Profile - To The Moon

Yap Wai Hong

在马来西亚赚美金不是梦

Posts · 462Posts · 462
Following · 0Following · 0
Followers · 0Followers · 0

- Yap Wai HongYap Wai Hong
  ·2023-12-29
  2024 Beat the market
  332 Comment
  1
  Share
  Report
- Yap Wai HongYap Wai Hong
  ·2023-03-02
  $SPDR能源指数ETF(XLE)$
  607 Comment
  3
  Share
  Report
- Yap Wai HongYap Wai Hong
  ·2023-01-22
  $SPDR Portfolio S&P 500 Growth ETF(SPYG)$
  519 Comment
  1
  Share
  Report
- Yap Wai HongYap Wai Hong
  ·2022-11-28
  $爱彼迎(ABNB)$
  500 Comment
  1
  Share
  Report
- Yap Wai HongYap Wai Hong
  ·2022-10-11
  $Opendoor Technologies Inc(OPEN)$[得意]
  848 4
  4
  Share
  Report
- Yap Wai HongYap Wai Hong
  ·2022-10-09
  $DraftKings Inc.(DKNG)$
  563 1
  Like
  Share
  Report
- Yap Wai HongYap Wai Hong
  ·2022-10-02
  Palo Alto
  特斯拉AI Day 2022全解读：能走会动的Tesla Bot，DOJO超算明年量产、还有FSD新进展
  几个小时前，特斯拉正式举办了 2022 AI Day，一场全球汽车、人工智能、信息科技行业翘首以待足足 13 个月的发布会。严格意义上 AI Day 不像是「发布会」，而是「交流会」——马斯克本人也在
  特斯拉AI Day 2022全解读：能走会动的Tesla Bot，DOJO超算明年量产、还有FSD新进展
  631 Comment
  Like
  Share
  Report
- Yap Wai HongYap Wai Hong
  ·2022-10-01
  $Sea Ltd(SE)$[得意]
  485 Comment
  Like
  Share
  Report
- Yap Wai HongYap Wai Hong
  ·2022-09-25
  $Opendoor Technologies Inc(OPEN)$[得意]
  614 Comment
  1
  Share
  Report
- Yap Wai HongYap Wai Hong
  ·2022-09-20
  $Kenon Holdings Ltd(KEN)$[得意]
  644 2
  1
  Share
  Report

Most Discussed

Terms and Conditions·Privacy Policy

Company: TTMF Limited. Tech supported by Xiangshang Yixin.

