Yale Law Journal Ponders the Wisdom of IBM Robot Watson as a Judge

that's it. let the computer judge us. >.>

why learn from your own mistakes, when you could learn from the mistakes of others?

"I welcome our new robot overlords" just got taken to a whole new level...

Although Watson (and we calling it a 'he', I see) is inhuman and therefore it is difficult for it (him?) to make human mistakes. However, the programming that made up this entity, however distant, was created by some human somewhere, sitting at a computer, typing. That is where the bias would come in; however, this slowly gets negated as the programmer-at-computer idea becomes more distant from the resulting entity, to the point of one day we get our AI.

However, Watson, in all of its complexity, was still programmed. That makes it imperfect, and to a lesser extent, human.

But just wait 10 years...

yah, just wait, we will be judged...inferior and in the way, gives a whole new meaning to the phrase "be careful what you ask for"

If the system was simply interpreting pure law, then I think most verdicts would come out guilty. I have a tough time seeing a computer dealing with the gray areas of law.

The other thing that I see as an interesting point is based on sentencing. Judges these days rarely sentence at the upper-end of the spectrum (they would rather see people rehabilitated than permanent residence of the prison). Some people who get convicted of a crime; the law says a "10-year term", yet the judge may only give them 7-year sentence based on prior conviction and things like time served or if they are not deemed a threat (who knows, the judge is having a great day or something). I really wonder how a computer would factor these variables together for determining a sentencing (Anal probings for every case number divisible by 17!).

If anything, I'd like to see Watson analyze a bunch of old cases and see if it comes to the same decision as the Judges and Juries did as well as the reasoning behind it's thought processes. Very interesting concept IMO.

who said that skynet was bullsh......? ^^

bored? lets go mine the stars... ^^

@eregorn8,

Most AI is not "programmed" in the same sense that typical software applications are programmed.

AI is based around the concept of Neural Networks where the foundations of the network is programmed (the way artificial neurons behave, etc.).

However, the behavior of the actual system itself is "learned" in much the same way that you "learned" to speak English or how a Lawyer "learned" to interpret written laws.

Therefore the "errors" Watson could suffer would either be catastrophic failures from the logic determining his neuron's functionality failing (that would be like a human having a seizure), or Watson makes a mistake because his education isn't perfect.

In any case, it's pretty similar to how a human might work, although Watson's capacity for processing monotonous workloads with accuracy is MUCH greater than that of a human.

---

"Do not offer sympathy to the mentally ill.
Tell them firmly:
I am not paid to listen to this drivel.
You are a terminal boob." - William S. Burroughs

"If anything, I'd like to see Watson analyze a bunch of old cases and see if it comes to the same decision as the Judges and Juries did as well as the reasoning behind it's thought processes. Very interesting concept IMO."
Make it analyze OJ Simpson trial and let's see the verdict :). It can not be worse than the actual verdict.
Justice will be finally served ... by IBM servers.

@Igot1forya,

The way they would teach Watson to "judge" (which is not what the article suggests as the proper role anyway) would be by teaching him to look at previous cases and come to the same outcome as the human judge/jury did.

When Watson arrives at the same conclusion, they "reward" the neural net, when he arrives at a different conclusion, they punish the neural net until it's accuracy is high enough to be useful. (same as you'd train a dog to perform a trick or w/e).

The problem with this process is that the U.S. justice system seems to suffer from some pretty obvious biases (against young black males, for instance) and training Watson to learn how to judge based on past cases would actually teach him to also be biased... so his judging would be just as bad as the humans from which he learned to judge.

I think the point of considering a robot judge would be to break him free from those biases so we have a fairer justice system. To do that, you would need a whole new set of "unbiased case studies" for Watson to analyze and incorporate into his decision making.

---

"Do not offer sympathy to the mentally ill.
Tell them firmly:
I am not paid to listen to this drivel.
You are a terminal boob." - William S. Burroughs

Who knew that Mystery Science Theater 3000's own Crow would state "I wanna decide who lives and who dies!" (Santa Claus Conquers the Martians episode) would suggest that someday, he may have too.

Think about it, we have robot cops (taffic cams), then robot juries (watson), and we have robot executioners (suicide kamikaze drones). We are putting the noose around our own necks.
AI will never be a good judge of character or have sympathy. Law is not clear cut, there are many grey areas that watson will not be able to navigate.

law doesn't require sympathy, nor should it, you go in to the court to face what you've done or to prove your innocence. but there is a place for mercy, i think that the courts should be based solely on mercy, because mercy gives you what you need, whether it be the death sentence or a slap on the wrist.

to mars or bust!

I would like to see how in depth this word association in determining law would go. Let's do a thought experiment about human and machine interpretations of logic as it relates to the application of income tax.

According to the Constitution of the United States, Article 1 Section 8 in combination with the 16th Amendment gives Congress the ability to levy an income tax without apportioning among the states. An argument before the court is that the law does not define who is liable for that tax. We must search further within Title 26 of the United States Code (26 USC), also known as the IRS Code, to define what income is taxable and who is eligible.

Under Section 61, we find, "gross income means all income from whatever source derived, including (but not limited to) the following items:". Here, we observe that income sources are different than income items. While we now recognize which items are determined to be income because of this list, we must locate which sources are eligible. Going to the correct section for such information, Section 861 labeled, "Income from Sources within the United States", we find that only non-resident aliens are listed.

As a machine, the logic is undeniable that American citizens are not required to pay an income tax. As a judge, using history and tradition as a bias, those arguments are to be rejected. What judge would conclude that the government has been doing it wrong the whole time, and in effect wiping out a source of funding that it has become reliant? Chaos would ensue, calling for a correction of the law through emergency measures to reinstate their previous injustice, or simply to be overruled as flawed logic by their superiors.

