Preskoči na sadržaj
Korištenjem servisa na Twitteru pristajete na korištenje kolačića. Twitter i partneri rade globalno te koriste kolačiće za analize, personalizaciju i oglase.

Za najbolje sučelje na Twitteru koristite Microsoft Edge ili instalirajte aplikaciju Twitter iz trgovine Microsoft Store.

  • Naslovnica Naslovnica Naslovnica, trenutna stranica.
  • O Twitteru

Spremljena pretraživanja

  • obriši
  • U ovom razgovoru
    Ovjeren akauntZaštićeni tweetovi @
Predloženi korisnici
  • Ovjeren akauntZaštićeni tweetovi @
  • Ovjeren akauntZaštićeni tweetovi @
  • Jezik: Hrvatski
    • Bahasa Indonesia
    • Bahasa Melayu
    • Català
    • Čeština
    • Dansk
    • Deutsch
    • English
    • English UK
    • Español
    • Filipino
    • Français
    • Italiano
    • Magyar
    • Nederlands
    • Norsk
    • Polski
    • Português
    • Română
    • Slovenčina
    • Suomi
    • Svenska
    • Tiếng Việt
    • Türkçe
    • Български език
    • Русский
    • Српски
    • Українська мова
    • Ελληνικά
    • עִבְרִית
    • العربية
    • فارسی
    • मराठी
    • हिन्दी
    • বাংলা
    • ગુજરાતી
    • தமிழ்
    • ಕನ್ನಡ
    • ภาษาไทย
    • 한국어
    • 日本語
    • 简体中文
    • 繁體中文
  • Imate račun? Prijava
    Imate račun?
    · Zaboravili ste lozinku?

    Novi ste na Twitteru?
    Registrirajte se
Profil korisnika/ce ali01
Ali Yahya
Ali Yahya
Ali Yahya
@ali01

Tweets

Ali Yahya

@ali01

Partner at @a16z. Investing in crypto. Formerly a software engineer and researcher at @GoogleBrain, Google X, and @Stanford Computer Science.

San Francisco, CA
outlast.me
Vrijeme pridruživanja: travanj 2008.

Tweets

  • © 2020 Twitter
  • O Twitteru
  • Centar za pomoć
  • Uvjeti
  • Pravila o privatnosti
  • Imprint
  • Kolačići
  • Informacije o oglasima
Odbaci
Prethodni
Sljedeće

Idite na profil osobe

Spremljena pretraživanja

  • obriši
  • U ovom razgovoru
    Ovjeren akauntZaštićeni tweetovi @
Predloženi korisnici
  • Ovjeren akauntZaštićeni tweetovi @
  • Ovjeren akauntZaštićeni tweetovi @

Odjava

Blokiraj

  • Objavi Tweet s lokacijom

    U tweetove putem weba ili aplikacija drugih proizvođača možete dodati podatke o lokaciji, kao što su grad ili točna lokacija. Povijest lokacija tweetova uvijek možete izbrisati. Saznajte više

    Vaši popisi

    Izradi novi popis


    Manje od 100 znakova, neobavezno

    Privatnost

    Kopiraj vezu u tweet

    Ugradi ovaj Tweet

    Embed this Video

    Dodajte ovaj Tweet na svoje web-mjesto kopiranjem koda u nastavku. Saznajte više

    Dodajte ovaj videozapis na svoje web-mjesto kopiranjem koda u nastavku. Saznajte više

    Hm, došlo je do problema prilikom povezivanja s poslužiteljem.

    Integracijom Twitterova sadržaja u svoje web-mjesto ili aplikaciju prihvaćate Twitterov Ugovor za programere i Pravila za programere.

    Pregled

    Razlog prikaza oglasa

    Prijavi se na Twitter

    · Zaboravili ste lozinku?
    Nemate račun? Registrirajte se »

    Prijavite se na Twitter

    Niste na Twitteru? Registrirajte se, uključite se u stvari koje vas zanimaju, i dobivajte promjene čim se dogode.

