Technology Review
A massive AI project called CALO could revolutionize machine learning.
The thing that makes computers a huge pain for everybody, says Pedro Domingos, an associate professor of computer science at the University of Washington, is that you have to explain to them every little detail of what they need to do. "It's really annoying," Domingos jokes. "They're stupid."
That's why Domingos is taking part in CALO, a massive, four-year-old artificial-intelligence project to help computers understand the intentions of their human users. Funded by the Defense Advanced Research Projects Agency (DARPA), and coordinated by SRI International, based in Menlo Park, CA, the project brings together researchers from 25 universities and corporations, in many areas of artificial intelligence, including machine learning, natural-language processing, and Semantic Web technologies. Each group works on pieces of CALO, which stands for "cognitive assistant that learns and organizes."
Adam Cheyer, program director of the artificial-intelligence center at SRI, explains that CALO tries to assist users in three ways: by helping them manage information about key people and projects, by understanding and organizing information from meetings, and by learning and automating routine tasks. For example, CALO can learn about the people and projects that are important to a user's work life by paying attention to e-mail patterns. It can then categorize and prioritize information for the user, based on the source of the information and the projects to which it is connected. The system can also apply this type of understanding to meetings, using its speech-recognition system to make a transcription of what's said there, and its understanding of the user's projects and contacts to process the transcription intelligently into to-do lists and appointments. Finally, a user can teach CALO routine tasks such as purchasing books online and searching for bed-and-breakfasts that meet specific criteria. CALO can interact with other people, taking on tasks such as scheduling meetings, coordinating among people's schedules, and making decisions, such as deciding to reschedule a meeting if a key member becomes unable to attend.
"It's an amazingly large thing, and it's insanely ambitious," Domingos says. "But if CALO succeeds, it'll be quite a revolution. Even if it doesn't, so much good research is happening under it that it will still have been worthwhile."
The goal is to build an artificial intelligence that can serve as a personal assistant for someone--not something with a rigid structure within which it can be helpful, like the animated paper-clip assistant featured in Microsoft Office products, but a system that can learn about a user's environment and needs, and adapt to them, without having to be programmed anew by engineers. "What's different and has never been done before in this way is the truly integrated approach of bringing all of these technologies and all of these capabilities into a single system," says Cheyer. "It takes a system of this size to give you something that can understand and organize so much information."
A startup has licensed heavyweight artificial-intelligence technology for its online recommendation engine.
Software lets a hybrid car learn your habits and optimize its power usage.
Great attempt. I wish you every success further.
But any approach which tries to simulate any functionality of the brain (learning) should take in to account the nature of forgetting. The way brain forgets over time is one of the main reason for the brain's speed. Its remembering capacity is due to the basis of the frequency of the data usage. Also forgetting is due to not using the data over long time. This helps to throw away or exclude the irrelevent data for the fact at hand.
Probably we can make intelligent system better by associating the forgetting threshold to each and every fact & data and modifying it over time based on the usage frequency.
www.browsetoknow.blogspot.com
I disagree. Though forgetting is an essential operation to human intelligence it is merely an evolutionary design constraint begging to be expanded upon.
Sure, to start with a working model such emphasis should be considered....but after making it work; to make it work _well_ such constraints should be solved.
futureprogress.net
Actually, forgetting is a very practical issue with CALO for two reasons. First, it is easy to fill the disks on current desktop and laptop computers with the CALO knowledge base, data base, event history, and harvested documents. Second, the user's "work world" changes over time as projects come and go, so it is important for the learning algorithms to adapt to these changes (e.g., by downweighting older data or by explicitly modeling the change process). So while in principle there is nothing to gain by forgetting things, in practice there is a practical need and a scientific justification for forgetting old data.
actually the human brain and computers are very similar already in the forgetting process....however in forgetting we are not disposing the information but instead like a computer telling us that there are unused icons on the desktop would you like to cleanup the desktop in which it compresses and archives the data
I agree for the point you made. It is crucial in the beigning to make the models work properly.
The point I wolud like to emphasis is that we can assign some weights to the facts/data and imporve the data access decisions. These weights are dynamically modified based on the frequency of the data references. As the weights vary for each data unit in the knowledge base, accessiing the frequently reffered data should be made faster than rarely accessed data(The way brain recalls). I do not know whether this is a relevant and required future for your system.
