Feature Friday: Google Now Stories

Comments (Archived):

1. Have you tried wildcard? https://trywildcard.com/This app is awesome and does not get the credit it deserves.

   1. hmm. android not ready yet. gave them my email to for the beta access. it looks promising. thanks.

      1. Just saw that Now’s remembered where I parked my car yesterday. Actually wrong address by 2 houses. Ha! Have to see if I don’t have maps running whether it continues to track my movement. I assume it does. **Oops, meant to put this under your point above about privacy….

2. Let me be the first to ask the daily question: is Google an AVC portfolio company?

   1. It may be a Fred Wilson portfolio company.

      1. I was just messing about Dave, following on from a similar question regarding Square yesterday.

         1. I got the subtext.

            1. Attaboy.

            2. Let’s have another political post. More interesting comments.

      2. he’s Google’s lab rat

         1. We all are.

            1. i’m the boy in Google’s golden cage. i can leave any time i want to 😉

            2. not if you use DuckDuckGo, and iPhones.

            3. and yahoo mail, and firefox, and mapquest, and vimeo, and try to avoid connecting with people who use google services (as well as businesses that advertise on google)

            4. I use Bing and an iPhone, but I have a gmail account, so Google gets me there.

      3. it is. see my comment above

   2. I see this post more like Fred trying to figure out Twitter 🙂 – talking about portfolio companies

   3. i wish. but my wife and i do own Google stock. we bought it recently when they changed to the Alphabet structure. our entry price was something like $640. how is that for disclosure?

      1. I thought you owned Google stock for ages. Just goes to prove the old adage about people broadcasting their buys but not their sells.

         1. The twist on that that I learned as a kid was something like “eh…he tells you all the stocks that he makes money on but not the ones he loses money on..”

      2. I suppose if you’re going to invest your money in something might as well be something you like and understand. 🙂

         1. Invest in something that you have an edge on over someone else. To the extent that you like something that can be a benefit, but also a detriment because perhaps you can’t be as objective (I am thinking of some of Fred’s music investments).Of course otoh if you like something you are more likely to understand it and consequently see what needs can be filled that aren’t. (I wouldn’t be able to do that with music or with sports). But contrary to that point (taking the other side) not knowing something also might allow you to see a niche that someone knowledgeable takes for granted (dropbox comes to mind as in “why use that when you can just use rsync”).As far as actually operating a business (as opposed to passive investments) anything that makes money will be fun. This idea that you will enjoy more making money by selling boats or operating a golf course if you like those as hobbies, well, I don’t buy into that at all. In fact it’s also possibly a way to kill something that you enjoy by mixing it with potential negatives and pain.

      3. I have to admit that it’s unfortunate to sit on the side and own no stocks at all while the market is rising to unseen heights. But I just can’t get by the basic concept of discipline to my principles of not being jealous and only putting money into things that I fully understand and have an edge on as opposed to things that are beyond my control and that are subject to forces unknown.But the most important reason is this. Every time I might think “yeah that is worth the risk take a chance”, like in a casino, if I do that enough there will be counter cases that account for losses. My dad did pretty well with stocks but he was retired and spent a great deal of time thinking about it and following certain patterns. I think it’s difficult for a person without that kind of time (or who is not a professional) to duplicate that. Over time. The house has the advantage over time unless the tide is constantly rising. (And infact I thought you had mentioned that you weren’t doing much stocks so I assume this is just some diversification or trivial to where your other money is. I just hope it’s not driven by the shoeshine boy mentality of not losing out..) My dad also did ok with real estate but to my point he did the best and hit it out of the park in a small way in areas which he knew very well, and had the beat and feel of the market, as opposed to areas that he had no special knowledge of.Interesting that Warren apparently doubled down on IBM and analysts polled overwhelmingly think he is nuts.http://www.forbes.com/sites…Warren of course unlike “jenny six pack” gets to hear the dog and pony show so he has more info than a regular investor does. Ditto for you perhaps with google. (Not saying the info comes from google either but your private info stream is a bit different than the rest of us).

