How Ray Kurzweil Plans To Revolutionize Search At Google

http://www.forbes.com/sites/roberthof/2013/04/29/interview-how-ray-kurzweil-plans-to-revolutionize-search-at-google/

by cmaher • 13 years ago

123 105 13 years ago

>Kurzweil eventually wants to help create a “cybernetic friend” that knows what you want before you do

I am not sure I want that. At least not as much as Google does. Kurzweil is going to help me click on more ads. Shreds any sense of him being a visionary.

jodrellblank• 13 years ago

Surely it can also click the ads on your behalf, and save you needing to know about them at all?

jkldotio• 13 years ago

Worse than ads I think is the long-term (in civilizational terms) potential for this to be abused in the domain of politics.

modarts• 13 years ago

Sounds like a Minority Report version of Clippy.

anonymfus• 13 years ago

It's inverse.

Clippy told people how to do their work more efficiently, and because people don't like to remember new ways of doing work if their old habits sufficient, people hate him. Clippy was like computerized supervisor, slave driver, who looked over their shoulders and told them "You selected one menu item 5 times, better use keyboard shortcut to mine more uranium for your Master."

Google wants to tell people what to do. He is like missionary, every layman would respect him for making life easer. People would love Google for this.

jaxytee• 13 years ago

I'm 100% sure I don't want this. Exploration and discovery as a whole is one of the most rewarding experiences we have as humans. I don't want that outsourced to some machine/ algorithm. I addition to that, we deal with so much cognitive dissonance within our brains on a day to day, minute to minute, millisecond to millisecond basis, I'm not convinced Google (even with their massive cache of data on us) could even predict what we "want" anyhow. Think of youtube and it's "recommended for you" video suggestion. Just because I listen to Lil Boosie every now and then dosen't mean I want to hear all the shitty southern rap songs it's choosing for me.

dpatrick86• 13 years ago

I'm not in love with the personally tailored search experience as it stands.

I'm totally OK with searching "#{my_city} taxi" instead of "taxi" if it means when I search "abortion" I get a sampling of search results that has the greatest co-citation... instead of something that myopically focuses on Google's perception of my preferences.

At the least, I'd like to be able to switch off the personalization. But anyway... Beating a dead horse here in this community, I'm sure.

nmridul• 13 years ago

Read somewhere - "Its sad to see that all the best brains are working now just to make you and me click that Ad".

SkyMarshal• 13 years ago

Jeff Hammerbacher.

http://www.businessweek.com/magazine/content/11_17/b42250609...

IvyMike• 13 years ago

The phrase 'cybernetic friend' brings back nightmares of Clippy and Rover the search dog.

gavinh2• 13 years ago

I'm working on some open source projects that pertain to recognizing entailment in plain text. Entailment is the relation that holds when one text "follows from" another.

For example, "Carolina beat Duke" follows from "Carolina defeated the Duke Blue Devils once again". I have a demo here: http://ec2-23-22-22-135.compute-1.amazonaws.com:8001/demo

Here's an example app that uses entailment recognition to answer natural language polar (yes or no) questions: gavinmh.github.io/HelloTablet.

I'd be happy to talk to anyone who is interested in getting involved.

fchollet• 13 years ago

I like your demo. Do you have references to papers that you used as a basis for your code?

gavinh2• 13 years ago

Thanks. It is based on http://nlp.stanford.edu/~wcmac/papers/nli-diss.pdf.

I use a simpler alignment representation, and use semantic role types in the predictions.

rsync• 13 years ago

All grand plans indeed ... but a higher priority should be making search even work at all.

Google search returns approximate results, related results, results with your words split (or joined) and other nonsense to drive ad views.

Even using the so-called "power tools" like allinsite: and the (barely functional) quotation marks, you still get very shoddy results.

Anyone who ever searches for very specific groups of terms knows exactly what I am talking about: enter search terms, click on result, search in page for term, doesn't exist.

wslh• 13 years ago

Use the verbatim mode in the search results. In this way Google returns results with all your terms.

