Slashdot Mirror
Why Google Hired Ray Kurzweil
An anonymous reader writes "Nataly Kelly writes in the Huffington Post about Google's strategy of hiring Ray Kurzweil and how the company likely intends to use language translation to revolutionize the way we share information. From the article: 'Google Translate is not just a tool that enables people on the web to translate information. It's a strategic tool for Google itself. The implications of this are vast and go beyond mere language translation. One implication might be a technology that can translate from one generation to another. Or how about one that slows down your speech or turns up the volume for an elderly person with hearing loss? That enables a stroke victim to use the clarity of speech he had previously? That can pronounce using your favorite accent? That can convert academic jargon to local slang? It's transformative. In this system, information can walk into one checkpoint as the raucous chant of a 22-year-old American football player and walk out as the quiet whisper of a 78-year-old Albanian grandmother.'"
The main thing he says he will be working on is artificial intelligence that can understand "context". The goal is for Google search to be able to find pages etc based on what you mean rather then on word counts of what you type.
"Have you ever thought about just turning off the TV, sitting down with your kids, and hitting them?"
Many of the "language processing" problems the OP describes are actually "cognition" problems. If Google is serious about algorithmically translating from "academic jargon to local slang", then they're looking at writing an AI which can in some sense understand what is being said.
I guess it's a good thing Kurzweil's on board.
DRM: Terminator crops for your mind!
You need to investigate the entire initiative Google is spearheading around its acquisition of Metaweb. They are building an ontology for human knowledge, and are ultimately building the semantic networks necessary for creating an inference system capable of human level contextual communication. The old story about the sad state of computers' contextual capacity, recounts the story of the computer that translates the phrase "The spirit is willing, but the flesh is weak." from English to Russian and back and what they got was "The wine is good but the meat is rotten."
The new system won't have this problem. Because it will instantly know about the reference coming from the Bible. I will also know all the literary links to the phrase, the importance of its use in critical historical conversations, The work of the Saints, the despair of martyrs, in short an entire universe of context will spill out about the phrase and as it takes the conversational lead provided by the enquirer it will dance to deliver the most concise and cogent responses possible. In the same way, It will be able to apprehend the relationship between a core communication given in context 'A' and translate that conversation to context 'B' in a meaningful way.
Ray is a genius for boiling complex problems down into tractable solution sets. Combine Ray's genius with the semantic toy shop that Google has assembled, and the informational framework for an autonomous intellect will become. The real question is how you make something like that self aware. There's a another famous story about Helen Keller, before she had language. symbolic reference, she lived like an animal. Literally a bundle of emotions and instincts. One moment, one utterly earth shattering moment there was nothing, then Annie Sullivan her teacher placed her hand in a stream of cold water and signed water in her palm. Ellen understood... water. In the next moment Ellen was born as a distinct and conscious being, she learned that she had a name, that she was. I don't know what that moment will look like for machines, I just know its coming sooner than we think. I also can't be certain whether it will be humanities greatest achievement or our worst mistake. That awaits seeing.
The problem with people like Kurzweil, Jeff Hawkins, the folks at the Singularity Institute and the rest of the AI community is that they have all jumped on the Bayesian bandwagon. This is not unlike the way they all jumped on the symbolic bandwagon in the last century only to be proven wrong forty years later. Do we have another half a century to waste, waiting for these guys to realize the error of their ways? Essentially there are two approaches to machine learning.
1) The Bayesian model assumes that events in the world are inherently uncertain and that the job of an intelligent system is to discover the probabilities.
2) The competing model, by contrast, assumes that events in the world are perfectly consistent and that the job of an intelligent system is to discover this perfection.
Luckily for the rest of humanity, a few people are beginning to realize the folly of the Bayesian mindset. When asked in a recent Cambridge Press interview, "What was the greatest challenge you have encountered in your research?", Judea Pearl, an Israeli computer scientist and an early champion of the Bayesian approach to AI, replied: "In retrospect, my greatest challenge was to break away from probabilistic thinking and accept, first, that people are not probability thinkers but cause-effect thinkers and, second, that causal thinking cannot be captured in the language of probability; it requires a formal language of its own."
Read The Myth of the Bayesian Brain for more, if you're interested.
Yeah, but there's a reason why statistical models are hot now and why the old AI-style of logical reasoning isn't: the AI stuff only works when the input is perfect, or at least, planned for. As we all know, language doesn't really have rules, just conventions. This is why the ML approach to NLP is powerful: the machine works out what was probably meant. That's far more useful, because practically nobody writes well. When Abdur Chowdhury was still Twitter's main NLP guy, he visited our department, and guess what-- people even write in more than one language in a single sentence! Not to mention, in the old AI-style approach, if you fill a big box full of rules, you have to search through them. Computational complexity is a major limiting factor in all AI problems. ML has this nice property that you can often simply trade accuracy for speed. See Monte Carlo methods.
As you point out, ML doesn't "understand" anything. I personally think "understanding" is a bit of a squishy term. Those old AI-style systems were essentially fancy search algorithms with a large set of states and transition rules. Is that "understanding"? ML is basically the same idea except that transitioning from one state to another involves the calculation of a probability distribution, and sometimes whether the machine should transition is probabilistic.
I think that hybrid ML/AI systems-- i.e., systems that combine both logical constraints and probabilistic reasoning-- will prove to be very powerful in the future. But does that mean these machines "understand"? If you mean something like what happens in the human brain, I'm not so sure. Do humans "understand"? Or are we also automata? In order to determine whether we've "cracked AI", we need to know the answers to those questions. See Kant and good luck.