         1. Well, that’s why the so called fin experts recommend a diversified portfolio, whether it’s equities, bonds, fixed income securities, commodities, RE, cash, etc., based on affordability and one’s risk tolerance. I trade equities for myself, and generally do pretty well, but over time I’ve found the best and safest strat is investing in S&P index funds. Large index funds have advantages and generally provide more stable risk/reward tolerance. I certainly could have done better w/ a more aggressive investment strat but as I get older I’m much more about capital preservation.

            1. Well then what kind of edge is that. And doesn’t something like that just track the general market?Also this:http://www.marketwatch.com/…(And I am sure this is like anything else subject to many opinions pro and con so perhaps the skill comes into learning enough to know whose spin to believe?)…

            2. “Edge” is a relative term. I could have easily pursued a more aggressive investment strat that more than likely would have yielded higher returns given market performance, but that strat would have been inconsistent w/ my level of risk tolerance. (One can always identify de facto a comparative investment that performed better or worse than the one you made, but that’s water under the bridge.)Years ago I managed the advertising for a fairly large mutual fund company. We did multi-city focus groups with various brokers and financial planners. In total I prob listened to more than 75 so called “financial experts.” Not to indict a whole industry, that would be wrong, but the experience left a bad taste in my mouth in terms of broker reliability, confidence, etc. True objectivity (e.g. underwriting) and churn was always a glaring issue for me.

            3. Yes, it will just track the market. But that’s exactly what you want (~10% returns historically, while requiring ZERO cognitive overhead), unless you have somewhere better to put the money. Warren buffet said it best (paraphrasing): “we’ve always considered our cost of capital to be the second best investment idea. Our current investment idea has to beat it.”As long as you have a decade+ investment horizon (which you should), an S&P 500 index is a great place to park $ you’re not actively investing. Then when you come across an idea you love, where you have an edge, you can move some or most of your capital to those oppty’s.I would only advocate holding cash to someone that is spending their full time investing and has a value strategy of actively hunting for cheap stocks and then loading up on those buys. (Otherwise your dry powder will spend too much time just staying there as dry powder, not going to work for you)

            4. Zero cognitive overhead is important point thanks.

            5. Exactly. The average person would be better off just putting their money in a low cost index fund (Vanguard for example) that tracks the S&P 500, investing a set amount every week/month/quarter and reinvesting the dividends. Assuming the fund actually tracks the market closely, given a long enough time horizon (10 years +), it’s likely to actually outperform if you reinvest the dividends.

            6. Take a look at the hedged portfolio method on my site and see what you think. The basic idea is to rank every hedgeable security (stocks and exchange traded products like ETFs) by its potential return over the next six months and by the cost of hedging it over that time frame, and then buy and hedge a handful of them that have the highest potential returns net of hedging cost.The goal is to get competitive returns while strictly limiting your risk.

            7. I took a look and my head is spinning. The basic concept looks good but I don’t understand enough to know fully what I am doing and I don’t have any time to study what you are presenting (which of course I have heard and know about) in order to feel confident about doing this. If I had the time I would study up on this and perhaps give it a shot. But your site is written to the level of people who are already doing this and know how to use it to their advantage.A step in the right direction of course would be a tab [1] showing a complete newbie example (including “so sign up for a Schwab account first and then…”) but even that takes time to read (the “101 course”).The other thing I always look for is a strong about page. For example your site says it was developed by you but doesn’t explain why anyone should trust what you are doing or suggesting. While that is not a big problem for me (the learning curve is) I am sure it might be for people who don’t know you. But I do like the concept I will say that.[1] Labeled “is this confusing to you or your Aunt Wilma then read this”.