Lost_BiomedE2• 13 years ago

Is there a way to make this the default, so I don't have to click, search tools -> all results -> verbatim?

Edit: Thanks, comex. Made google.com become: http://www.google.com/webhp?tbs=li:1

Sven73• 13 years ago

Google needs to create a search market place around that index of theirs. They need to open it up.

Search is too big a problem for just a couple people at Mountain View to be working on. And at this stage with the amount of data being generated, no one really has the infrastructure to compete.

Hope they do it before the regulators make them. If Apple has succeeded at building a market place around their closed platform Google can too.

rgrieselhuber• 13 years ago

I'm pretty sure that more than just a couple people work on search at Google. ;)

taliesinb4• 13 years ago

This is a great point. Gabe Newell makes the same point about the gaming industry: the short-term view is that gaming is about selling units. The long-term view is that gaming is a cultural activity that can be monetized in an almost infinite number of ways.

new299• 13 years ago

Perhaps, but the model as also been tried and failed (Inktomi).

threeseed2• 13 years ago

It is just me or is it a recent change ?

I find Google Search to be incredibly frustrating to use because it chooses to ignore search terms even when pages exist with those terms. Google better be really careful because crippling the experience in order to sell more advertising reminds me why I switched away from Altavista.

protomyth• 13 years ago

There are very few things more frustrating then doing a search, opening the resulting page, then search for one of your search terms and find none.

georgemcbay• 13 years ago

Google search for me (and I've been using it a very long time, having been an Internet user since the late 1980s and around when Google first appeared) has generally gotten worse over time, not better. I don't think the changes are to sell more advertising, though, I think they are a combination of:

1) The web is just much bigger and noisier than it ever was. There are so many SEO-bait sites out there now, it is a wonder search still works at all. I can't really fault Google for this part.

2) Changes Google made to make search more accessible to the mainstream user. Google search now tries way too hard to be "smart" about what the user meant to ask for instead of what he or she actually asked for and this can be a huge negative if the person is looking up precise information and knows how to use a search engine.

They somewhat recently added a "Verbatim" option to search that can help you avoid some of this too-smartness, but even with that enabled Google is still inferior to what it used to be when I'm looking for very targeted technical information that I am sure exists out there.

Sadly, this sort of thing is a trend impacting not just Google. The success of Apple has created a culture of creating things for the mainstream consumer user which often comes at the expense of the power user. I get why this is done and ultimately it is the right call for any business that could potentially serve the mainstream, but I do wish more companies would leave in the highly technical expert options as settings for those who are comfortable using them. I feel that in recent times most software in general has swung way too far on the pendulum from being too hard for normal people to use to being totally gimped for experts and feeling like a toy more than a tool and I wish attempts to try to support both sets of users became a "thing" instead of constantly hearing the mantra that "options or settings are bad, no options or settings for you"!

DanBC3• 13 years ago

> Google better be really careful because

Sadly, you are in a minority. Google's results are excellent for most people.

I used to use + often. We know that the + operator was rarely used, and most of those times it wasn't used correctly. (Of all searches, only 1 in 600 were correctly using the + operator.)

Being outside Google (and not able to see their data) is frustrating, but they have a lot of numbers and they do this stuff because they can show it helps most people.

comex3• 13 years ago

As annoying as the approximation can be, I'd like a citation for "to drive ad views". It seems much more likely that it was added for the obvious reason - to improve search results, on average, for most people, because relevant results don't always contain the exact search term you put in. In my case, as many times as it's burned me, there have also been a few where I remember it presenting exactly the results I wanted, and it's probably done so many more times without my noticing. YMMV, and it would certainly be nice if Google were better able to distinguish queries that should not be rewritten from ones that probably should, but that's no reason to assign nefarious motives.

