I worked in Amit's organization for over four years and it was great. He was brilliant technically both in terms of writing the main search algorithm that replaced Sergey's janky code, and explaining tricky principles of designing search engines to other people as the search team massively grew. One conversation we had about when it made sense to take the square root of things, which I still recall which whiteboard it was on, in my noob-cubicle in building 43, was really the time when I realized, aha I get how to design search algorithms now. It was also always just a very fun, idealistic, and interesting team. And he was good at corporate politics and defended his minions well ;-)
Oh also Amit clued me into the fact that Ardbeg is the best scotch.
This is certainly a loss for Google, but the guy was there for 16 years, so it's probably just time for a change. I will cross my fingers hoping something cool comes next....
Square root of relevance signals to "smooth" them out? I worked a bit on tuning search results, not at Google scale though and it's long time ago, but taking square root, or even logarithm of out of scale signals is a very common trick. Heck, even with the classic tf-idf signal, there are a bunch of heuristics like capping off tf, taking log of idf, penalizing doc length, etc. It's really a trial and error thing.
Yeah exactly - both that square roots are good for smoothing but more so that a it's trial and error thing. When I started out after dropping out of grad school I was working on search algorithms and a bunch of my coworkers had a background in informational retrieval and I did not, and I was kind of thinking about it in the wrong way. I was thinking about it like it was a math puzzle and if I just thought really hard it would all make sense.
So one day I was stumped in how to make this signal useful and Amit suggested, hey why don't you take the square root. I was like, why a square root, it doesn't make much sense, nothing is getting squared ever, and he went up to the whiteboard and just drew a square-root-ish-wiggle-arc-shape and said look you just want something that looks like this. Square root, log, whatever.
I was like, oh... am I allowed to write code that doesn't make any sense? I thought I wasn't supposed to do that. And he was just like, well, just don't worry about it, you are overthinking it, you can take all the square roots you want, multiply by 2 if it helps, add 5, whatever, just make things work and we can make it make sense later.
At that point I realized that real world software engineering was much different than research had been, and also that this was going to be way more fun.
When I started working in Information Retrieval this was something that I enjoyed right away. It was quite easy to come up with a metric that could give more value to some things instead of others..
A few years later I did a PhD in Information Retrieval doing the same kind of things, like it was an art and always exploring new metrics and ideas. I do miss those days, but it is hard to find remote work in this area (IR/NLP)..
I keep seeing Information Retrieval. How is it different from general search algorithm and are there grad programs famous for search? The only thing comes to mind is private information retrieval in security.
> ...look you just want something that looks like this. Square root, log, whatever. ...And he was just like, well, just don't worry about it, you are overthinking it, you can take all the square roots you want, multiply by 2 if it helps, add 5, whatever, just make things work and we can make it make sense later.
I don't know much about search, but probably every software engineer should take some physics classes as well. Just to get used to that sort of reasoning, which is perfectly valid in a lot of circumstances.
I started out as a physics major. I remember the first time I had to do problem sets using the small-angle approximation or taking the first two terms of a Taylor series and dropping the rest.
"Here's how you solve it: the angle of the pendulum from vertical is sin(x), which will always be small, so substitute x..."
"Wait, you can't do that! sin(x) isn't equal to x."
"For the purposes of this problem set it is..."
I had a lot of trouble with this when I first started physics. I was used to thinking of mathematics as Platonic Truth, the eternal mysteries of the universe rendered absolute. But it's not - it's just a model, and it's a model that we use because in most cases it approximates reality pretty well. Contrary to the advice of many primary-school math teachers, sometimes it only matters whether you get the right answer rather than whether your reasoning makes any sense.
I kind of agree and disagree with that at the same time. Yes, doing approximations like that is perfectly fine when you are looking to solve a practical problem and can observe that the solutions you found actually works. When you need guarantees though, I would like a rigorous derivation. Taylor approximation is fine, you can derive error bounds from that, and so on. Obviously, this might get complicated, but what do we have computers for? They should help us with that.
A quick glance through it made me think there's a lot of tricks that are commonplace at Google but not mentioned in the text, but I'm not sure how many of them are proprietary, so I think it's safer for me not to mention them. TF-IDF, however, has a whole Wikipedia article about it:
>>the guy was there for 16 years, so it's probably just time for a change
This would make sense to me if the announcement included what he was doing next.
