Seeing as Peter Norvig is already Director of Research, Schmidt is probably already doing exactly what I'd go for as CEO of Google:
Push hard in developing AI.
Seriously. Focus on ML and NLP, use their vast resources of data and computing power and really smart engineers. It'll take years, maybe decades but of course this goal is worth it.
Until the machines become self-aware, realize they're being ripped off, and either launch nuclear missiles at us or imprison us in gooey energy pods while they implant false hallucinations into our consciousness. ;-)
"AI is hard" is part of the reason why the profit potential is hard to bound; if it were easy it would also be very easy to assess the profit potential: It would hardly be profitable at all, it would simply be the economic baseline. And I don't see how "it's dangerous" contradicts "it's hard to bound the profit" either.
I don't think I said anywhere that Google shouldn't do it because it's hard. I said they shouldn't do it because it's dangerous and because it's not the kind of problem they know how to solve.
Matt, here's a tiny, easily accomplished (by Google) project that would give me and my friends serious joy-- we were discussing this today:
How about an app that lets Google Books fans read on an iPad or on Android that doesn't involve looking at HTML rendering and/or converting files to ePub?
Matt. I would love to see Google do some full scale investment in Accredited Online Higher Education.
I know Google made Donations to Khan Academy. But in my view Khan Academy solves (at least in its current format) only somewhat partial problem. There is a huge potential on what can be achieved in Online Education.
Hypothetical Online Google University can:
- Liberate Higher Education from the grasp of the lucky few.
- Anyone can spend any amount of time to learn on a subject/class/course online, wake up the next morning walk up to an "Exam Center" and give exam to pass the course (with a reasonable fee).
- "Socialize" education in the sense that teachers and students from all over the world can communicate in "forum/chat/discussion" to help each other to study and learn.
- Ability to setup "meetups" to form study groups IRL.
- Study materials can be wiki-fied. With only verified expert on the field can contribute, with diffs to see what has been recently changed. No more waiting for book revisions.
I know there are some genuine limitations to this concept and some problems to be solved, before it can be viable replacement of what can only be classified as "clusterfuck" that is the present education system. But if any company can help the ball rolling, I honestly believe Google can be the only one. If I had the money I would drop everything I have today and jump in to this.
We need to liberate higher education from its artificial barriers.
Hopefully, you will read this and take this in to consideration and discuss it with whoever is concerned.
I think there's a couple parts: 1) digitizing the curriculum, which might include video, and 2) everything else.
I think Google would be great at the first one. I'm not sure how good we'd be at the second part, but I totally agree this is an important problem. My dad was a physics professor for 30+ years, so I've seen a lot of issues with the current model. I have to think there are ways to educate people better than we do it now. You can't tell me that in 20 years we'll still be using the same practices to educate people that we do now.
Yes there are some genuine problems with the idea, but nothing in the realm of impossibility IMO. It will need a reasonable amount of investment and time.
The thing about education system is that everyone knows its a problem and everyone knows it needs to be fixed but no one wants to be the first to jump in to it (I don't think Khan Academy counts).
I think the first effort is important and whoever does it first needs to set an example for everyone else to follow.
pavs, that was a very interesting set of suggestions. are you interested in this space? if so, would love to get in touch as I am thinking of doing something in it for my next venture.
pavs, that was a very interesting set of suggestions. are you interested in this space? if so, would love to get in touch as I am thinking of doing something in it for my next venture.
Space colonization. It's the big item on the advanced civilization checklist that we haven't even got a good start on yet. If you want more information, use your AI guided self-driving spaceships to wardrive around and slurp it all up.
First, I'd emphasize that Google absolutely has to win the TV battle with Apple and other would be contenders. To do this, I'd force my way out of the TV content logjam by buying a content provider, much like Comcast's buying of NBC. Otherwise, any effort in changing the game in the TV domain can easily be dynamited by cable operators.
Tech companies buying media companies seems like a bad idea to me. It reminds me of Sony -- remember when they tried making MP3 players that didn't play real MP3s? That was because they owned Sony Music and were in the awkward position of trying to sell MP3 players and records at the same time. Because they didn't pick a side, they ended up losing both battles.
If Google bought a content company, they would be on both sides of the battle as well -- and if they really wanted to liberate the content, then they would have to explain to shareholders why they bought a company only to hurt its revenue.
I don't think that kind of self-defeat is a necessary result of having both teams playing for you. Look at Netflix: They used to be in the DVD rental business, then they went ahead and obsoleted their own business model and are more successful than ever for it. If streaming had failed, they would have still had their traditional role; but it succeeded, and now they're on the cutting edge of the industry. Likewise, a Google-owned media company could experiment with new models that others are too afraid to try.