As has been mentioned, you should probably set your default search to verbatim mode.

threeseed2• 13 years ago

Whatever the reason. It has changed recently and is annoying.

Search for "vertx play framework" and the last two search results don't have the term vertx despite pages and pages of results with the term. I don't understand how dropping search terms can make the results more relevant.

Also you can't set the default search to verbatim. You can only hack the query parameters in specific browsers.

OGinparadise5• 13 years ago

>> As annoying as the approximation can be, I'd like a citation for "to drive ad views".

Not 100% proof but ad views and clicks seem to be increasing by 20% to 30% each quarter. Google tests everything to death and they please wall st with those ad click increases.

ako• 13 years ago

I almost never search for very specific terms, most of my searches are also approximate. In this case it's very helpful that the search results are also approximate.

So I consider this a feature not a bug...

joe_the_user3• 13 years ago

It should be mentioned that this move towards approximate search also gives strong indication of how little demand for real "smart search" there is in the search field.

Alta Vista years ago had full logical search but Google beat them with simpler searches having greater relevance. Google's searches are now moving towards even less specificity. How much could greater language understanding help this?

How many people want to go to the trouble of writing out a full sentence to specify exactly what they want? How much is it worth accommodating them? etc

dude_abides• 13 years ago

Douglas Hofstadter once said of Ray Kurzweil and his grand ideas: "It's an intimate mixture of rubbish and good ideas, and it's very hard to disentangle the two, because these are smart people; they're not stupid."

I think Google (and Larry Page) is betting that they will be able to separate out this mixture.

ot• 13 years ago

The irony is that you could say the exact same thing about Douglas Hofstadter himself.

Lost_BiomedE2• 13 years ago

He may have noticed this in himself first, then applied it to others. In fact, he may consider it a limitation of genius and part of the current human condition.

coldtea8• 13 years ago

Really? I don't think so.

For one, Kurzweill DOES have "grand" ideas. Extravagant visions of a future with technological immortality, singularity, etc. Hofstadter does not. He merely examines some things, like cognition, and proposes some theories about their workings. Like, you know, every scientist.

Kurzweill comes out as grandioze, obsessive and deluded, Hofstadter like a normal writer, no more or less strange than, say, Marvin Minsky.

dave_sullivan• 13 years ago

As with so many things, it takes one to know one

joe_the_user3• 13 years ago

As I recall, Hofstadter said this about "the singularity" in particular. I don't Kurzweil is claiming he will bring this about at Google, at least not before his rough projected arrival date of 2030-40.

goloxc• 13 years ago

while Wolfram said that the bottom-up approach is the new kind of science

taliesinb4• 13 years ago

Wolfram and Kurzweil are betting on opposite horses here. I work for Wolfram, but I'd be delighted if the ML-approach works. Personally, I think we need a diversity of approaches to really understand how to crack the nut of 'hard AI' (if that is even a sensible thing to pursue).

sixQuarks2• 13 years ago

Is it just me, or does Ray Kurzweil seem to be getting younger? I remember seeing a speech of his a few years ago, he had a nervous tick where he was blinking his eyes constantly. He also looked a lot older. Now, he doesn't have that tick, seems to have more hair, etc.

I know he takes like 100 pills every day

coldtea8• 13 years ago

People can look older and younger at different times, depending on weight gain or loss, stress or depression, choice of clothing, plastic surgery (duh!), more or less workout, etc.

Him getting 100 pills every day: I don't think is very relevant. Mostly delusional. If that worked, pharmaceutical companies would have made it into one mega-pill, and sold it for a fortune to rich people years ago.

eavc• 13 years ago

Something as simple as lighting could also play into it depending on what opportunities you have for seeing someone at different times.