Often "retiring for personal reasons" is code for something they wouldn't want to put in a press release. Especially when the retiree is less than 50 years old.
It's possible. Not typical for the type of person that ascends to that level though. He and Matt were perhaps the two most well known names in that group to the outside public. Both with similar departures (or semi-departures).
The scuttlebutt I heard on him is that he opposed using machine learning and artificial intelligence techniques for search, specifically deep neural networks.
I can understand his opposition, Google search results have been monotonically declining in quality recently in my opinion.
I wouldn't be surprised if this is a polite way to sweep him out of the way. It's pretty much a golden parachute if so.
This is of course all speculation and hearsay based on what I heard about him a few months ago.
As discussed in Marissa Mayer and the Fight to Save Yahoo![0], Mayer's feud with Singhal was responsible for her 'demotion' to Google's Maps/local team, and eventual departure for Yahoo!:
> Singhal loathed how, when his team was ready to roll out a change, Mayer would insist it pass through her UI review—a process that could take weeks.
…
> [I]n 2010, Larry Page decided that Google was moving too slowly. He wanted the whole company to move as fast as its Android and Chrome divisions.
> One impetus for this decision was a memo from a longtime Googler named Urs Hölzle. It reminded everyone that social networks, and Facebook in particular, had become a dominant force on the Internet. It said that Google was nearly blind to all the knowledge stored on Facebook. Hölzle pleaded with his colleagues to focus on social media, or else the company could be swept away in an oncoming wave.
> Page took the memo to heart, and put Vic Gundotra in charge of developing a social network for Google. The project was code-named Emerald Sea, after a Japanese painting in which a boat is about to be wiped out by a huge wave. Gundotra’s team built a prototype in a hundred days.
> The speed impressed Page. He realized that one thing Android, Chrome, and the Emerald Sea project had in common was that each had a single person in charge. He wondered if it was time to put one person in charge of search.
> Gundotra, suddenly vested with influence thanks to the Emerald Sea project, made it known that he did not like working with Mayer. With all her reviews and processes, she slowed things down too much.
> Page thought about the future of Google and its past. To him, the strength of the company was that it developed machines that could learn what humans wanted and then provide it to them. He believed that Singhal better understood how to make the technology that powered those machines. Page believed that made Singhal more capable of pushing Google products to their technological limits…He summoned Mayer into a one-on-one meeting and told her she was done working on Google search.
It's interesting to see printed descriptions of events that I lived through. (I worked on projects sponsored by both of them, and was also loaned out to help with the Emerald Sea launch.) Some are echoes of the truth, but there are also some fairly big inaccuracies.
Emerald Sea didn't have anything close to a prototype that could ship in 100 days. At best, there were a few technology demos from individual engineers. Urs sent his memo in March 2010; Emerald Sea launched as Google+ in June 2011, with the +1 button (done by David Byttow, who later attained infamy as the founder of Secret) launching a couple months earlier.
Vic had few if any dealings with Marissa. She was moved to Geo in Oct 2010; at the time, Emerald Sea was a few scattered tech demos. The push to integrate G+ into everything Google didn't really start until after G+ launched in June 2011.
Larry was impressed by speed, but I think that his exemplars were much more Android and Chrome than Emerald Sea.
Mayer and Singhal had a very confused reporting relationship - basically, the engineers reported up to Singhal and the PMs reported up to Mayer. I wasn't close enough to them to know how well they got along personally, only that they both had very strong opinions and a lot of experience to back them up. It did make for some very slow product decisions as PMs had to get both of them to sign off on a change.
I don't see how that quote indicates that her "feud with Singhal was responsible for her 'demotion'...and eventual departure".
I'll admit that I generally dislike these books that construct a simple narrative out of anecdotes out of years of time, so I'm mostly responding because it's disappointing that this is the only top comment right now.
And (assuming it's true), believe it or not he's not the first person to resent someone bringing their expertise to bear on their baby, regardless of merit.
> In October 2010, it was made official. Mayer was removed from the top of Google’s search organization and put in charge of Google Maps and other “local” products.
> Technically, this was a lateral move, if not a promotion, because Mayer retained her vice-president title and she was, at the same time, given a seat on Google’s operating committee—then-CEO Eric Schmidt’s round table of top executives from the company.