            8. Those are all legitimate points.It’s challenging to write precisely about a technical topic and keep it at a complete newbie level. I’ve erred on the side of being precise and thorough there, but another tack would be to take the Wall Street approach: more descriptive, colorful, and illustrated in the front (though not quite at complete newbie level), with 100 pages of CYA prospectus in the back.Ultimately, I think, the Wall Street approach is the way to get traction with this, but it needs to be sold as a packaged product instead of DIY: You invest in this, and we’ll guarantee you won’t lose more than X% over the next six months if the crap hits the fan, and we’ll try to get you the best possible return while guaranteeing that. Aunt Wilma is more likely to buy that than do it herself in a Schwab account.The “why I should trust you” objection is another good one. I had that objection myself, and took over a year off from promoting it (“promoting” is probably a generous description of my poor efforts at that) and worked on backtesting it to make sure it worked. Of course, that raises the question of why you should trust my backtests: after all, anyone can draw a line on a graph.My initial answer to that was going to be a backtesting tool where you could pick any starting and ending date within the time period of the data (1/2/2003 – 4/30/2014) and run ‘live’ backtests, showing the returns and holdings. That turned out to be infeasible. It would have required full time use of an industrial server (which I had temporary access to during the year of backtesting), which was too expensive.What I ended up doing instead was posting interactive backtests (which you can find about two thirds of the way down this page, depending on your screen resolution: https://portfolioarmor.com/… ). The starting and ending dates of the portfolios are fixed, but you can see every stock and option position throughout the duration of the backtests.But, those are still backtests. Better than backtests would be to run real money with this approach and post audited returns. But I don’t have the money to do that. So, I am going to “front test”: post hedged portfolios regularly (e.g., below), and then follow up six months later with the actual performance.A hedged portfolio for an investor unwilling to risk a >2% drop over 6 months. Data as of 11/19/2015 $ESRX $VOD pic.twitter.com/aNg7Zntf97— Portfolio Armor (@PortfolioArmor) November 20, 2015

         2. When Google was going to IPO I had asked my parents to give me access to whatever inheritance I would get. I was going to sell it and buy Google stock. That didn’t happen …Prior to that cable internet was just starting to get traction and I had wanted to buy stock for it. Within a year it went up 80%, within a few years was up 200%.Now I just invest in myself..

      4. Thought you owned no public names….

3. i’ve been thinking about what you said about newspapers. the “top down” verses “bottom up” process, and twitter as the new newspaper. is the moments product native to bottom up?moments feels like a detour from the roadmap.

4. I like when it reminds me to pay a bill and when it shows me sports scores I’m interested in.But I’m curious about knowing how do they do this? There is a level of Inbox and other browser history intrusion that takes place, and I’m not sure if I like it or just need to accept as the price for getting that predictive info.

   1. My understanding is they are mining your inbox. Apple does it as well but the approach is different (cloud processing v. local).

      1. I’d like to learn more about how they do that.

5. Is this comment useful ?If upvoting and downvoting are driven by agreement to sentiment (ie fair play its an opinion) you cannot argue (one expects opinions to diverge and a contrary opinion can still be welcome)But if Disqus had an “is this comment useful” button – would it mean (is my opinion valid?)The difference being that commentators have feelings where algorithmic recommendations do notI fear unintended consequences[edit for clarification]

   1. Agreement and respect are separate variables; perhaps they should also come with separate up/downvotes?

   2. I am not going to pretend I understand the comparison between “is this comment useful” button and like buttons in the context of Disqus.I this context, and FB had a similar feature on newsfeed, it is clear the app wants to know if to keep algorithm x (that returned a piece of info) in relation to user y

   3. Useful may imply agreement or disagreement. Interest may imply like or dislike. Disagreement and dislike are not properly captured (in absolute or relative senses) and therefore all aggregation and voting systems are fundamentally flawed. I may like a subject but dislike (disagree with) how it was presented or written.