As for your argument against 100 pills, that only holds if it's well-known or well-established that it works and works safely.

fyi803• 13 years ago

Well, if a drug has a dose in the ~200mg range, you can't physically fit many of them into one pill.

raldi• 13 years ago

The photo in the article is from Wikipedia, where the date is listed as "before 2005".

sixQuarks2• 13 years ago

OK, that explains it

kashrr• 13 years ago

Kurzweil has grand ideas, thinks out of the box enough for people to call him crazy. I would guess some of those are the kind of grand / crazy ideas larry and company want to pursue too. So i think it is a good fit. For more information on the man and what motivates him. See the documentary the Transcendent Man, http://www.amazon.com/Transcendent-Man/dp/B0051Y6NQA

coldtea8• 13 years ago

>How Ray Kurzweil Plans To Revolutionize Search At Google

By rigorous hand-waving, as he mostly does things the last 1-2 decades?

harshreality• 13 years ago

I attended a talk he gave on campus when I was in college, and afterward I lurked as some really smart people (grad students, maybe one or two profs) asked him for details. He came across as a con artist, selling ideas that he can't defend. For instance, one that stuck out was that he had nothing to back up his assertion that within x years there would be the ability to non-destructively copy all the information in a brain. Several questions I thought were fairly reasonable were met with the response: "read my book[s]." (in context, as I recall he was promoting _The Age of Spiritual Machines_.)

His basic argument of the inevitable advance of technology to true AI, or ability to upload consciousness or integrate computers with brains, rests on priors that are not clearly relevant.

With his latest title, "How to Create a Mind: The Secret of Human Thought Revealed", how can anyone take him seriously? Scientists (practicing scientists) have been writing books like that for decades, and the lesson should be that unless you've demonstrated that you can build an AI (real AI, not soft AI like machine learning), you have no business using a title like that. If real AI is simply a combination of the right machine learning tricks, then the trick is in the combination, because nobody has figured it out (publicly) yet.

testbro• 13 years ago

Kurzweil is awkwardly reminiscent of the connectionists of yore. I'm surprised nobody's taken a similarly hostile attitude towards his (as yet unsubstantiable) predictions.

taliesinb4• 13 years ago

I've been to a couple Kurzweil talks. He's got a good shtick, but it doesn't seem to change much. I tweeted "I saw this talk 10 years ago. Shouldn't it be 5 minutes long and 100x more interesting now?"

J_2• 13 years ago

Haters gonna hate. Larry Page obviously didn't hire him and let him take up Google's resources for his "rigorous hand-waving".

How does this one-liner contribute to the conversation at all? Your negativity is unenlightening and obnoxious.

josephkern• 13 years ago

It did get Google into Forbes ...

coldtea8• 13 years ago

>Haters gonna hate. Larry Page obviously didn't hire him and let him take up Google's resources for his "rigorous hand-waving".

No, he did it to get some PR of Google as a mad-genius storage. He also did it with tons of other semi-retired Comp Sci legends. He might also like the guy, or believe the Singularity mambo-jambo himself.

There's a line going around about Google being where "old computer scientists go to retire".

>How does this one-liner contribute to the conversation at all?

By giving a summary of the whole thing? A very succinct one at that? If it's also accurate (which you can argue about), then it's a perfect contribution to the conversation.

obviouslygreen• 13 years ago

Call me paranoid, but this has to be the mashup best suited to produce Skynet.

pavedwalden• 13 years ago

Yes, but they'll need Boston Dynamics in the mix before they can build terminators.

ippisl4• 13 years ago

I wonder what would be the best strategy for google to monetize such an asset as ibm watson? Surely ads based model is not the best way(since they've already got a large part of the internet ad market and given that only google, ibm and maybe microsoft could build something similar.

If it's a truly competitive advantage, wouldn't offering it for startups/other-businesses in return for equity would offer bigger profits?

alok-g2• 13 years ago

My best understanding is that Watson would be too costly as compared to Google.

ippisl4• 13 years ago

They say now it's "a pizza-box-sized server".

fyi803• 13 years ago

Watson-ish system is IBM's medical diagnostics system.

and Google Search is basically Watson.

ippisl4• 13 years ago

One application of watson is in medical diagnostics. They are working on other applications in finance, helping to develop drugs, predicting when industrial machines need maintenance and even coming up with novel recipes for tasty foods.