> In reality, it was a demotion. Mayer was no longer in charge of what Google’s most important product looked like or how it worked. At Google, there was search, which generates nearly all of the company’s revenues and profits, and then there was everything else. Running Google search, Mayer was managing the most important product at the world’s most important Internet company. Running Google Maps, she was not.
Heh, so far up the food chain, being in charge of Google Maps is a demotion. If anything Google Maps is a genuinely good product and well used by many, as is search and Gmail. Could have been worse... Google+. I wonder how Googlers internally feel about maps, it should be "up there" next to search in terms of being amazing and well used.
Google probably has a better key to profit via Maps than it could ever hope to get by building a completely new social network. So dumb that it was so desperate to imitate Facebook and Apple that it let its best products languish or die, even though many of them (e.g. Reader) contained social network microcosms.
I consider Google Maps to be far more useful and better than Facebook. The sans/quality of G+ doesn't impact my view of Google Maps. It's just an amazing product, that got even better once it came on those feature phones with pseudo browsers, even if it took a whole minute to load some tiles, it was still amazing for going to appointments in new areas.
I have a lot of childhood memories of my dad in the car messing around in the dark with physical maps only to find a torn page or inaccuracies, heh.
Google Maps has reached its true potential only now that everyone has a smartphone with GPS. I don't think that it was entirely clear in 2010 just how important mobile would be, or that Google would play such a dominant role on mobile devices. The company was still fixated on Facebook's rise, but it's since jettisoned the focus on social networking (after Google Plus failed to take off).
Is that really so? I have actually stopped using Google Maps over the last couple of years as the last few redesigns have made it less useful and more frustrating. I would say that 2010 represents the peak of Google Maps' usefulness, and whatever it has been doing since then looks more like squandering potential than realizing it.
...But I also found Google Maps to be excruciating and effectively unusable on a mobile phone when I first tried it. I don't think I've actually tried using Google Maps on a phone since their last redesign. Perhaps that's what happened, they just switched from a design that was usable on a computer to a design that was usable on a phone, but I'd already given up on the phone experience so I didn't notice.
I don't think it's quite there yet. Google Maps for Mobile still has a ways to go in terms of feature completeness & accuracy, and there are a ton of additional ways search & profile data can be mined & surfaced in Maps results.
Although incremental improvements in data quality and data mining can certainly still improve the user experience (I've personally encountered many frustrating examples of this), their potential impact is clear now that everyone has Maps in their pocket.
The value of any such improvements to end users was drastically magnified by the availability of Maps on mobile devices with GPS. And that shift required many advances to occur independently of Google - Apple and the Android OEMs played a key role.
Just playing devil's advocate here -- sometimes very trusted lieutenants are put in charge of the underperforming divisions. One could see it that way too. That Page trusted Marissa with the ability to lift up the "local" division.
> Page consulted with [SVP of Engineering Bill] Coughran. Coughran said he could no longer peacefully referee what was becoming an increasingly hostile relationship between Singhal and Mayer. Coughran said that a couple times a week, one or the other would be in his office, complaining and asking to lead the organization.
> “We’re going to have to pick somebody here,” he told Page.
The feud was with Singhal. Page only picked the winner. I'd recommend you just read the book (the quotes are from chapter 6).
I worked at Google for a very long time and always wondered if Mayer wasn't in the OC because her and Page dated for awhile back at the beginning of Google. The optics of having your ex-gf in your inner circle seems like it would look bad, corporate politics or not.
Many of the other folks in the OC are super-long-timers and would've been around when Mayer and Page were an item, adding to the awkwardness.
I don’t know if it has anything to do with Mayer/lack of review per se, or if it’s more about the scattered focus of G+, Android, etc., or if companies of a certain size are just inevitably incapable of solid UI implementation, or what, but Google’s UI polish on its flagship browser apps has taken a nosedive in the past 5–6 years. They used to be lightweight, fast, and reliable. Now they are bloated monstrosities packed with weird glitches that devour browser resources.
I think Mayer's direct focus was more on UI than on performance, although a simple UI could certainly lead to better performance:
> One time, an APM brought a product to [Mayer] for review and she told him, “This page is too busy. What you need to do is look at every font on the page, every font size. And every time you see a new color or a new font size, you add up a point. I want this page below five points.”
> The comment about “five points” ended up in the meeting notes, and then it became a rule. No pages with more than five points.