      1. A Claim: That a user indicates a like definitely really is significant data and, in particular, should be useful for estimating what else they or someone else will like.We can think and discuss all day what truly, really, actually, precisely like means, ways with reliability and validity for measuring like, what like has to do with true, false, up/down votes, hearts, smiles, frowns, respect/contempt, entertaining, boring, irritating, insulting, infuriating, fact, opinion, emotion, reason, agree, disagree, Mary’s artistic taste or not, Joe’s interests in computing, finance, or whatever, etc., but, still the claim stands to hold.I thought about it and am counting on it!Or, mathematically that some user Mary indicates a like for some instance of content is a random variable. Truly random? I would never say such a thing — and it is not necessary to. Does it have a Gaussian distribution? I doubt it, but maybe so, but, really I don’t much care. Instead, need to understand the mathematical foundations of random variables. In this way, we can get values of a lot of random variables. Then there are some things we can do with random variables, e.g., we can ask if they are independent or nearly so.Such things are not computer science and not really software and certainly not coding but are relevant to information, and that can be valuable. Yes, with some trillions of like values, some computing and software can be crucial for doing the data manipulations.

         1. With incomplete “information” sets we can draw terribly wrong conclusions. That’s why it’s so easy to revise history the moment after it occurs (ex poste); because more often than not people/historians do a poor job of knowing all the necessary information and certainly not in an objective fashion. Need to accept this in principle and assume that this has and always will happen. My point is the information drawn from the current process is far far from complete, and therefore the premise that one can “target” better ex ante is open to question.As it relates to history, subjective interpretation often enters because people want to “believe” that things happened for a reason; and not give into uncertainty and the law of unintended consequences or exogenous factors. That’s another problem with big data analytics.

            1. Yup. You are correct.But, and a little off topic, as we know it’s still possible to get results as solid as physical and medical science, and, while occasionally those results do get revised, say, gravity from Newton to Einstein, in relative terms they do seem quite solid. E.g., one of the fundamentals of the scientific technique is that results be reproducible over time, and that filters out a lot of junk.Setting aside science and, instead, as you did, considering history, etc., again, you are correct — yup, what we get from history, etc. is not perfect. And, yup, sometimes what looks like small errors in information can result in big problems in applications of that information.But, there is a big world of information that is often useful but not perfect and between (A) perfect information and (B) no information at all.Yes, we can insert a big theme: Often we can’t get to 100% but still can get to old Ivory soap, 99 44/100% correct.And for information that is not 100% correct, there are some impressive results in math that, with some surprisingly meager assumptions, as we collect more and more information, our conclusions will converge to the right answer — e.g., the astounding martingale convergence theorem. From those results we can expect that in some cases we will get improving accuracy as we collect more information.What I am doing with random variables about like makes sense intuitively and mathematically and looks powerful and promising to me: Right, the results will be approximate, not perfect.The work is not blue sky or just abstract nonsense; instead I have good software for all the mathematical manipulations, data in SQL Server, and a Web site in front of all of it, with the software apparently ready for at least early production. If I will quit posting at AVC, I will more quickly get into the data base some good initial data, which I have, and start alpha testing!But, reading AVC does let me see some of what else is going on, e.g., what Fred explained today about some of what Google is doing. From such reading, at AVC, Hacker News, etc., I’ve seen a little on a lot of work by others, but so far I’ve seen nothing at all competitive with what I have done. I doubt the other guys will think of my basic business idea or duplicate or equal the corresponding crucial, core applied math derivations.You mentioned big data. Of course, that topic is heavily a recent fad with a lot of hype. Four things are not clear: (1) What are the interesting sources of big data? (2) What important, valuable information do people want to get from big data? (3) What approaches, e.g., applied math manipulations that are useful and why, do people want to use to get the information from big data? (4) What is really new, correct, and significant about the work in big data?I’m not against big data: My back of the envelope estimate is that my startup should grow to about 150 TB of data. Shockingly, now could put that data in a briefcase, pick it up, and walk off with it — e.g., as Samsung 16 TB, 2.5″ form factor, NAND solid state drives! That the drives are so fast also helps. That the drives can wear out after too much writing mostly doesn’t hurt. Amazing world we live in!But the applied math I’m using on my case of big data is unique — e.g., won’t find it in any PDF files, YouTube lectures, motel seminars, vocational training classes, conference presentations, etc. Why? My work is too specific to details of my startup. Would be like expecting that a part from a Ferrari transmission would apply widely to BMW, Mercedes, Ford, Chevy, Cadillac, Chrysler, Toyota, etc.My view is that often that situation, that is, the need for techniques not general but quite specific, will be common, maybe usual, in effective exploitations of big data. In that case, the big data people will need not just big disk drives, fast fibers, racks of computers, lots of infrastructure software, and lots of common algorithms but also some math, likely at least in part original. It stands to boil down to some applied math problems. I’ve been saying such things for months at AVC, but not many people will enjoy hearing such a thing, especially when currently the big deal in computer science is supposed to be neural networks for recognizing, say, faces and handwriting. E.g., yesterday I downloaded a PDF of several hundred pages on machine learning. I glanced through the material — it was relatively mathematical, but the prerequisites were not very advanced or especially powerful. Point: Reading that PDF would provide essentially no progress on work that would duplicate or equal mine. Net, it is not promising to have a cookbook of tools for the best work in big data.