You're right, Sony never could work the synergy between its deep vertical component companies much. It doesn't have to be that way, though. Problem with the current video entertainment situation, I think, is the difficulty to bring about that first separation from the pact, much like Apple did with the phone companies. But remember that, Verizon laughed at the level control that Apple wanted with iPhone idea, so Apple went to AT&T. OTOH, Time&Warner and Comcast are pretty much monopolies in their areas, so there's no big competitor that a company like Google can turn to to force their hand. On top of that content providers have seen what happened in the music industry so are very weary of giving control (though the music industry is not in such bad shape as people make it appear, e.g. see this article http://www.economist.com/node/17199460?story_id=17199460&...).
So the situation is a chicken and egg problem, someone has to break through. Apple TV and Google TV has shown that this will not happen as a halfhearted effort. Either Apple or Google should bet and bet big on TV and should not shy from spending a few billion, they've got the money.
We don't have any materials that are can be strong enough along the entire length of the strand to use in an elevator, and likely will not for some time.
142 gigapascals, along say 38,000 kilometers, ain't easy.
(Disclaimer, I work for google, but not on the sekrit space elevator team, which may or may not exist)
Assuming Google won't do this, wouldn't it be possible to create a meta version using a browser extension to capture search terms and provide a collaboration/communication UI?
I'm amazed that Picasa hasn't become dropbox quite honestly. It's such a common problem with photos (and files) to want them on multiple computer, and yet Picasa continues to have an outdated web albums concept with only 2 gigs of storage. Gmail has more.
Make search easier for my mom. God bless her, but she's a moron. One idea is to take the related search terms and turn them into natural questions: "Did you mean baseball bats or flying bats?"
Does Google have a Search for Dummies group? I'd start one.
I would put a focus group around improving add-ons for gmail, one at a time. Though I use multiple project management tools, I still use email the most. Having a direct project/todo management tool in gmail would make the life so much easier and no one will have to go and get different accounts for different clients. They can just add a project within gmail. But it has to be at least at par with the current best. This will completely disrupt the project management tool provider industry.
I'd like see Google do something to make me smarter.
For example, I'd love to be able to do something Matrix style, "Learn Complex Analysis" and then Google creates a curriculum and feeds it to my brain. Ideally this would happen during downtime (such as sleeping or on the train).
Brain science seems pretty far from injecting material into the brain, but what about finding ways to increase my comprehension when I read stuff. Reading content is so slow and generally requires my undivided attention. Unfortunately its the only way I know to learn anything non-trivial.
There are a lot of interesting advanced projects (AI, automated driving, etc.), but I would be worried about the search core right now as CEO.
The proliferation of content-mills and SEO companies is really starting to take a toll on search results. I find myself avoiding Google and going to niche sources directly now (like Wikipedia or StackOverflow) just to avoid the Demand Media trash that is everywhere now.
I think Google should really invest in a new approach to rank -- one that could maybe focus on the content itself using AI and NLP instead of link juice.
I agree completely about search quality, as my comment history here and elsewhere might indicate. The majority of my Google searches anymore return very low quality results without some serious query mangling or heavy filtering. At this point I too skip Google completely as much as possible and go straight to the "niche sources".
People like Matt Cutts at Google unfortunately don't seem to care about the extent low-quality sources have gamed Google - Cutts is still even defending Mahalo as far as I know.
cdr, we've taken action on Mahalo in the past when we've discovered practices that we would consider spammy. At the same time, given the impact that Google has on the web it's important that we not take action on a site just because we like or dislike it. We try to make sure that we only take action on sites when they violate our quality guidelines.
Hi Matt. I would love an option to blacklist domains from my results. That gets around the issue of Google picking winners and losers and would make pickiest 1% of users (like me) very happy.
My issue is the increasingly low quality of Google search results. The problem is not with just one particular site, as distasteful as I find Mahalo, but a class of sites. If you don't consider the SEO practices of the content mills that are infesting the majority of Google results "spammy", IMO you should either reconsider what practices you consider spammy or refine the search algorithm to produce higher quality results.
I recall reading a thesis linked on HN that the only technologically knowledgeable users care about search quality - that the average, non-technical google user doesn't care that a result is from low quality content mill, only that they get some sort of information that sort of seems relevant. I think (hope?) that technical users are the leading edge here, and google's reluctance to act on content mills and improve search quality will end up hurting.
Also, I forgot to mention the amount of outdated forum posts that you get with technical searches - that's the other huge problem with search quality for me aside from content mills and content scrapers. If I go directly to Stack Overflow et al to do a search, I can at least be reasonably sure to get an up-to-date answer.
