Slashdot Mirror
Cognitive Machines Help Decision-Making
Roland Piquepaille writes "At Sandia National Laboratories, new "smart" machines can accurately infer your intents and help you to take better decisions or avoid mistakes. They could change in a near future how we interact with computers, according to this news release. The team who developed the concept associated cognitive psychologists and robotics researchers. The Sandia team thinks that "it's entirely possible that these cognitive machines could be incorporated into most computer systems produced within 10 years." This summary contains more details, including a photo of a "Sandia software developer operating a simulation trainer while a cognitive model of the software runs simultaneously.""
A new type of "smart" machine that could fundamentally change how people interact with computers is on the not-too-distant horizon at the Department of Energy's Sandia National Laboratories.
Over the past five years a team led by Sandia cognitive psychologist Chris Forsythe has been developing cognitive machines that accurately infer user intent, remember experiences with users and allow users to call upon simulated experts to help them analyze situations and make decisions.
"In the long term, the benefits from this effort are expected to include augmenting human effectiveness and embedding these cognitive models into systems like robots and vehicles for better human-hardware interactions," says John Wagner, manager of Sandia's Computational Initiatives Department. "We expect to be able to model, simulate and analyze humans and societies of humans for Department of Energy, military and national security applications."
Synthetic human
The initial goal of the work was to create a "synthetic human" -- software program/computer -- that could think like a person.
"We had the massive computers that could compute the large amounts of data, but software that could realistically model how people think and make decisions was missing," Forsythe says.
There were two significant problems with modeling software. First, the software did not relate to how people actually make decisions. It followed logical processes, something people don't necessarily do. People make decisions based, in part, on experiences and associative knowledge. In addition, software models of human cognition did not take into account organic factors such as emotions, stress, and fatigue -- vital to realistically simulating human thought processes.
In an early project Forsythe developed the framework for a computer program that had both cognition and organic factors, all in the effort to create a "synthetic human."
Follow-on projects developed methodologies that allowed the knowledge of a specific expert to be captured in the computer models and provided synthetic humans with episodic memory -- memory of experiences -- so they might apply their knowledge of specific experiences to solving problems in a manner that closely parallels what people do on a regular basis.
Strange twist
Forsythe says a strange twist occurred along the way.
"I needed help with the software," Forsythe says. "I turned to some folks in Robotics, bringing to their attention that we were developing computer models of human cognition."
The robotics researchers immediately saw that the model could be used for intelligent machines, and the whole program emphasis changed. Suddenly the team was working on cognitive machines, not just synthetic humans.
Work on cognitive machines took off in 2002 with a contract from the Defense Advanced Research Projects Agency (DARPA) to develop a real-time machine that can infer an operator's cognitive processes. This capability provides the potential for systems that augment the cognitive capacities of an operator through "Discrepancy Detection." In Discrepancy Detection, the machine uses an operator's cognitive model to monitor its own state and when there is evidence of a discrepancy between the actual state of the machine and the operator's perceptions or behavior, a discrepancy may be signaled.
Early this year work began on Sandia's Next Generation Intelligent Systems Grand Challenge project. "The goal of this Grand Challenge is to significantly improve the human capability to understand and solve national security problems, given the exponential growth of information and very complex environments," says Larry Ellis, the principal investigator. "We are integrating extraordinary perceptive techniques with cognitive systems to augment the capacity of analysts, engineers, war fighters, critical decision makers, scientists and others in crucial jobs to detect and interpret meaningful patterns based on large volumes of data derived from diverse sources."
"O
"I'm sorry Dave, I'm afraid I can't do that"
Gibble: Descriptive of an emotional state in which one's mind is scrabbling for some purchase on reality
deserve's got nothing to do with it...
Will this machine allow me to install Windows on my PC?
Omnis amans amens
the preceding comment is my own and in no way reflects the opinion of the Joint Chiefs of Staff
"You didn't really want to make that choice, did you? Of course not... let me fix it."
All I want is a kind word, a warm bed and unlimited power.
I could use a smart machine to aid my decision making relative to posting on Slashdot.
It could warn me when I'm about to submit a post that's impulsive and likely to be modded down.
Hmm.. maybe I could use one right now.
Dave% vi PodBayDoors.c ... ...
Message from HAL@localhost on pts/2 at 09:56
HAL: I'm sorry Dave, I'm afraid you can't do that.
EOF
Dave% echo What\'s the problem\? | write HAL
Message from HAL@localhost on pts/2 at 09:57
HAL: I think you know what the problem is just as well as I do.