            2. If you can, drop me a line at michael at ivpcapital dot com. I’m interested in cross-platform horizontal exchange models that afford an infinite number of vertically complete solutions and are far more generative and sustaining than today’s advertising driven and privacy invading informational stack consisting of an infinite number of islands/silos. The opposite of the winner takes most model; which is a repeat of 1913. Have we learned nothing from history!?!

   4. isn’t that what the upvote does?

      1. No – A Disqus upvote either expresses agreement, or it expresses respect or both.But for example (picking some usual suspects:) @JLM or @LIAD sometimes make brilliant points, but I may disagree with their conclusions – Do I upvote or downvote?In Google – voting implies information content but not respect (there is no argument)In Disqus it does seem to imply respect – therefore does downvoting imply mere disagreement or also disrespect ?I very rarely downvote for exactly this reason – it is problematic (as others imply)It is therefore broken.

         1. An interesting point. I don’t have data on this but I think people generally view up votes as capturing a larger range of possibilities than down votes which tend to be reserved for comments not conducive to respectful conversation.

            1. I entirely agree – and better safe than sorry [or a miserable git 🙂 ]

         2. No – A Disqus upvote either expresses agreement, or it expresses respect or both.An upvote is also a PR ping. Commenter’s avatar/name appears when others check to see who has upvoted. [1] So in that sense it promotes a greater sense of community as an unintended consequence. It also gives additional juice to the person who has made the comment by providing group support and feedback. Which encourages even more upvotes. (Some people don’t want to be the first one to arrive at the party..)For example I saw Stephen Voris upvote this (and other comments) and he is new to AVC. So his name and avatar are becoming familiar to me now and I am more likely to pay attention to his remarks as well. Same thing happens when people upvote my comment. I am more likely to pay attention to what they say.[1] At least I do this so I assume others do as well.

            1. “People don’t notice you’re there, they notice you’re still there.” Though granted, the picture’s new, and most of my comments have been talking with Twain and sigma.Anyway, though, to add on to your upvote-as-PR comment – not all upvotes are created equal, so to speak. One from a regular has a different, deeper impact than one from a guest, at least in here; I suspect where guest upvotes show up, well, regularly (say, in the hundreds), they might do something more than just pad the numbers.