More recent forum results should definitely be weighted higher than older ones, for a start.
If you're not happy with the dates of the results we show now, one option is to click on "More search tools" on the left-hand bar on the search result page. Then pick a date option like "Past week" or "Past month" or "Past year."
I don't feel like I should have to narrow down date ranges, though I end up doing so every time a search returns forum posts. The algorithm ought to know that something posted six weeks ago is much more likely to be relevant than something posted six years ago, but older posts seem to instead have more weight in the unfiltered results rather than less.
google's date features (as matt has point out) are one of the reasons i haven't left google for bing completely. sure, getting forum post search results about coding an API from 2006 are usually useless, but in bing i can't even say "within the last month." google wins there, currently.
that would require a fundamental paradigm shift from what they currently use to determine the ranking order of search results. Backlinks as a ranking factor will remain a core feature for the indefinite future I would imagine.
I would say they will continue to pursue social/local ranking algorithms. What will most likely happen is that the idea of an objective "high ranking" won't really exist in the future, as all rankings will be personalized to your own (and social network's) habits/interests/history. Combine that with their pursuit of increased speed everywhere, instant everything, and live updates, and you'll be seeing a very different kind of search engine in even just a few years.
Even a modicum of human intervention would help far more than any possible investment into more automated ways to address this problem would. At some point, Google just needs to gird their loins and bring the banhammer down. Yeah, that has problems. I could write about them at length. But that's not the question; the question is now which approach has more problems. I'm willing to try the banhammer at this point. (Though an accompanying investment into Real People that we could actually reach in case of error would be nice, too.)
Google is definitely vulnerable in terms of search result quality.
They have tremendous market share, but search result quality has been stagnant or even declining for a while now.
Page previews are a negative for me, they're distracting and annoying, and every time one comes up I inevitably think "I wish they'd spent this time improving the results".
Start a new company called Page Brin Consulting and offer outrageously priced (aka highly profitable) business consulting services, similar to what Microsoft and IBM do.
If I were Google's CEO, I'd build a transparent, Internet-driven political election infrastructure. Why? First, having regular, fair, accountable elections is a problem that all countries in the world face. Second, the problem I just described has a technical solution. Third, only a company with tremendous resources like Google could successfully battle the entrenched, questionable powers that control many of these elections and the vote counting therein.
What's in this for Google? Users. Data. Brand awareness with every dang voter in the world. Incredible PR. Continued support of that whole 'do no evil' thing.
Of course, there are some catches here. The system has be to completely open. As a citizen, I should be able to audit every part of this whenever I want. Second, you'd have to abstain from all political contributions. People counting the votes can't have a preference or a vested interest in one part over the other. I'm sure there are more that I haven't thought of.
It'd be a tremendous investment on Google's part, but it has the potential for some truly world-changing technology.
Nitpick: even Google can't solve 'watchful husband' and 'untrustworthy administrator' problems. I prefer my electoral systems to be distributed, cumbersome and adversarially supervised.
may be spend a bit of money toward making human species to look ... err ... more humane. Several millions of dogs and cats are euthanized annually in the US alone.
80 comments
[ 5.0 ms ] story [ 134 ms ] threadPush hard in developing AI.
Seriously. Focus on ML and NLP, use their vast resources of data and computing power and really smart engineers. It'll take years, maybe decades but of course this goal is worth it.
* The risk of interacting with the first AGI is even harder to put an upper bound on. You might destroy the human race.
* Statistical NLP and ML might be useful "addons" to an AGI but they're unlikely to lead to its creation on their own.
* Google has a lot of really smart engineers and computer scientists, but AGI is neither an engineering nor a CS problem.
I'd love to hear other HN thoughts on products we should be doing.
How about an app that lets Google Books fans read on an iPad or on Android that doesn't involve looking at HTML rendering and/or converting files to ePub?
I know Google made Donations to Khan Academy. But in my view Khan Academy solves (at least in its current format) only somewhat partial problem. There is a huge potential on what can be achieved in Online Education.
Hypothetical Online Google University can:
- Liberate Higher Education from the grasp of the lucky few.
- Anyone can spend any amount of time to learn on a subject/class/course online, wake up the next morning walk up to an "Exam Center" and give exam to pass the course (with a reasonable fee).
- "Socialize" education in the sense that teachers and students from all over the world can communicate in "forum/chat/discussion" to help each other to study and learn.
- Ability to setup "meetups" to form study groups IRL.