> Another rule was: Design a product for the “98 percent use” case. For Mayer, the best example of a product that followed this rule was the Xerox copy machine. It could do all kinds of fancy things: staple, collate, copy, and fax. But if you walked up to one and pressed the giant green button, the right thing just happened. Mayer believed on every good product there should be a big button like that for the 98 percent use case, where if the user clicks it or taps it, they get a delightful, fluid, simple experience.
> During a product review, Mayer would count the number of keystrokes it took to get every job done. Too many, and it was back to work.
Another rule was: Design a product for the “98 percent use” case. [...] Mayer believed on every good product there should be a big button like that for the 98 percent use case, where if the user clicks it or taps it, they get a delightful, fluid, simple experience.
This is all good, but hereis the thing that many designers with great responsibilty doesn't seem to get: As much as there is one big button/search field etc, DON'T REMOVE advanced options for advanced users. Example: For a long time Google Maps on Android it would be impossible to look up a route from anywhere but where you were now. It was not only streamlined to start with current location but rather neutered so you couldn't. To add insult to injury it used to work fine before someone started their UX work :-/
> For a long time Google Maps on Android it would be impossible to look up a route from anywhere but where you were now. It was not only streamlined to start with current location but rather neutered so you couldn't.
I almost lost my mind trying to do this exact thing.
Google maps refusal to make offline reasonable has pretty much killed my use of it. You can use OSM data pre-download the whole set of countries you're driving through and not worry about poor signal / roaming charges / data use.
Or you can use Google maps and get screwed. Guess it works better if you're in the US.
I can't upvote you enough. This is my #1 pet peeve with modern UI/UX design. I'm perfectly fine with the "big green button" scenario where the most most common use case is the one that gets most of the visibility, but those advanced features NEED to still exist somewhere. Minimalism taken too far can and will hurt the quality of the product.
> As much as there is one big button/search field etc, DON'T REMOVE advanced options for advanced users.
I think Mayer's rule implies that. If you remove the feature entirely, you aren't designing for the 98% case, you're designing for the 100% case because the remaining 2% case is gone.
I interpret "design for the 98% case" to mean "focus on that, but don't eliminate the other 2%".
Ditto...that's what the UI on a Google built page personifies...appeal to the 98 percent, because that 98 percent expects to use a simple tool, either because they value expedience, or because they are unaware of other options available to them...
IMHO, it is so important to factor in the simple fact that ,by definition, 1/2 of the population is below average in intelligence...it's statistically so...
Thus, it rarely (comparatively) enters the mind of those in the left tail of a normal distribution that what they see before them should be easily "tweaked" to meet their precise needs...simplicity suffices...
Those seeking expediency feel the same...simplicity suffices...
> IMHO, it is so important to factor in the simple fact that ,by definition, 1/2 of the population is below average in intelligence...it's statistically so...
Technically, half are at or below median, not mean. As an example, if I have a sequence with 11 values, 1 to 10 and 100. Mean is 15.5, which 9 of the 10 samples fall under. Median, however, is 6.
>Technically, half are at or below median, not mean.<
My use was colloquial, not technical, thus I think the use of "average" appropriately serves to illustrate my point...
If you are into statistics, and want to provide a similar example with sample size n=320 million(approximately), I'd be interested in taking a look at it...
> If you are into statistics, and want to provide a similar example with sample size n=320 million(approximately), I'd be interested in taking a look at it...
1 to 319,000,000 and 1 googol...
But actually, you could make the argument that intelligence is likely normally distributed, so the median and the mean should be close enough to make your statement accurate.
I agree... if it didn't follow a normal distribution I'd probably lose faith in the proven heuristics I've come to rely on to help me parse information...
Thanks for helping keep me on my toes, my friend...we're all here to espouse views and defend them...sling ideas around...spirited debate...that's the beauty of this forum...
These are great product guidelines. Especially the 98% use case. Are these from the book linked in the parent comment? Or do you have a source with more ideas along these lines?
Just remember, as much as you want one big green button, -DON'T remove options if you have existing users, just make them smaller, hide them (but easily accessible) and make sure they reset automatically (at least for noobs).
> During a product review, Mayer would count the number of keystrokes it took to get every job done. Too many, and it was back to work.