            2. > [1] At least I do this so I assume others do as well.Yes but not religiously

      2. They could easily add more granularity in the upvote/downvote scene. It’s lame and as expected binary as opposed to analog. What you expect from computer types or people interviewing politicians “so would you put troops on the ground or not?”. Nothing (except labor and creativity) to prevent them from making things more interesting by giving multiple easy ways to agree or disagree with something (granularity) rather than just one way “upvote or downvote” which is entirely open to interpretation as to what that actually means. [1] Of course they aren’t going to do that because they have made literally no user facing changes in the product for several years or longer and my guess is that they have no intention of making any improvements. Products needs to be continually improved, even if they are free products. Disqus is stuck in rev 1.1.[1] Because you know it makes total sense that the top comment on a particular day should be something that is funny that is posted first thing in the morning.

6. If only Now had Safari’s ‘reader view’ it would be wonderful.

7. It seems to me that what you might call an info concierge — a conglomerater of news, blogs, and social media that’s smart and learns about you — could threaten so many business models if it was done right. Google is smart to be working along those lines.

8. There was a company in Chicago TechStars called “Hooks” that did similar. http://www.gethooksapp.com/…

9. The cards look like Twitter posts

10. I only like Google to Google.

11. Usually, recommender systems ask for feedback when they are uncertain about a particular recommendation. Sometimes, they will also ask if an example doesnt match a preexisting pattern or is identified as being “very informative”. The machine intelligence that selects when to ask for user feedback is fascinating, I think it can even be more interesting than the recommendations themselves. There is a sub class of machine learning called active learning that explores these concepts.

12. No Zingsanity news!!! ;p

13. I love Google Now, it is so useful, gives me current traffic conditions from work to home, even knows on a Sunday morning I take my son to play rugby. The voice recognition is excellent from making a call, getting it to navigate or play a track on Google Play. I can see Google Now being integrated into a home automation system.

14. Thanks for posting this – I had the app but I didn’t “get” it – I had looked at it quickly and never thought to scroll down – I didn’t immediately see the difference between Google Now and Chrome! This is really cool.On another note, I hate to be a disclosure dork but it may be worth considering a simple standardized blog disclosure about whether you own the stuff you write about – to avoid all this noise, which distracts from your point. Maybe a little star at the bottom of the blog along the lines of “*I do / do not have an ownership interest in X.”

15. Their scale, pervasiveness and sheer dominance is awe-inspiring. Yet a little disturbing as well. Google feels very much like a monopoly. They have sooo much data and I hear they grab 10 cents of every advertising dollar spent on the planet. Seems more like a public utility than a competitive business. Yet, like the end of 1984, I love big brother too. A tad creepy.

16. There is a deep point here.Google Now depends on access to the kind of data that Apple is saying that it doesn’t believe a company should exploit as a revenue stream. Tim Cook has emphasized this point repeatedly.Clearly, Apple’s positioning of Google’s data capture is an attempt to set itself apart from Google by undermining consumer trust in Google – google as ‘big brother.’But here’s the problem I see for Apple with this positioning. To the degree that they truly don’t access such data they will be unable to provide a product like Google Now. ( I made precisely this point in an online ‘discussion’ on Benedict Evans’s site.)Here’s the question – do consumers care enough about day to day privacy to forgo better relevance/user experience? Google thinks not. They will try to firewall to improve privacy at the same time as they data mine to increase relevance and hence customer value.Google is a data science company above all else. Products like ‘Now’ are the tip of the iceberg. The question that Apple is going to have to face is, can they maintain their espoused position on privacy? Because access to data and the resultant analytics isn’t just a way of selling ‘your’ data to advertisers, it is a means to better relevance.Fred likes Now. I like Now. And I believe consumers in the large will, like us, share this appreciation for relevance.Long live data science.

    1. Yes. Relevance (including relevant none-sense); The only thing that actually matters online. The whole point of Google. 🙂

17. I wish Twitter would enable me to swipe tweets I don’t find interesting out of my feed, and then use machine discovery to find me other people to follow, rather than to be dependent on… I don’t know what.