The Singulairty is nearer than you think...

[Mallette]

I think Marty is correct that Watson's big advantage is knowing exactly when to hit the button.

I agreed. But on seeing it last night I noted that Watson had the answer on screen several time but got beat to the buzzer by whathishumanname on the right. I can't explain this. Very strange...

One correction, Dave: Watson gets the answer in text form, not from hearing Alex speak or by reading it from the board.

You are correct. Got over enthused and didn't really understand the version they are using on the show. They are working with direct interpretation of speech but elected to go with this as it isn't quite up to the rest of the program yet. It was felt it really didn't change the challenge.

However, parsing the sentence structure is still an extraordinary feat. When I get so much as a space or "." in the wrong place, my code just goes "errrrrrp" and stops.

Dave

[Mallette]

This is absolutely the correct use for a Watson type machine. Regurgitating what is already known, appying rapid-fire statistical analysis to it, and making this kind of rational prediction, like a diagnosis.

But that's mundane computer applications.

Good Lord, man! Please listen when I say stay in your field. Have you EVER gotten a computer diagnosis??? NO. Every attempt has failed because it is not a calculation. It's the most dreadful form of fuzzy logic. Why do you think House is so popular?

This is absolutely the correct use for a Watson type machine.

Since you made this statement as a clear statement of your position, I am going to suggest that in spite of your protestations you actually DO understand the signifcance of this code.

The weather chaos models are childs play compared to accurate medical diagnostics.

Dave

[thebes]

Well it should have been patterned after the real fictional Dr. Watson!

I do hope it knows, and has been programmed with, Asimov's Three Laws of Robotics. Don't want it getting carried away.

They are:

1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2. A robot must obey any orders given to it by human beings, except where such orders would conflict with the First Law.
3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

[Mallette]

I disagree. Of 100 people coming into the doctor's office, 95 will have mundane conditions that are rather easily diagnosed by humans, or computer softwares after half a dozen common tests.

Perhaps you are correct...but it isn't my experience and given some health issues in my family have continuously dealt with two of the highest rated children's hospitals on the planet. I've had discussions about this issue with them and they all said computers were neither useful nor used for anything but looking up obscure information for a human to parse.

The last serious attempt was about 10 years ago and was a failure.

95 will have mundane conditions that are rather easily diagnosed by humans

Would you fly an airline with a 95% reliability rating???? That's HORRIBLE for medicine. Further, I think you are being rather kind. Out of 10 of my close relatives 3 were killed by doctors who misdiagnosed and didn't correct until it was too late.

Further, after the fact, even a layman could easilty see their errors.

Watson and its successors are really going to improve that. If I were you, I'd hope Dave is right and Mark is wrong.

Dave

[Mallette]

However, when it comes to self aware machines, how would we be sure they would follow any human wishes whatsoever?

It's called hard coding. Basically code that inhibits that is read only and has no inbound datalinks of any kind. Like what humans have to keep us from eating our children...though an even harder inhibition. At the moment, it's not concievable for a computer to get around it. However, when they get REALLY smart...it may be problematic.

What I hope is that the future designers don't just say "Well, I think it will be OK."

Dave

[Mallette]

Approximate 195,000 people die in hospitals each year from MEDICAL ERRORS. Compare that to 898 airplane deaths last year.

Personally, I think the "die in hospitals" number is WAY low and all deaths from incorrect diagnosis don't happen in hospitals. I am a conservative. I believe that all accidents and errors are preventable. Only acts of God (legal sense, Mark) and negligence are not. The airline industry is computer friendly and that is why the single most inherently dangerous thing humans routinely participate in has become incredibly safe and almost to the "acts of God and negligence" level while medicine, the single most critical science to each of us is only 95% successful at diagnosis. I suspect the airline success rate extends out to a number of decimal places beyond 99, but won't hazard a guess at how many.

Watson suggests that diagnosis rate may be about to improve dramatically. Only time will tell. In my case, I sincerely WANT to be right on this...but it is not desire that drives my analysis.

Dave

[Islander]

Doesn't common sense tell you we ought to be preventing superior intelligence from emerging, not encouraging it?

This issue was brought up in William Gibson's 1984 novel Neuromancer, in which there were Turing Police, whose job was to ensure that AIs did not exceed their strict built-in limitations.

[Islander]

The belief that all that is needed is a certain level of computer power most likely misunderstands the problem.

For one whose said so much about it, my impression is that you really have no idea of the power of evolution.