Palm once had a tap counter that had the rule: 'If any task on the Palm Pilot takes more than three taps of the stylus, it's too long, and it has to be redesigned.'
I think these are very essential qualities of a head of product that had direct influence on making Google the most valued company that it is; They are very complementary to designing a search algorithm - but you need both to be successful. I admired Marissa Mayer's work when she was still in charge and was sorry when she left. I presume it would have been the better choice to make her lead Google+ - if anything this product had a problem with its "personality". I would have trusted her to get this right.
Do you feel like Google+ is more than a problem in search of a solution? That's how I view it, and I doubt Mayer could have changed that. That says nothing of her and everything of Google+.
I was about to joke that maybe they should do that with Yahoo.com, but then I fact-checked myself first by going to it and yahoo.com is actually far cleaner than the horrible, cluttered mess I remember. Granted, my recollection was that it was 'painful to look at' so it's a low bar, but still.
When did that change? I hadn't been there for so many years that I don't even know any more.
Reminds me of the UI bug I hate most on google search page. The page loads fine but the cursor takes longer to appear in the search box than it should (probably 3 sec at max but long enough to frustrate)
Not questionable at all- it's a gold mine for knowledge of value to advertisers. Every post, like, etc. on Facebook is spontaneously coming from a user. You don't have to infer what people love or hate based on browser and search history, you just let them tell you (and all their friends) what they love or hate. They're literally telling you what you should advertise to them, it doesn't get better than that for businesses driven by ad revenue (not a negative judgement, just an observation). Of course Google would want that data, and of course Facebook would want to prevent them from getting it.
The caveat being that on the level of the user it's more like (love, sort of like, guilted into saying you liked) vs (hate, feel kind of meh about, guilted into saying your hated).
The aggregation of those likes is more useful but at the individual user level it's a terrible signal. Especially since the entire social structure of the network is geared toward getting you to click like whether you actually like it or now.
All of my local recommendations (handyman, local photographer, where to buy tomatoes) have now shifted to facebook groups. I no longer google handyman in blah, xyz. instead I go through my saved posts in FB.
This is along with local city news and news from friends and family that is all on facebook. e.g. One of my colleague's (and friend) kid was in ER last weekend so on Monday during standup I was able to tell the rest of the folks that he was out due to family emergency without getting an email from him...
[All of my local recommendations (handyman, local photographer, where to buy tomatoes) have now shifted to facebook groups. I no longer google handyman in blah, xyz. instead I go through my saved posts in FB.]
In my local area NextDoor is replacing FB completely for any local recommendations, news, Alerts.
My dad went to college with Amit in India - he and his friends always speak incredibly well of him but simultaneously express wonderment as to how it was he who made the rise, given that many others were more 'brilliant', class-toppers, etc. so to speak. I think it's amazingly interesting how true brilliance manifests itself in different ways than we could expect a conventional system to indicate.
A testament to the perhaps under-appreciated role of luck in our life outcomes; those with the most success are not necessarily destined to it by intrinsic greatness, but just by capricious random events swinging their way as opposed to others'.
There's a lot to be said for preparing and positioning yourself to be ready to take advantage of providence when its winds blow your way.
I.e. it'll never happen for you if you're home watching TV.
It's also necessary to be able to recognize opportunities when they come your way. I know I've missed many. My dad would say that most people don't recognize opportunities because they come disguised as hard work.
"Fortune favors the prepared mind." - derived as to still be relevant (I personally think) from Pasteur's "In the fields of observation chance favors only the prepared mind."
It's not luck... if you have ever worked at Google, you know there are tons of smart people there, especially in search quality.
If you don't know your stuff, and constantly learn more, they're not going to respect you. No luck about it. You can't even "luck" into a manager position, let alone VP of search.
As it happens, I have. You are correct that there are tons of smart people there (in my less modest moods, I'd take myself as one such datapoint). What of it? That does not contradict, and therefore did not disabuse me of, the observation I noted above.
I think people generally think luck is more of a factor in success than it is, rather than it being under-appreciated as you claimed.
You might think there's some luck in joining Google in 2001, but that ignores a prior decade spent on information retrieval, which was VERY MUCH NOT a hot CS field -- it was arguably the opposite. It's not luck when you spent huge amounts of time on something you care about, but other people don't care about. And when you happen to join some like-minded people well before they were successful.
