was never meant to have. People are killed, wars are prevented. Is that “success”?

Many people believe that strong leaders are “good” even when all the evidence says otherwise. They translate an insecurity into a wishful fact. If the wish of these people – often the majority – is granted, is that “good?” Is it good to allow a person to reject medicine which would save them because of personal belief? When all evidence suggests that the belief is wrong? Is it good to force happiness on people?

We want AIs to have an impact on the real world – avoid collisions with other people and cars, select the best medicine, make people spend more money on things they “need”, detect “abnormal” behavior of individuals in groups, kill enemies efficiently. Some of those goals are only “good” for a very small group of people. For me, that sounds like the first AIs won’t be created to serve humanity. The incentive just isn’t there.

Conclusion

AIs are built by flawed humans; humans who can’t even agree on a term like “good”. I feel that a lot of people trust AIs and computers because they are based on “math” and math is always perfect, right? Well, no, it’s not. In addition, the perceived perfection of math is diluted by greed, stupidity, lack of sleep and all the other human factors.

To make things worse, AIs are created to solve problems beyond the capability of humans. We use technologies to build them which we cannot understand. The goals to build AIs are driven by greed, fear, stupidity and hubris.

Looking back at history, my prediction is that the first AIs will probably be victim of the greatest mental human power: ignorance.


Technical Solutions to Amok Runs

3. August, 2016

Every now and then, an idiot realizes that his life isn’t exciting enough and decides to do something about it. Note: I apply humor to horror.

Some people (I think of them as idiots as well, just a different flavor) think that arming everyone is the best solution to this problem. Maybe these people probably never get angry.

Anyway. Here is my attempt at a solution: Data contracts.

A data contract is a contract which is attached to data.

Example: I could attach a contract to data which my cell phone produces, for example, “code looking for the signature of gunshots can access data which the microphone produces.” Similarly, I could attach “code looking symptoms of mass panic can access data from my mobile’s acceleration sensors.” And lastly, “code which detected mass panic or gunshots is allowed to access location data on my mobile.”

To build such a system, all data needs to be signed (so it can be attributed to someone) and it needs to contain the hash code of the contract. Big data services can then look up people by their signature (which would also allow to create a public / shared signature for an anonymous entity) and from there, get the data contracts.

Now that in itself doesn’t protect against abuse of data by greedy / evil corporations. The solution here is the same as in the “real” world: Auditing. People applying for access to this system need to undergo an audit where test data is fed into the system and auditors (which can be humans or bots or both) validate the operation. This results in a digital document signed by the auditors which will then allow them to access the data feeds.

This approach would then protect my privacy from people wanting my movement profiles to annoy me with adverts while safety services could still use the data to automatically detect disasters and dispatch help without me having to fumble for my phone while running for my life.

On the downside, attackers will start to shoot mobile phones.

If we look into the future, unstable people could be sentenced to share some of their data with automated systems which monitor their mental state – I’m positive that several companies are working on systems to determine the mental state of a person by looking at sensor data from their phones or fitness sensors as you read this. Of course, we’d need an improved justice system (our current one is too busy with things like patent lawsuits or copyright violations) with careful balance and checks to prevent another kind of idiot (the one which doesn’t believe in “everything has a cost”) to run amok with this (i.e. putting “unwanted” people into virtual jails).

There is a certain amount of “bad things happening” that we have to accept as inevitable. Everyone who disagrees is invited to move to North Korea where they have … ah … “solved” this already.

For everyone else, this idea has a few holes. It needs computer readable contracts, a way to negotiate contracts between computers (with and without human interaction), it needs technology for auditors where they can feed test data into complex systems and see where it goes.

I think the computer readable contracts will happen in the next few years; negotiating contracts and knowing what contracts you have is a big issue with companies. Their needs will drive this technology. Eventually, you’ll be able to set up a meeting with a lawyer who will configure a “contract matching app” your mobile. When some service wants your data, the app will automatically approve the parts of the contract which you already agree, and reject those which you’ll never accept. If the service still wants to do business with you, then you’ll get a short list of points which are undecided, yet. A few swipes later, you’ll be in business or you’ll know why not.

The test data problem can be implemented by adding new features to the big data processing frameworks. Many of these already have ways to describe data processing graphs which the framework will then turn into actual data processing. For documentation purposes, you can already examine those graphs. Adding signature tracking (when you already have to process the signatures anyway to read the data) isn’t a big deal. Auditing then means to check those signature tracks.

It’s not perfect but perfect doesn’t exist.


Paris

16. November, 2015

The foundation of civilization is the ability of the community to withstand their own death wishes and murderous instincts — André Glucksmann (source; my own translation)

There are people who will tell you that it’s a dog-eat-dog world. That’s a white lie. The building in which you sit while you read this, is the result of cooperation of hundreds of thousands, maybe even millions of people. They dug the earth for ore and cement. They build trucks to transport them. They built factories to refine them and turn them into steel and tools. The process of smelting and forging steel has been developed by thousands of people over ten thousand years. Thousands of people all over the globe worked to build the device(s) which you use to read this.

Civilization is a result of cooperation by millions of people who have never met. Cooperation is the foundation on which we all stand. No bomb can change that – unless we allow ourselves to be manipulated by people that we despise.


How Much do You Have to Hide?

16. September, 2015

When confronted with surveillance the usual reply is “nothing to hide.”

This answer is wrong. Let me tell you a story.

For over one hundred years, the city of Amsterdam had a census. They know your gender, relation ship status, number of children, parents, where you lived. All this information was used to make life better for everyone. And it worked. People were happy. The city government was efficient. It could base decisions on statistics and data instead of gut feelings. They were the first ones to use computers to efficiently store and handle the data.

May 10, 1940, the Nazis took the city. Suddenly, one bit of information – faith – decided over life and death. The Nazis took the data which had been collected and efficiently rounded up all the people they wanted to murder.

Surveillance is not about what you have to hide, it’s about how you can be hurt. It’s the question how much someone hiding in a faceless organization wants to ruin with your life.


Balancing Security

3. October, 2014

For your IT security, you want

  • Security
  • It must be cheap
  • And comfortable

Now choose at most two.

As always in life, everything has a cost. There is no cheap way to be secure which is also comfortable. Home Depot chose “cheap” and “comfort” – you’ve seen the result. Mordac would prefer “secure” and “cheap“.

Those example show why the answer probably is “secure” and “comfortable”. Which means we’re facing two problems: “cheap” is out of the question and the two contradict each other. Secure passwords are long, hard to remember, contain lots of unusual characters (uncomfortable the first time you travel to a different country – yes, people there use different keyboard layouts). Turns out there is a “cheap” part in “comfortable”.

Taking this on a social level, the price for security is freedom. To quote Benjamin Franklin: “Those who would give up essential Liberty, to purchase a little temporary Safety, deserve neither Liberty nor Safety.” I don’t know about you but I feel bad about terrorists dictating us how much of our freedom we have to give up.

In a similar fashion, you can either punish criminals or prevent future crimes but you have to choose one. We have learned through bad experience (witch hunts, flaws of the US penal system) or good (like the Norwegian system) that punishment doesn’t always help nor does it make victims happy. Which leaves us with the only conclusion: We, as a society, pay money to prevent future crimes because that’s the most reasonable thing to do.

Even if it leads to people mistakenly attribute modern penal system as “holiday camps.”