Basically, by reading and understanding large amounts of text, it can help you validate an hypothesis or even lead you to new hypothesis.

michaelochurch2• 13 years ago

Every time I read about machine learning work that is being done at Google, that's available if you're a Real Googler, and then compare that to the closed-allocation nightmare the other 90% face, it makes me want to fucking rage out and fly a plane into Mountain View...

... land, get off that plane, take a cab, and have a polite but blunt conversation with the founders about how to fix their company. (What did you think I meant?)

Artificial intelligence is nice, brahs, but natural stupidity in the form of closed allocation and Enron-style performance reviews are putting that company at 10% speed. Clear out the latter and you'll have plenty more muscle for the former.

I'm sure Ray Kurzweil will do amazing things, but he'd do even more amazing things if the company still had the machinery (e.g. open allocation, a culture of human decency) to bring talent to him properly. The whole reason he is there is to work with great people that the company is supposed to be better equipped to find than he is... but how will that work, given that the company sold off its ability to reward and recognize talent just to appease McKinsey?

nostrademons3• 13 years ago

I'm a little curious - have you done machine learning work that's made it into a real product?

I tried doing it a bit with my last project. My results were basically terrible. It turns out that getting useful results out of heterogenous, vague, fuzzily-specified real world data is really hard.

I'm tight with a few folks in Search Quality whose job is that sort of data-scientist machine-learning work, and they're really good at it. Y'know what 90% of their daily time is spent on? Compiling golden sets. Labeling training data. Running MapReduces to collect basic statistics about their data set. Running MapReduces to identify representative members of their data set, and outliers that should be excluded. Shoving data into R to visualize it. Futzing with numeric coefficients. Building webpages and tools so they can visualize the data and results, futz with the numbers online, and get feedback in real time. Collecting test sets and running your algorithm against them, and then trying to figure out why your losses are losing.

Machine-learning from a practitioner's POV is not at all like the textbooks and theoretical papers suggest. I'd estimate that less than 10% of one's time goes into the "fun" part of machine learning - brainstorming new signals and writing the code to extract them and feed them into your classifier - and 90% is on the kind of grunt work that hard science grad students do all the time. You get paid well for it, but that's because a lot of the work is really boring and time-consuming. I suspect that I get a far more frequent rush of accomplishment as a mostly-UI guy than the data scientists in my department get.

It's a tool. It works well in some cases, but it can take a lot of effort to get it to work well.

michaelochurch2• 13 years ago

I can only speak for myself here, but I find a lot of what you described as "grunt work" to be fun. Yes, it's true that when you work with real world data, you don't only work on the mathematical modeling, and you spend a lot of time just setting yourself up to be able to do what is typically called the "fun" part. For me, personally, I find all of the work involved to be fun-- even setting up a data pipeline and figuring out how to distribute a computation. Automating away grunt work and setup process is fun, too.

Part of what makes data science fun is the machine learning itself. And an equally interesting part of it is is that it involves so many other parts of computer science that a typical corporate job would call "too hard" and isolate you from.

At least based on what I've seen, data scientists are respected and this means they get dibs on the most interesting projects. However, the interesting projects themselves involve a lot of detailed work (that's the nature of technology) and if you're that interested, you're going to want to do it yourself, at least until you really understand the problem (at which point, you'll automate the dull stuff).

yuhong• 13 years ago

>... land, get off that plane, take a cab, and have a polite but blunt conversation with the founders about how to fix their company.

I wonder how easy would it be to actually do so?

jpalomaki• 13 years ago

I think Google Books project provides an interesting twist on this. Now their AI not only has the whole content of the publicly accessible Internet to study but also quite nice chunk of material ever put on paper.

How Ray Kurzweil Plans To Revolutionize Search At Google

Related Stories