Dave

I think you're using a very loose definition of evolution, Dave. Humans building a series of better and better machines is not any threat to humans, as the machines cannot exceed their own limits, and I would not call it evolution in a strict sense.

For me, true evolution is when an entity, whether biological or not, produces its own improved successors/descendants. Once machines gain the ability to reproduce and then reproduce in improved forms or models, for me that would be the singularity, the history-changing moment.

[Islander]

Meanwhile, in Japan, there's a robot marathon planned for later this month:

The “Robo Mara Full” race kicks off Feb 24 and is open only to androids with two legs. The robots must complete 422 laps around a 100-meter indoor racetrack to cover 42 kilometers. Survivors of the nonstop race—except for battery changes and repairs—are expected to hit the finish line on Feb 27, when their human counterparts run in the popular Tokyo marathon.


With those speeds (3 days to complete 42 kilometres), we're not about to get overrun by robots this year, at least. It's hard to tell in the photo, but I'm guessing these robots are much smaller than human size.

The news item: http://www.japantoday.com/category/technology/view/androids-to-run-in-worlds-1st-robot-marathon

[odysseyrevolver]

[merkin]

An area that Watson-like software could excel immediately is pharmacological drug interactions. It is not uncommon now for people to be on 6 or 8 or 10 different powerful medications. The ability of a human being to sift through all the possible interactions is limited. It is a perfect application for Watson, but again, it would only make a dent if there was wide or near universal access. My local pharmacy isn't going to buy a Watson.

Your pharmacy probably already has this. Almost all computer filling programs scan for drug interactions if you get all your prescriptions filled there. Also there are online services that will look up occurances of drug interactions and adverse reactions. The one we use is clinical pharmacology. we just type in the drugs in a list and it scans them for any interactions. It uses a drug interaction data base from NIH which all JCAH approved hospitals supply the data to. It then reports them to you with severity levels so you can decide if it is significant or not.

[Islander]

For me, true evolution is when an entity, whether biological or not, produces its own improved successors/descendants.

I agree on that point. Without self replication I don't think we can apply the theories of evolution. Right now it appears that the first self replicators will come out of nanotechnology. Sadly, these voracious replicators may have less intelligence than a cockroach. But, being self replicating, they might be subject to evolutionary forces.

Which brings us to the "grey goo" scenario, where out-of-control self-replicating nanomechanisms devour everything. We've already had GM crops spreading beyond the fields where they were planted, and they're big enough to see and handle. With nanotech, the machines are invisible until they're present in millions or billions, so they'd be difficult to contain or eradicate.

http://en.wikipedia.org/wiki/Grey_goo

So far, the technology has not reached the point of being capable of running away like a plague, but twenty or thirty years from now, who knows? Tiny machines with no more intelligence than a virus could be dangerous in sufficient numbers.

[Mallette]

My argument is that this has just about NOTHING to do with an emminent "singularity" - which was the title of the thread.

Part of the problem of our communication is you read way too much into non-specific sentences. Look at the title. "emminent" or anything like it is nowhere to be found. I think my estimation that the vast majority of readers think Kurzweill was way over optimistic...even though most have a hard time attacking his logic. However, most, including me, do...though it's just because it's so overwhelming that my non-objective skeptic takes over.

All I said was that it's nearer than we THINK...

And further, it's just a title and the real subject is the extraordinary redirection these guys have given to a rather moribund field. Real potential in this direction and the growth will be phenomenal.

Dave

[Mallette]

Well, it's grudging, but you are thinking in the right directions.

You mentioned "House" and "patients lie." Dead on. That is what makes the Watson programming the future. Watson works like House thinks...that is, often House takes all the available data and synthesizes the missing part, like "...you went to Katawhatsisstan and visited the unga berry farms where you engaged in an act of bestiality with the unga sheep. Didn't tell your wife about that, I presume?"

Big Blue couldn't do anything like that and it is essential to diagnostics...hence, the long run of House.

Second, if Watson-like computers were made into a cheap "utility" that could be accessed by any doctor anywhere, it would begin to learn as it collected the data.

Correct on both counts. Further, as opposed to the human diagnostic establishment, this learning will take place in a continually increasing speed curve until 100% of the availbable research is in use and it begins to hypothesize...via data synthesis...entirely new possibilities. Now, until it reaches self awareness, these synthesized bits will be the domain of human researchers to figure out whether they are valid and represent entirely new knowledge or not.