    Registrirajte se
    Imate račun? Prijava »

    Dvosmjerni (slanje i primanje) kratki kodovi:

    Država Kod Samo za korisnike
    Sjedinjene Američke Države 40404 (bilo koje)
    Kanada 21212 (bilo koje)
    Ujedinjeno Kraljevstvo 86444 Vodafone, Orange, 3, O2
    Brazil 40404 Nextel, TIM
    Haiti 40404 Digicel, Voila
    Irska 51210 Vodafone, O2
    Indija 53000 Bharti Airtel, Videocon, Reliance
    Indonezija 89887 AXIS, 3, Telkomsel, Indosat, XL Axiata
    Italija 4880804 Wind
    3424486444 Vodafone
    » Pogledajte SMS kratke šifre za druge zemlje

    Potvrda

     

    Dobro došli kući!

    Vremenska crta mjesto je na kojem ćete provesti najviše vremena i bez odgode dobivati novosti o svemu što vam je važno.

    Tweetovi vam ne valjaju?

    Prijeđite pokazivačem preko slike profila pa kliknite gumb Pratim da biste prestali pratiti neki račun.

    Kažite mnogo uz malo riječi

    Kada vidite Tweet koji volite, dodirnite srce – to osobi koja ga je napisala daje do znanja da vam se sviđa.

    Proširite glas

    Najbolji je način da podijelite nečiji Tweet s osobama koje vas prate prosljeđivanje. Dodirnite ikonu da biste smjesta poslali.

    Pridruži se razgovoru

    Pomoću odgovora dodajte sve što mislite o nekom tweetu. Pronađite temu koja vam je važna i uključite se.

    Saznajte najnovije vijesti

    Bez odgode pogledajte o čemu ljudi razgovaraju.

    Pratite više onoga što vam se sviđa

    Pratite više računa da biste dobivali novosti o temama do kojih vam je stalo.

    Saznajte što se događa

    Bez odgode pogledajte najnovije razgovore o bilo kojoj temi.

    Ne propustite nijedan aktualni događaj

    Bez odgode pratite kako se razvijaju događaji koje pratite.

    Ali Yahya‏ @ali01 12. lip 2019.
    • Prijavi Tweet

    While I was at Google X, I worked on an effort to build a robot hivemind––a network of robots with collective intelligence. Our work was part of a broader strategy to gain a data advantage. Here’s what I learned about data network effects. A mega thread 👇 https://outlast.me/robot-hiveminds-with-network-effects/ …

    06:41 - 12. lip 2019.
    • 357 proslijeđenih tweetova
    • 1.339 oznaka „sviđa mi se”
    • Muhuuh Abridged ⚗️ Sagun Garg (1.01³⁶⁵ = 37.2x) A Ed Sahil Dhingra Spencer McConoughey Jimmy Wang PΛBLO MΛRTINEZ BERNΛL Leo von Kleist
    21 reply 357 proslijeđenih tweetova 1.339 korisnika označava da im se sviđa
      1. Novi razgovor
      2. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        About a month ago, my partners @martin_casado and @peter_lauten argued in a great blogpost that data network effects are mostly bullshit. I could not agree more! Every one of their insights maps directly to some aspect of my experience at X.https://a16z.com/2019/05/09/data-network-effects-moats/ …

        1 reply 13 proslijeđenih tweetova 66 korisnika označava da im se sviđa
        Prikaži ovu nit
      3. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        First, some context: The ultimate goal of our robotics project was to leverage the combination of Machine Learning and Distributed Systems to enable robots to learn from each other’s experience in order to effectively operate in unstructured environments with limited sensor data.pic.twitter.com/vsMGFh2muG

        1 reply 3 proslijeđena tweeta 20 korisnika označava da im se sviđa
        Prikaži ovu nit
      4. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        One important sub-problem that my team and I found ourselves having to solve is teaching a robot how to open a door using this approach. This may sound easy (and it is, if you’re doing it for just a single door), but it’s a hard problem to solve under *all* circumstances.

        1 reply 2 proslijeđena tweeta 14 korisnika označava da im se sviđa
        Prikaži ovu nit
      5. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        This is because there is a very long tail of variations in lighting and resultant shadows; the color, texture, and reflectivity of the door and handle; the weight of the door; the type of the handle and the specific forces required to operate it; and so on.

        1 reply 1 proslijeđeni tweet 22 korisnika označavaju da im se sviđa
        Prikaži ovu nit
      6. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        There are even slight variations between different “identical” robots that are running the same software. They inevitably have different sensor/camera calibrations and motor characteristics. A true solution to the problem has to be general enough to contend with all of this.