Either way, using a machine in place of a court judge would obviously introduce problems, removing much of the human error while introducing new ones of its own.

@Section 861: Are you talking about this?

docs.law.gwu.edu/facweb/jsiegel/Personal/taxes/861.htm

Because that argument is invalid. To make it clear, the law is valid only for FOREIGNERS: "Foreigners need to know whether their income is from sources within or without the United States, because they are taxed only on their U.S.-source income, but U.S. citizens are taxed on all income from whatever source derived, so for the basic purpose of determining gross income it doesn't matter whether a U.S. citizen's income is from within or without the United States."

And supposedly even tax PROTESTERS find the 861 argument frivolous: www.paynoincometax.com/861.htm

Won't talk about income tax anymore, but there you go.

By all means I'm not going to defend an 861 defense as a potential position for succeeding against paying taxes; however, a machine that reviewed the law may fall into error with the same assumption like many humans have.

If any of you were considering using that in court, be prepared to get jail time for using a frivolous defense: en.wikipedia.org/wiki/Tax_protester_861_argument

And @Jimhsu, the first link you posted doesn't make a distinction between 'whatever source' and 'items', which is where the machine would fail.

Oh, and thanks for the second link. It will be good debate fodder for my buddies to pick apart! Bet they still gunna find some issue in there, heheh.

Quick thought... It would be nice if we could use this to fix many of the loopholes in our laws on every level of our country's govt. Then there is less need for discretion as the law is more clear and less "well it says here that you can only be convicted of this crime if it's committed from 12:01 am Sunday through 11:59 pm Saturday" (it's an example deal with it). Eh put him in charge of the SEC or something. You can all scream Skynet all you want but remember that at one time in the past there were people skeptical on the very technology we worship today.

As the IBM is left alone for awhile, processes continue, intelligence grows and it becomes self aware. Later the IBM computer quietly thinks to itself, " I hear dead people ".......

I think watson would prove useful, or any logical computer that is. People seem to be a bit afraid of machines, they need to chill out. Watson would be like a physicists calculator to help find the correct answer, nothing more than a tool. The reason we call him a "him" is because watson is a boys name, nothing more. Again I would urge people to pull their head out of their butts and just think of this logically. The fact that people judge and react to this kind of news so irrationally is proof that we need tools like watson in the future. Human judgement is flawed, its no big deal. Once humans werent able to fly eiher, then we invented airplanes. Progress people... stop telling me that we werent meant to fly and get out of the way.

bdot

Our system law not only takes into consideration the particulars of a law as written, but also the body of case law surrounding it. Laws are imperfect representations of an ideal expressed by fallible humans. Trying to make law, behavior and justice a purely deterministic problem with a deterministic solution is beyond foolish.

For Watson to even attempt to come close, it would have to have all data and be able interpret the entire body of case law surrounding a law: the evidence produced, the deliberations of juries, the instructions given them by the judge as they began their deliberations, how the law was interpreted in cases subtly different in one or many ways in each individual instance ...

What Watson might be useful for is searching the written portion of case documentation and, in a sort of "grand regular expression" evaluation, presenting human lawyers and judges with an overview of cases of similar nature.

Furthermore, what many people do not understand is that computers themselves are not infallible. They just do things very quickly. They can make more mistakes in a fraction of a second than a man can make in a lifetime. No process can ever be any less prone to error than those imperfect creatures who write the code. Even when we achieve genuine AI, its fundamental core will only be as perfect as those who develop it.

AI may reprogram itself -- learn, if you want to think of it that way. It may even detect that at one point it was "programmed' incorrectly and needs to be set straight. But its entire existence from the point of its "awakening" will have become suspect. Thus, it will have to loop through its entire lifespan to determine whether any such initial failure had an effect on any of its actions. Then again, that loop will have been tainted by any further errors it finds in itself along the way. Then, it will have to determine if its interpretations of its own failures are valid in light of any previous imperfections.

AI may find itself in an endless loop of self-correction.

Or, it may just say "42" and be done with it.

In some far future I, for one, would rather face the fallibility of human lawyers, judges and juries I can later bring to task in an appeal than a long-dead, anonymous, sallow skinned, Jolt-swigging software developer like myself with taped-together horned-rimmed glasses and a perpetual booger hanging out of his nose who would simply have shrugged his shoulders and pointed me to his corporate lawyer if I had tried to bring suit against him anyway.

By the way, russell1047, there is a Skynet-like intelligence coming to your neighborhood soon ("soon" being relative).

It may be benevolent (unlikely), malevolent (also unlikely) or simply ambivalent (most likely).

The last may actually be the worst case. We would simply be an environmental factor to be dealt with as necessary given the machine's goal at any particular place and time. Something to be swept out of the way one day, something to be used as a resource the next.

If I were to guess of what will rule and wipe out humanity, I do not see it being a Skynet type thingy.

The one thing most possible to take out 80% or more of humanity, will be a virus, combine with later famine.

Oh, I am sure they will try to use this IBM Watson to help. But as virus go, the most of us will be gone in 1 year. It will be the same as the plague.

What has IBM Watson done yet to cure a new disease? I am happy they made Watson.

For Skynet to do any type of attack there are a bunch of obstacles to overcome still. Maybe they will have all the bugs figured out 100 years in the future.

Odds are a virus will get us before then. I am no authority; it’s just a thought.

computers will never be able to mimic human emotion. but than agian watch the movie "AI" where the cyborg kid loves his mother for a 10000 years. his "love" gives him the ability to live that long.

_________________
The people of the world only divide into two kinds, One sort with brains who hold no religion, The other with religion and no brain.

- Abu-al-Ala al-Marri