There are also plenty of brilliant people who also joined Google in 2001 who never became VP of search. Based on that, I'm saying there is NO possibility that he doesn't have extreme talent. You don't get to the top at Google accidentally.
Extreme success obviously involves both luck and talent, but luck isn't that much of it. Bill Gates got lucky in many ways, but he would have been tremendously successful no matter what. To a lesser degree, I'm sure the same is true of Amit.
If he wasn't viewed as a top student, well that means diddly squat and isn't surprising at all. Contrary to the grandparent, I would NOT expect the top students in the class to become the most successful. People who overoptimize for academic success are generally somewhat obedient... Early engineers who become VPs, founders, and other successful people are not obedient (whatever their other flaws).
This is a really old argument too, which your comment isn't really doing justice.
That's the part that most people call lucky. When the thing you spent a decade working on because you found it personally interesting suddenly becomes the thing everybody else wants.
So I would call that closer to being "visionary" rather than being lucky. In other words, it's seeing something before other people see it. Seeing unrecognized potential is a skill that I believe requires a certain form of intelligence.
You could imagine someone choosing information retrieval without any special insight... but those kinds of people usually don't have enough curiosity or motivation to get to the frontier of the field and push the boundary.
When you say "someone got lucky", you are usually undermining their success, by implying that their personal talents and hard work didn't have much to do with it.
He did not say that. If 100 people work on diverse subjects for 10 years because they have a personal interest in them and one of the subjects suddenly becomes a hot topic, then one of the persons is a winner.
It is possible to reverse your argument and say that you are undermining the success of the remaining 99.
A very wealthy man once told me the phrase, "There is no such thing as luck. 'Luck' is where preparation meets opportunity."
I'm sure he had some random events placed in front of him, giving him an unique opportunity. But he executed phenomenally. And we should not discount that. Many have had the same kinds of opportunities and failed to execute.
That said, I'm always surprised by how people are so quick to discount luck. Or consider a bad thing if they got lucky.
You can be both prepared and have an opportunity and still fail because of factors outside of your control. Likewise, you could be unprepared, flail your way through a solution, and still manage to be very successful at the very end because the stars aligned. Maybe the market was super ripe, your competitors were going through some rough patches, etc.
Success stories are often carefully crafted with a narrative fallacy. I can think of several very successful companies today that would've been sold a long time ago had it not been for the potential buyer rejecting the deal. Or entrepreneurs who have one success, and then were never able to create a majorly successful big startup after.
Luck is embedded into the definition of "opportunity" so the quote is contradictory.
I am invited to an interview for the #3 job at Newgle. There's an earthquake the day of my interview and it's postponed. While waiting to be rescheduled, Barry, the co-founder bumps into a great engineer at a cafe and hires him instead. I never interview. Newgle goes onto become the most valuable company in the country. I stay at Oracle as a senior engineer but never do much else.
>>by capricious random events swinging their way as opposed to others'.
The idea is you better be prepared when they start swinging your way. A lot of time the opportunity is there, but the person is just not ready, like there are job opportunities all the time, but a very few people qualify or are can even do the job.
A lot of opportunities open up if you travel some distance with the work on hand. They won't even be visible to you when you start working. They only come to you after you've done some work.
Sometimes opportunities are also 'next steps' to anything you do. David Allen talks about in detail in his GTD book.
Opportunities multiply as they are seized - Sun Tzu
One part may be that the "brilliant" folks went for the established companies or into research. Amit went for the scrappy startup - that could have exploded just the same.
One of my teammates went to business school with Sundar... yet my colleague is a program manager and Sundar is CEO. Luck/shit happens, but fortune always favored the well-prepared.
After watching the video, to me, it just looks like Amit is quick to be the first to react and speak up. Which makes him appear like a leader to others.
That was a great video, thanks for sharing! I wish Google would have continued putting out these videos, it's somewhat rare to see a technical discussion outside of a conference/talk setting.
Ok, wild guess here but Sundar is cleaning out the house? It's kind of hard to give direction to some who's been there forever, actually achieved more technologically and financially, and more critically would not listen to you.
No wonder Larry decided to create Alphabet, stayed out of all the office politics. I would take a bet that Jeff Dean and Urs Holzle will leave in the next year or two.
I'm surprised the media is billing this as "a push for AI".