        1 reply 3 proslijeđena tweeta 16 korisnika označava da im se sviđa
        Prikaži ovu nit
      7. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        Our approach was to depart from traditional techniques in robotics that favor modularity (e.g. one subsystem determines the pose of the handle, a completely independent one controls the arm of the robot, and yet another system then operates the gripper, etc).

        1 reply 2 proslijeđena tweeta 8 korisnika označava da im se sviđa
        Prikaži ovu nit
      8. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        Instead, we set out to train a single deep neural network to do everything, end-to-end. The goal was to get the network to go directly from the robot’s raw sensor data all the way to the final output: 7 torques to apply to the robot’s 7 joints for 20 ms.pic.twitter.com/LhzozlVrQG

        1 reply 2 proslijeđena tweeta 28 korisnika označava da im se sviđa
        Prikaži ovu nit
      9. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        After 20 milliseconds, the neural network runs again (this time with slightly different sensor data, for the robot has since moved). It thus produces seven *new* torques. And so, in this manner, the network guides the robot toward opening the door 20 ms at a time.

        1 reply 1 proslijeđeni tweet 11 korisnika označava da im se sviđa
        Prikaži ovu nit
      10. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        To do this, it must internalize a model of everything from the high level look of a door (and its handle) in different conditions, down to the proprioceptive “feeling” of interacting with the handle and imparting on it the right forces so that the door opens.

        1 reply 1 proslijeđeni tweet 13 korisnika označava da im se sviđa
        Prikaži ovu nit
      11. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        And, to make this more fun, we wanted to get the same neural network (with the same weights) to work across different kinds of doors with different handles, and across distinct “identical” robots, each with diverging camera and sensor calibrations.

        1 reply 1 proslijeđeni tweet 12 korisnika označava da im se sviđa
        Prikaži ovu nit
      12. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        This required a lot of data. And, collecting it required running multiple robots simultaneously, each learning from trial and error on a different door with a unique handle.

        1 reply 2 proslijeđena tweeta 8 korisnika označava da im se sviđa
        Prikaži ovu nit
      13. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        We wanted to get the robots to learn from each other’s experience in real time and to use the data produced from each individual robot’s experience to train a single, global neural network that is shared across all robots. We wanted to build a hivemind :)pic.twitter.com/VwYwHZ5bNs

        1 reply 3 proslijeđena tweeta 23 korisnika označavaju da im se sviđa
        Prikaži ovu nit
      14. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        We called this Collective Reinforcement Learning. The proof of concept was a collective effort with Adrian Li, @mkalakrishnan, @YevgenChebotar, and @svlevine. Check out the full paper here! https://arxiv.org/pdf/1610.00673.pdf …pic.twitter.com/8Cpb4bRV8m

        4 proslijeđena tweeta 36 korisnika označava da im se sviđa
        Prikaži ovu nit
      15. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        But, what does all of this have to do with data network effects and product defensibility?

        1 reply 1 proslijeđeni tweet 7 korisnika označava da im se sviđa
        Prikaži ovu nit
      16. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        Well, at the surface it might seem like this one case (of all cases!) should benefit from strong data network effects because each robot that is added to the network learns (immediately and continually) from the experience of every robot that precedes it.

        1 reply 1 proslijeđeni tweet 11 korisnika označava da im se sviđa
        Prikaži ovu nit
      17. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        If you believe, as we did, that each robot would contribute to the collective a differentiated set of experiences, then the marginal utility offered by the network to a new user bringing her robot online should scale superlinearly with the number of robots in the network.

        1 reply 2 proslijeđena tweeta 12 korisnika označava da im se sviđa
        Prikaži ovu nit
      18. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        That right there is a network that gets more valuable for the next guy who joins, the more people are already on it, right? That’s a network effect by definition, no? It’s Metcalfe’s Law! (:

        1 reply 1 proslijeđeni tweet 11 korisnika označava da im se sviđa
        Prikaži ovu nit
      19. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        And indeed, all of that is partly true. The nuance here, as Martin and Peter would observe, is that this effect can actually be better explained by the scale of the collected data than by the fact that the robots are incidentally connected to one another in a network.