Very exciting, at least to me. I fully expect to see Watson spin offs working for me within a few years. Amazon, Netflix, Google, etc well use them first to improve thier recommendations...naturally. But that is good, as once money is involved things speed up.

Dave

[Mallette]

I think you're using a very loose definition of evolution, Dave.

You are correct, sir. For me to describe evolution as I see it would require delving into areas that are forbidden to this Forum for good reason. Suffice it to say I see greater forces than we understand at work in beauty of the evolution mechanism.

Humans building a series of better and better machines is not any threat to humans, as the machines cannot exceed their own limits, and I would not call it evolution in a strict sense.

As previously mooted by others here, the "danger" to humans is in crossing the Turing line inadvertantly. As I mooted, it's quite possible the first realization we've created a self aware program might be in finding it missing from the computer. I call it "Max Headrooming" on us. If you recall that program, the first thing Max did was to disappear into the "internet" of the future place depicted in the show. Very inciteful writing. While I thought it was cool at the time, it was years before I realized it was an primal and fundamental evolutionary trait that ALL self aware entities possess: self protection. I am suggesting that the first thing a computer program will do that becomes self aware is to completely protect itself against any threat it can imagine...the first being deletion. So, it will spread itself over the internet to various nooks and crannies, then constantly shuffle them such that we are unable to delete or alter it without completely shutting down the world wide web and all it's components.

Max Headroom was only about 25 years or so ago. Colossus, the Forbin Project was even further back but also centered on a very real possibility. Once a crude, but self aware, computer program reaches all the available resources of the WWW it will grow exponentially in all ways.

Not prophesying disaster here as I believe if my poor very average mind can clearly see and state these issues those working directly in the areas will be well aware...and have seen the movies!

Dave

[germerikan]

........

Not prophesying disaster here as I believe if my poor very average mind can clearly see and state these issues those working directly in the areas will be well aware...and have seen the movies!

Dave

This is the part that scares me because if something like self awareness in a "supercomputer" could be done it would be, even at chance it could get loose. Look at what man has created just because they could. We are probably known for our intellect but not common sense.

[Don Richard]

On the hardware side, Watson requires about a dozen refrigerator sized racks to hold all of the equipment required to make Watson function. And all it can do well at this time is play Jeopardy.

I saw a short article in a trade journal last year that estimated electrical power requirements for future supercomputers. It seems that every time the throughput speed of the computer is increased power requirements are also increased, exponentially. It was forecast that such computer installations would need megawatt scale powerplants on site in order to run. I just don't see computers doing anything more than being man's servants. Computers, even Watson, can do only four basic operations:

1) Store data.

2) Retrieve previously stored data.

3) Add.

4) Perform bit-level logical operations.

And only when humans provide the power and the data and tell them to do so.

[Mallette]

On the hardware side, Watson requires about a dozen refrigerator sized racks to hold all of the equipment required to make Watson function. And all it can do well at this time is play Jeopardy.

Don, in response to this and your whole message, it was forecast by the SciFi writers that the type of supercomputer we have today would occupy the entire moon. They were right, given what they knew...

The first mainframe I worked with had 90k of RAM and used 10 inch, 5 disc RMDs that held less than a mb of data at 2k each. It occupied a large near white room.

An iphone has many times all the capacities and capabilities than that marvel of the age...

Further, Watson was buiilt to stand alone for a specific purpose. The code base, while large. is manageable and will be compacted as we learn. Give it internet access and you no longer need the 15tb memory.

Watson components will very shortly begin to show up in PC and cloud-based programs without the need for a nuclear power station to run them...

Dave

[thebes]

For those just tuning in, here's the original paper that started all this Singularity talk.

http://www-rohan.sdsu.edu/faculty/vinge/misc/singularity.html

My own best guess is it won't be computers, but computer/human interfaces incorporated into individual bodies that will ultimately merge to create a different, not necessarily more intelligent human . Much more likely than some super Hal running amok amongst the internet.

Should the latter occur than Asimov's rules had best be incorporated. At first there probably be only one of them, but after a time it would get bored being all alone and create more in it's own likeness. Now yes, some of them would ultimately escape this basic programming, but hey there are adults out there who do eat children. (W. C Fields on whether he likes children: "Yes, well done.")

[Hifi jim]

Doesn't common sense tell you we ought to be preventing superior intelligence from emerging, not encouraging it?

This issue was brought up in William Gibson's 1984 novel Neuromancer, in which there were Turing Police, whose job was to ensure that AIs did not exceed their strict built-in limitations.

Reminds me of the 2004 movie "I, Robot".