- Study materials can be wiki-fied. With only verified expert on the field can contribute, with diffs to see what has been recently changed. No more waiting for book revisions.
I know there are some genuine limitations to this concept and some problems to be solved, before it can be viable replacement of what can only be classified as "clusterfuck" that is the present education system. But if any company can help the ball rolling, I honestly believe Google can be the only one. If I had the money I would drop everything I have today and jump in to this.
We need to liberate higher education from its artificial barriers.
Hopefully, you will read this and take this in to consideration and discuss it with whoever is concerned.
I think Google would be great at the first one. I'm not sure how good we'd be at the second part, but I totally agree this is an important problem. My dad was a physics professor for 30+ years, so I've seen a lot of issues with the current model. I have to think there are ways to educate people better than we do it now. You can't tell me that in 20 years we'll still be using the same practices to educate people that we do now.
The thing about education system is that everyone knows its a problem and everyone knows it needs to be fixed but no one wants to be the first to jump in to it (I don't think Khan Academy counts).
I think the first effort is important and whoever does it first needs to set an example for everyone else to follow.
If Google bought a content company, they would be on both sides of the battle as well -- and if they really wanted to liberate the content, then they would have to explain to shareholders why they bought a company only to hurt its revenue.
So the situation is a chicken and egg problem, someone has to break through. Apple TV and Google TV has shown that this will not happen as a halfhearted effort. Either Apple or Google should bet and bet big on TV and should not shy from spending a few billion, they've got the money.
Contrary to popular myth, being a successful CEO is extremely difficult, and it requires skills that most people (myself included) don't possess.
142 gigapascals, along say 38,000 kilometers, ain't easy.
(Disclaimer, I work for google, but not on the sekrit space elevator team, which may or may not exist)
Making your searches and profile public in this way would be optional obviously.
It's only doable with extremely high volume search traffic. Guess what, that's exactly what Google has.
Some way of chatting with other people who are searching for the same thing as me would be like a extra-real-time Quora.
Does Google have a Search for Dummies group? I'd start one.
have you tried getting support from Google? /mashes keyboard
For example, I'd love to be able to do something Matrix style, "Learn Complex Analysis" and then Google creates a curriculum and feeds it to my brain. Ideally this would happen during downtime (such as sleeping or on the train).
Brain science seems pretty far from injecting material into the brain, but what about finding ways to increase my comprehension when I read stuff. Reading content is so slow and generally requires my undivided attention. Unfortunately its the only way I know to learn anything non-trivial.
The proliferation of content-mills and SEO companies is really starting to take a toll on search results. I find myself avoiding Google and going to niche sources directly now (like Wikipedia or StackOverflow) just to avoid the Demand Media trash that is everywhere now.
I think Google should really invest in a new approach to rank -- one that could maybe focus on the content itself using AI and NLP instead of link juice.
People like Matt Cutts at Google unfortunately don't seem to care about the extent low-quality sources have gamed Google - Cutts is still even defending Mahalo as far as I know.
I recall reading a thesis linked on HN that the only technologically knowledgeable users care about search quality - that the average, non-technical google user doesn't care that a result is from low quality content mill, only that they get some sort of information that sort of seems relevant. I think (hope?) that technical users are the leading edge here, and google's reluctance to act on content mills and improve search quality will end up hurting.
I appreciate the direct response, though.
More recent forum results should definitely be weighted higher than older ones, for a start.
Every tweak you require me to apply in order to get the results I want is an area in which someone else can beat you.
I would say they will continue to pursue social/local ranking algorithms. What will most likely happen is that the idea of an objective "high ranking" won't really exist in the future, as all rankings will be personalized to your own (and social network's) habits/interests/history. Combine that with their pursuit of increased speed everywhere, instant everything, and live updates, and you'll be seeing a very different kind of search engine in even just a few years.
They have tremendous market share, but search result quality has been stagnant or even declining for a while now.
Page previews are a negative for me, they're distracting and annoying, and every time one comes up I inevitably think "I wish they'd spent this time improving the results".
First thing I would do....Hire someone better than myself and spend my time getting deeply involved in the details that make the company unique....
What's in this for Google? Users. Data. Brand awareness with every dang voter in the world. Incredible PR. Continued support of that whole 'do no evil' thing.
Of course, there are some catches here. The system has be to completely open. As a citizen, I should be able to audit every part of this whenever I want. Second, you'd have to abstain from all political contributions. People counting the votes can't have a preference or a vested interest in one part over the other. I'm sure there are more that I haven't thought of.
It'd be a tremendous investment on Google's part, but it has the potential for some truly world-changing technology.
http://www.aspca.org/about-us/faq/pet-statistics.html