EOF
...and I'll say it again. No, I don't want to go there today.
Upcoming feature: these cognative models will soon all talk to each other through a new protocol called Skynet :-P.
Computer Science is all about trying to find the right wrench to bang in the right screw. -T.Cumbo?
Microsft "Bob".
Roving Web-Teleoperated Robot
I see your trying to write a letter....
Noooooooo! Bill! Stop trying to help me.
Visit Jonesblog and say hello.
kind of technology that won't let me open the pod bay doors when I want.
The Kruger Dunning explains most post on
Thank goodness! Now we might be able to bargain for our lives with the military's Super Death Robots. "Cognitive" means "Understands bribes", right?
Or if that fails, we can just sprinkle some rust-monster microbes on them.
Best read in good ol' Monaco 9 point.
Think of the possible effects that this sort of technology could have on gaming. Although AI gets better and better every day, are we looking at a future where playing against a machine would be almost the same as playing against a human?
Me thinking hard: I will start the day with a +5 Insightful post or a +5 Funny. Insightful or Funny...Insightful? Funny? No no,funny=>No Karma=>Post insightful comment. But Funny comment=>feel witty and warm inside. Funny, insightful....funny, insightful...*ggnnnn*.
Sandia machine: Seg fault core dumped.
Seriously, more than half the time, I can't even figure out what the next human I meet intends to do. It's really REALLY hard, even if you use the current/past actions as a guide. Face it, we humans are REALLY unpredictable creatures. And women more than men.
Now we will have something tangible to vent frustrations on when PaperclipRobot v1.02.39 says "You seem to be paying bills.."
1) Use this technology to get you to links before they are slashdotted.
2) Have Slashdot create ever-increasing 'Super' Subscriber options. For an extra $20 you get stories before subscribers do. (followed iteratively by the Super-Super and Super^3 subscriber levels).
sarchasm: The gulf between the author of sarcastic wit and the person who doesn't get it.
Some of the ideas presented in the Anti-Mac interface (Google Cache) guildlines. Also, this reminds me a lot of some research that was done by Douglas Hofstadter and Melanie Mitchell and described in "Fluid Concepts and Creative Analogies". I highly recommend the book if you are interested in AI.
Helping with organizational effectiveness is our job.
At Sandia National Laboratories, new "smart" machines can accurately infer your intents and help you to take better decisions or avoid mistakes.
Me: Dude, I'm so trashed. Is that girl hot?
My Smart MachineNegative. Your beer goggles have wrongly given her a +5 hot. The correct answer is -1 fat.
I predict that if/when such a technology becomes prevalent, it will greatly reduce the human ability to make decisions.
Take for example any simple video game, how about MahJong (the stack of tiles that you have to match pairs on to remove them).
If you play it without the computer's aid, you develop a good eye for it and can do quite well. However, if you constantly hit the 'help' or 'hint' button, you become dependant on it showing you the next move, and never develop the skill for yourself.
To put it in context with other situations:
How many of you need a calculator to find a 10, 15, or 20% tip amount? Worse, how many of you need a calculator to add that extra 3.25 onto your 21.75 bill? I admit, it takes a great bit of effort for me to add simple numbers in my head simply because I don't exersize that ability enough.
no comment
I think calling this cognitive computing is a bit of an overstatement. It seems more like a heuristic tool that learns the behavioral patterns of a human and alerts the human when something deviates from the norm. We have a long way to go before we have real computer cognition.
The Moore-Murphy Law: The number of things that will go wrong will double every 2 years.
it seems that "intelligent machines" is a bit too much of a generalization. what they are doing is teaching a machine/software how to do something correctly, then have it correct humans when they do it wrong, based on their cognitive model. this is all well and good, but "intelligence" implies some sort of learning. this learning has to be online, i.e. the machine learns how to do something without a stimilus to learn it specifically. what sandia labs has done is get software to infer how a human made decisions to get a certain "state", but this is not exactly "intelligence". just my $0.02
BSD is for people who love UNIX. Linux is for those who hate Microsoft.
- Clothing ("You're not really wearing that to my mother's, are you?")
- Money ("It costs HOW much? Forget it")
- Housekeeping ("NO, I already told you - glass cleaner on the top shelf and bleach on the bottom")
- Driving ("SLOW DOWN! Watch the guy on the bike")
- Entertainment ("Give me the remote. Bridges of Madison County is starting in a minute")
I believe the simplest solutions are always the best...What this group has going sounds good, but so have many other things I've read about AI related. How about a video or something to show what it really does? I mean, if they have this mega software then making a video of it in action can't be all that hard can it?