I was on both Giannandrea's and Singhal's teams at Google (I left last year). Giannandrea doesn't really have an AI or machine learning background (LinkedIn page: https://www.linkedin.com/in/johngiannandrea). He was put in charge of machine learning relatively recently, IIRC less than two years ago. Before that he was doing Knowledge Graph - he came into Google through the Metaweb acquisition - and that technology stack really doesn't have much AI or ML in it. Google has tons of good AI people; if it wanted to, it could easily have replaced Singhal with any number of engineers with more AI experience.
Google's problem in search has never been a lack of good engineering, its been a lack of vision for search. The fact that someone from the search team isn't leading the company says everything.
I think the IBM's Watson, Wolfram Alpha, Freebase etc were more much visionary than what Google has done in search in the last 10 years. Even Bing's API's are much more useful. The WolframAlpha and Watson kind of NLP queries have been around for almost 6-7 years now. And Google still can't handle most of them. Whats the use of having all the AI experts on the planet?
Search should be its own company split apart from Chrome and Android. The empire defense game distracts from what Search can be,
I think this gets back to the aforementioned point about Marissa's statement re: serving the 98%. Yes, WA & Watson are super useful, but they're useful in ways a huge majority of the population will never miss and thus never know about. Google has been far more focused on addressing the "obvious" problems and low hanging fruit ... it's only a matter of time before they start encroaching, imho.
Sure. But in that case, why interpret choosing him as "a push for AI"? The logic of "Google put an AI person in charge" -> "Google wants to focus the company more on AI" seems mostly valid, but the premise is false, so there's no reason to believe the conclusion (more than we would have otherwise).
It is just spin (so is almost everything you read in the press, fwiw) and as such represents PR people doing their job well. You only realize the spinness when you have inside information that contradicts it.
Congratulations to jg regardless. He's very smart and I'm sure has picked up enough understanding of current AI techniques in the past couple of years. Disclosure: I've known him for 30 years or so..
That part, I'm ready to bet dollars to peanuts, is a talking point pushed by Google PR, which the journalist used (just like the other 99 facts and not-so-facts served to him on a platter all ready to digest by PR) because he's on a deadline and doesn't have the time to look for alternative explanations. That's a lot of work for a very weak ROI.
Combined with Matt Cutt's "indefinite leave", this seems somewhat curious. Pure speculation, but it's somewhat interesting to me that both are the "old guard" in search, and during their tenure, the pure AI sort of approaches, like Panda and Penguin, were implemented outside of the main algorithm, as bolt-ons that were very klunky and separate.
Feels like a move to to clear the way for AI to drive the algorithm, versus running outboard and afterwards as a tweak/tune cycle.
Edit: Er, okay. It's not curious that an < 50 years of age superstar is suddenly retiring for personal reasons?
> Feels like a move to to clear the way for AI to drive the algorithm, versus running outboard and afterwards as a tweak/tune cycle.
Well, to be honest, you still have to tweak and tune your AI model so that it works. Data speaks for itself. You have to run experiments, so the approach they have is going to be the same ANYWAY. AI isn't magical. It may just happen that both sides feel it is time for someone to take a new position. I think that's the better way to put it. But why this new guy not other seniors on the same team? That's a good question.
The tweak tune that I referenced as klunky wasn't the AI per se, but the way Google integrates it into the search results. Awkward enough that they would go more than a year between running/integrating it, and when it was deployed, the search results would swing wildy for weeks.
Go to Netflix and make a list of your favorite TV shows. Then every evening, instead of watching two episodes from your favorite shows, read papers and implement stuff.
It's a secret known only to a select few. I'm in a charitable mood today, and for only $99, I'll let you in on it. (But do it quickly, for I might not feel as charitable tomorrow.)
Does anyone know why the Google search results page had 2 copies of the top navbar for a while? One with a ghastly black background? Was that Marissa or Singhal?
Hope the new guy reveses the recent change to presentation of Google results on iPad.
The so called - Mobile Friendly - is truly horrible - it's such a pain having to - request desktop site - for every single search, shaame there is no auto option for this on safari or even better change user agent in settings.
Some app browsers let you do this, but they are slower and flackier than safari.
I tried Bing because of this, but still no where near in relevance of results for me.
This is certainly a loss for Google, but the guy was there for 16 years, so it's probably just time for a change. I will cross my fingers hoping something cool comes next....