        1 reply 3 proslijeđena tweeta 19 korisnika označava da im se sviđa
        Prikaži ovu nit
      20. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        If you could somehow collect the exact same data but without the network, the result would be the same. The network is an implementation detail — it’s just the mechanism by which the data is collected and updates to the hivemind model are distributed.pic.twitter.com/xifx433ZIq

        1 reply 4 proslijeđena tweeta 26 korisnika označava da im se sviđa
        Prikaži ovu nit
      21. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        This is different from the dynamics of, say, a social network where every edge connecting one user to another really does play a functional role in increasing the network’s value. Our robot hivemind benefits far more from data scale effects than it does from data network effects.

        1 reply 2 proslijeđena tweeta 39 korisnika označava da im se sviđa
        Prikaži ovu nit
      22. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        And, as it turns out, scale effects tend to be less powerful (and less defensible) than network effects. Why?

        1 reply 4 proslijeđena tweeta 15 korisnika označava da im se sviđa
        Prikaži ovu nit
      23. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        Well, at least in principle, the utility curve that best models a network effect is a true superlinear curve. Social networks do indeed continue to get more valuable with each additional user that joins. And each new user brings to the network more value than the last.

        1 reply 1 proslijeđeni tweet 7 korisnika označava da im se sviđa
        Prikaži ovu nit
      24. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        This superlinear effect tends to hold indefinitely regardless of how big these social networks get. The same cannot be said for scale effects. The utility curve that best models a scale effect tends to be an S-curve. At some point marginal returns to scale flatten out.pic.twitter.com/6j8Jw2zKZP

        5 replies 6 proslijeđenih tweetova 41 korisnik označava da mu se sviđa
        Prikaži ovu nit
      25. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        This is because scaling any one part of a system tends to cause other parts to have to scale disproportionately. At some point, some physical limit begins to challenge the assumptions that originally held the system together and causes returns to scale to slow down.

        1 reply 2 proslijeđena tweeta 16 korisnika označava da im se sviđa
        Prikaži ovu nit
      26. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        The reality of scaling our hivemind is that, as the network grows, the new data that is contributed by each additional robot becomes ever more redundant with the data that has already been collected by other robots. It thus no longer offers as much useful signal.

        4 proslijeđena tweeta 24 korisnika označavaju da im se sviđa
        Prikaži ovu nit
      27. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        And so, at scale, the amount of additional data that must be collected in order to yield the next useful datapoint begins to increase. This slows down returns to scale and allows trailing competitors to narrow the gap.

        1 reply 2 proslijeđena tweeta 21 korisnik označava da mu se sviđa
        Prikaži ovu nit
      28. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        A related problem that is especially important when dealing with the physical world is that systematic errors (biases) in how data is collected might have no effect at a small scale. But, at larger scale, they often accumulate and begin having a real impact on models’ performance

        1 reply 3 proslijeđena tweeta 21 korisnik označava da mu se sviđa
        Prikaži ovu nit
      29. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        This is compounded by the fact that scaling a robotic fleet tends to coincide with the loosening of quality controls in data collection (or with an increase in infrastructure and QA costs). This creates an additional source of bias (or cost) slow down returns to scale even more.

        1 proslijeđeni tweet 9 korisnika označava da im se sviđa
        Prikaži ovu nit
      30. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        As Martin and Peter point out, there is nothing about data that inherently confers defensibility. Any data advantages must be connected to a more holistic product, technology, or business story.

        1 reply 4 proslijeđena tweeta 31 korisnik označava da mu se sviđa
        Prikaži ovu nit
      31. Ali Yahya‏ @ali01 12. lip 2019.
        • Prijavi Tweet

        Scale effects, on the other hand, can and do confer some defensibility but only during the accelerating portion of the S-curve. In the world of robotics, how significant this effect is comes down to the shape of the data distribution.

        3 proslijeđena tweeta 19 korisnika označava da im se sviđa
        Prikaži ovu nit
      32. Još 8 drugih odgovora

    Čini se da učitavanje traje već neko vrijeme.

    Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

      Sponzorirani tweet

      false

      • © 2020 Twitter
      • O Twitteru
      • Centar za pomoć
      • Uvjeti
      • Pravila o privatnosti
      • Imprint
      • Kolačići
      • Informacije o oglasima