Questions blatantly not answered in these articles:
- Does it read screen text? If so, how? OCR?
- What api's is it compatible with?
- What operating systems does it work with?
- What exactly is the special hardware?
- What spoken languages is it compatible with?
- Specifically, what kinds of decisions can this AI make, what decisions can it not?
- How long does it take to make a decision?
I could go on, but I think you get the point...If old people in Florida actually MEANT to push 1 thing and missed, could this catch it and say "No, I think you meant to vote for Candidate X"? I think this could revolutionize voting in the USA...Maybe it could even be used to replace congress...it's like I, Robot...but without the 3 laws! Hooray!
Arthur: I mean what's the point? .. ...
;-)
Machine: Nutrition and pleasurable sense data. Share and Enjoy.
Arthur: Listen you stupid machine, it tastes filthy! Here take this cup back!
Machine: If you have enjoyed the experience of this drink, why not share it with your friends.
Arthur: Because i want to keep them. Will you try and comprehend what i'm telling you? That drink..
Machine: That drink was individually tailored to meet your personnal requirements for nutrition and pleasure.
Arthur: Ahh! So I'm a masochist on diet am I ?
Machine: Share and enjoy!
Arthur: Oh! Shut up!
Machine: Will that be all ?
Arthur: Yes! No look! It's very very simple, all I want.. Are you listening?
Machine: Yes.
Arthur: Is a cup of tea. Got that ?
Machine: I hear.
Arthur: Good and you know why i want a cup of tea?
Machine: Please wait..
Arthur: What ?
Machine: Computing
Arthur: What are you doing ?
Machine: Attempting to calculate answer to your question, why you want dry leaves in boiling water.
Arthur: Because I happen to like it, that's why.
Machine: Stated reason does not compute with program facts.
Arthur: What are you talking about ?
Ventillation: You heard.
Arthur: What? Who said that?
Ventillation: The ventillation system, you had a go at me yesterday.
Arthur: Yes, because you keep filling the air with cheap perfum.
Ventillation: You like scented air, it's fresh and invigorating.
Arthur: No I do not!
Seriously! No thanks
In 2020, your average $1000 Wal-Mart Computer will be roughly complex enough to emulate a human brain in realtime. Toss in some cognitive modelling, and you have your new plastic pal who is fun to be with.
Perhaps something like this could be used for Slashdot. Before posting a story to the front page, the congnitive system can read the story, and then make all the obvious comments such as adding Simpsons/Futurama quotes, comparing it to one or another movie, and finding ways to attack Microsoft regarding the story. Then it can be posted, and we don't have to read the same 4-5 comments 50 times throughout the story.
Of course, if that was added, what would be left for the people on Slashdot to talk about?
"You know your god is man-made when he hates all the same people you do."
Hey, CmdrTaco, Slashdot staff, is there any way you can get an interview with these guys?
/. interview with these guys would just be awesome. I bet they are /. reader, too.
This is some awesome work, but the article is so thin. A real
Anyone care to second the motion?
"I don't know half of you half as well as I should like, and I like less than half of you half as well as you deserve."
I don't know if these guys have something for real or not. Their press release is - perhaps unsurprisingly - fluff that says nothing about how their system is supposed to work. Without knowing some technical details its almost impossible to evaluate how sound their approach is and whether they've got anything new.
However a couple of things are suspicious. First they say "work on cognitive machines took off in 2002". So in less than 2 years they have essentially cracked several of the major problems that AI researchers have been struggling with for at least the last 4 decades? That seems unlikely. Second they seem to think that a combination of software engineering, cognitive psychology and robotics is the silver bullet of AI and that this is a radical new breakthrough. I hate to break it to them but these disciplines have been working together for many years in the AI community. This just isn't new.
Until we have a techical paper that describes their approach in detail and can be peer reviewed I will remain sceptical. AI got overhyped enough as it is, we don't need more extravegant claims and fluff press releases.
Sailing over the event horizon
The good thing about these systems is allowing them in an advisory capacity. I think it's just human instinct to want something to do all your work for you, but luckily it's also human instinct not to fully trust machines. This is why I don't think people will ever have cars that totally drive themselves or computers that decide when your nukes launch. It's not that they don't have the potential to do these jobs, it's just that there's always the feeling that human error is more reliable than computer error.
Oh yeah? You mean like that annoying microsoft text selection that prevents you from selecting what you actually selected by deciding for you that you wanted to select the entire word/sentence/paragraph/page (depending on its mood?