Thank You.
sman
Why is everyone so giddy about a machine that learns? Lest we forget - what we dream of, can (and usually does) happen eventually. Given our imagination (ie, movies/TV show/etc.) we should be asking ourselves 'Is this going to lead somewhere that we do not want to go?' This should not be considered part of evolution on human terms, but on machine terms. Look at our past to look at our future. Could it be better with our roles reversed? Only you can decide...
From what I understand our brain is FAR more capable of remembering things than pretty much anybody on the planet uses it. I don't think is so much HAS to throw out information, but that through lack of use the connections die. Which is very important as logic is not an instinct, without killing off old connections I think we would be EXTREMELY prejudice and never be able to get over addictions and habits we had.
A lifebyte is 151.3 quadrabytes
Lifetime of experience
In neurophysiology it has been observed that with direct electrical stimulation to the brain anything from your past is available and nothing is forgotten. A lot of old knowledge is waiting for an association to a stimulation or need to be recalled. I hope computers can store everything I have ever experienced or thought and associate and recall it when needed. I made a rough calculation that if you had a video camera in HD recording everything you saw, heard or said for 100 years it would require 151.3 Quadrabytes to store it all, perhaps called a lifebyte. Add the sum total of human knowledge the AI system can access, I think we would have a much faster development of new knowledge.
Re: A lifebyte is 151.3 quadrabytes
Wouldn't that be the data only? Interpretation of the data (knowledge) would require additional overhead. But a good approximation none the less. I wish I could see and remember in high def though :)
Re: A lifebyte is 151.3 quadrabytes
I don't think the brain saves data, especially visual data, in the linear approach we use in computers. I think our brains use a very complex association sparking neurons that combine create patterns that help us kind of reverse engineer the visual scene from our memory. And that is why witness testimony is so unreliable and subject to the emotions and other input while the memory is being created. All in all with the (off the top of my head) something like 7 billion neurons in our brain. Assuming that each neuron is capable of multiple equivalent bits of information, I think it would still be considerably smaller than 151.3 quadrabytes.
so I'm just wondering how this compares with cyc?
http://www.opencyc.com
This sort of technology would be a perfect fit for a business technology strategy called BPM, or Business Process Management.
Just like a living cell has millions of chemical processes which are routinely executed and orchestrated to achieve the purposes and goals of the organism, so too do large corporate organizations have work "processes" in which many people engage in patterns of email, phone calls, paper forms, software usage, and face-to-face communications.
These interactions stream together into processes, which are now being targeted for identification and automation through the "BPM" methodology of IT system design, and execution in BPM software platforms.
The task of identifying and designing these work processes in corporations is the current bottleneck in this billion dollar, emerging industry.
Automatically mining process data from organizations is a holy grail in the BPM industry, and it would be great if this technology were applied to this specific business problem.
Manufacturing in the United States is in trouble. That's bad news not just for the country's economy but for the future of innovation.
This document is part of the “How-To Guide for Most Common Measurements” centralized resource portal. This tutorial provides a detailed guide for measurement and device considerations to take temperature measurements using thermocouples. Get an introduction to thermocouples, which are inexpensive sensing devices widely used with PC-based data acquisition systems. Also review some specific thermocouple examples and learn how thermocouples work and ways to integrate them into a data acquisition measurement system.
View full PDF >
Listen to story > 
Our list of the 50 most innovative companies, including the following:
On Twitter
Become a Fan on Facebook
Subscribe to the Feed
fiberman
186 Comments
AI!
When MIT started looking at AI sometime in the last century, I used to say "I won't believe in 'artificial intelligence' until someone proves to me that 'natural intelligence' really exists. As the IBM instructor who taught me programming 45 years ago (IBM 650s with tubes) used to say: "Computers do what you tell them to do, not what you want them to do." Not much has changed, eh?
Heck, I'd be happy if MS Word would simply leave me alone to write what I want instead of guessing and suggesting!
Reply
Watcher
2 Comments
Re: AI!
"Computers do what you tell them to do, not what you want them to do."
No, not much has changed - for now...
Natural intelligence is almost absent due to 'less-than-desirable' factors inherent in modern society.
In the absense of intelligenge comes ignorance and arrogance, both are bad enough on their own, but together they can be absolutely deadly.
And, Brother Bill won't leave us alone. MS is intergrated into almost every system on the planet. We all have the MS blues now. The question now is: Should we do without MS, Computers, or anything associated with them and the oncoming emergence of AI? Can we?
Reply