I have cursed so much because of that "feature"!
I am the apostle of the "leave me the fuck alone" tao of programming: Every application should have one button, in one simple easy to find menu, that would turn off all automatic thingamajigs. Instead of the current system in wich the are 1.5 times as many places in wich you need to select "no/off" as there are annoying automatic features.
When I place my cursor in the middle of a word, its because thats where I want the selection to end!
You can't take the sky from me...
That could replace most of my family right there.
If you were blocking sigs, you wouldn't have to read this.
We've all seen this so many times before. Artificial Intelligence is a sham. It's analagous to alchemy. If you just put enough ingredients together, you've got intelligence. Bullshit. We don't even know what is necessary and sufficient for intelligence. We can't agree on the concept/definition, and I fear that if we could, no human would qualify.
As pertaining to this article...it's easy to get something that resembles intelligence in a closed, restricted, experimental environment. When you try to expand it, you get something like clippy. Annoying and unhelpful, and certainly not intelligent.
There are good, efficient algorithms that can help humans in many ways. But don't call it a "synthetic human," don't call it "intelligence," and don't believe it's going to start thinking for us in general terms. That fad went out in the 70's.
...just my 2 gil.
Of course, once the cognitive model becomes good enough, the temptation (economic imperative?) will be to offload some of the actual work onto it. This idea, taken to an extreme, is the topic of an excellent short story:
"I was six years old when my parents told me that there was a small, dark jewel inside my skull, learning to be me." (Greg Egan, "Learning to be me", Axiomatic collection)
Back on the topic of augmented intelligence, Kasparov has been advocating allowing mixed human/computer teams in "Advanced Chess" tournaments. It seems that the human/machine combination, with the right interface, yields far better chess play than either alone.
Tell me this: If Microsoft hadn't released the details, how on earth would an attacker have known how to write an exploit for the "RPC hole?" If you think full disclosure is a good idea, tell that to the families of the victims in New York, Detroit, and Cleveland. Microsoft has willingly provided the tools necessary to cripple the United States' information infrastructure.
You do not work in the IT / Security field do you?
Microsofts explanations come LONG after the whole blackhat community knows about them. The only people who LEARNED about the RPC whole from MS were the legions of "boot camp MCSEs" that MS turns out. Very similar exploits existed long before MSBlast came out.
Before I write the next paragraph I want to add that I also question MS security practices but not like you. The patch for this as well as protective measures were described in the same release you are speaking of.
Now to get scathing:
First, if a single PC can bring down a power grid that large, many people should be fired (from the power company). Not only computer admins but those who designed the system. Most small businesses can survive a few PCs going down now days.
Second, the patch was available weeks ago. I have over a hundred PCs and a half dozen servers and have not been infected.
Finally, you give no sources. It is an interesting diatribe but it lacks any and all seblance of substance.
In Notes From The Underground, he discussed tables (now would be computers) which would show and calculate the response to evertything and what people would do. People would not want to see that, if they did see it, they would rebel that and choose crazy decisions, just to rebel.
This doesn't exactly relate to the article, but the article reminded me of this.
. . . how will we know wether the error is software or user?
.
If your machine starts arguing with you, how do you determine the flaw? When it keeps making consistently wrong decisions, who is to blame.
I'm seeing a WHOLE new way around tech support here. Just keep telling the users that the machine is right and they're wrong. How will the average user know?
All jokes aside, as we humanize software, we need to develop ways to evaluate it and debug it that will require whole new ways of thinking.
Ten years? I'm not so sure it'll be that quick . .
"The Sage treasures Unity and measures all things by it" - Lao Tzu
I share your frustration totally (sometimes Word expands my selection to include the punctuation at the end of a word... wtf!?). However, when people say "I don't need any help from my computer!" I feel they aren't thinking it through -- your computer is always assisting you to some degree. This notion of "overzealous assistance" is all relative. My mom needs AOL in order to "see the Internet" (it's like fingernails on a chalkboard when she says that to me), but I find the level of "help" that AOL provides as frustrating and cumbersome.
Perhaps that's where this technology could be put to it's best use: correctly intuiting exactly how much assistance you may need. So, for example, Clippy v2.0 would ask you if you need help with that letter when you are a newbie, but scale back the assistance when it intuits that you don't need help.
And in ten years, when natural language parsing and voice recognition are perfected, it could go like this.
...the neighborhood of 10,000,000,000 neurons, each having up to thousands of connections. All of these are working in parallel.
I think you just described an epileptic seizure.
A programmer is a machine for converting coffee into code.