Email：uservice@ttm.financial

{"i18n":{"language":"en_US"},"isCurrentUser":false,"userPageInfo":{"id":"4088247795049920","uuid":"4088247795049920","gmtCreate":1625149386939,"gmtModify":1703292880001,"name":"Yap Wai Hong","pinyin":"yapwaihongyapwaihong","introduction":"","introductionEn":"","signature":"在马来西亚赚美金不是梦","avatar":"https://static.tigerbbs.com/e7710b70b68b33532687ac65fc65f96d","hat":"https://static.tigerbbs.com/4c5ceab71416b120bab4604c0754b52a","hatId":"shopping-5c7909103804749e0b21df6dfddc9bae","hatName":"2024新年头像框","vip":1,"status":2,"fanSize":20,"headSize":135,"tweetSize":462,"questionSize":0,"limitLevel":999,"accountStatus":4,"level":{"id":3,"name":"书生虎","nameTw":"書生虎","represent":"努力向上","factor":"发布10条非转发主帖,其中5条获得他人回复或点赞","iconColor":"3C9E83","bgColor":"A2F1D9"},"themeCounts":0,"badgeCounts":0,"badges":[],"moderator":false,"superModerator":false,"manageSymbols":null,"badgeLevel":null,"boolIsFan":false,"boolIsHead":false,"favoriteSize":1,"symbols":null,"coverImage":null,"realNameVerified":"success","userBadges":[{"badgeId":"44212b71d0be4ec88898348dbe882e03-1","templateUuid":"44212b71d0be4ec88898348dbe882e03","name":"Boss Tiger","description":"The transaction amount of the securities account reaches $100,000","bigImgUrl":"https://static.tigerbbs.com/c8dfc27c1ee0e25db1c93e9d0b641101","smallImgUrl":"https://static.tigerbbs.com/f43908c142f8a33c78f5bdf0e2897488","grayImgUrl":"https://static.tigerbbs.com/82165ff19cb8a786e8919f92acee5213","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.10.19","exceedPercentage":"60.82%","individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1101},{"badgeId":"972123088c9646f7b6091ae0662215be-3","templateUuid":"972123088c9646f7b6091ae0662215be","name":"Legendary Trader","description":"Total number of securities or futures transactions reached 300","bigImgUrl":"https://static.tigerbbs.com/656db16598a0b8f21429e10d6c1cb033","smallImgUrl":"https://static.tigerbbs.com/03f10910d4dd9234f9b5702a3342193a","grayImgUrl":"https://static.tigerbbs.com/0c767e35268feb729d50d3fa9a386c5a","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.04.27","exceedPercentage":"93.38%","individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1100},{"badgeId":"1026c425416b44e0aac28c11a0848493-2","templateUuid":"1026c425416b44e0aac28c11a0848493","name":"Senior Tiger","description":"Join the tiger community for 1000 days","bigImgUrl":"https://static.tigerbbs.com/0063fb68ea29c9ae6858c58630e182d5","smallImgUrl":"https://static.tigerbbs.com/96c699a93be4214d4b49aea6a5a5d1a4","grayImgUrl":"https://static.tigerbbs.com/35b0e542a9ff77046ed69ef602bc105d","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.03.28","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1001},{"badgeId":"cbe1c45c584340f3bd9ae6c77e0e9981-1","templateUuid":"cbe1c45c584340f3bd9ae6c77e0e9981","name":"Academy Experiencer","description":"5 lessons learned","bigImgUrl":"https://community-static.tradeup.com/news/fb5ae275631fb96a92d475cdc85d2302","smallImgUrl":"https://community-static.tradeup.com/news/c2660a1935bd2105e97c9915619936c3","grayImgUrl":null,"redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":1,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.03.24","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":2006},{"badgeId":"3e5a625780d3411098f7e72115f69f0a-1","templateUuid":"3e5a625780d3411098f7e72115f69f0a","name":"Total trades","description":"Cumulative for 100 trades","bigImgUrl":"https://community-static.tradeup.com/news/f6040a65e0c6e125e4dcfe0101de45ee","smallImgUrl":"https://community-static.tradeup.com/news/f803dba33fef1aa87b64be09f4f064ee","grayImgUrl":"https://community-static.tradeup.com/news/5dfa45d4a1e98a2d8c2e1b4253ddd916","redirectLinkEnabled":0,"redirectLink":"https://www.tigerbrokers.com.sg/activity/market/2023/stock-contest?adcode=StockContest2023","hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.03.01","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":5100},{"badgeId":"23f878ed27704312ae41c7f17e73b319-1","templateUuid":"23f878ed27704312ae41c7f17e73b319","name":"Annualised return","description":"Top 50%","bigImgUrl":"https://community-static.tradeup.com/news/ffa0ed2850aef35663a435f4fb24f40c","smallImgUrl":"https://community-static.tradeup.com/news/4f67f095c241bed7b45d41ed5b78dd58","grayImgUrl":"https://community-static.tradeup.com/news/01339d2a5c822a77e0c581ef8f959d81","redirectLinkEnabled":0,"redirectLink":"https://www.tigerbrokers.com.sg/activity/market/2023/stock-contest?adcode=StockContest2023","hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2023.10.21","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":5300},{"badgeId":"f00de38e20574ad1abf55a15966547c8-1","templateUuid":"f00de38e20574ad1abf55a15966547c8","name":"Average winning trade","description":"Top 50%","bigImgUrl":"https://community-static.tradeup.com/news/f53cdecfbff6499d133bad354ae10ff1","smallImgUrl":"https://community-static.tradeup.com/news/2554f84686bbcccc035b93fe29308e43","grayImgUrl":"https://community-static.tradeup.com/news/84c66bc70ac1cf09f4e89b1d4ca9d017","redirectLinkEnabled":0,"redirectLink":"https://www.tigerbrokers.com.sg/activity/market/2023/stock-contest?adcode=StockContest2023","hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2023.09.07","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":5200},{"badgeId":"82ffa69a97264fba9002044f4f5f535b-1","templateUuid":"82ffa69a97264fba9002044f4f5f535b","name":"Tiger Think Tank","description":"Tiger Think Tank","bigImgUrl":"https://community-static.tradeup.com/news/193f8255673f42bdc610097b2074ec3e","smallImgUrl":"https://community-static.tradeup.com/news/06af124f59f026a383d35b268362424f","grayImgUrl":null,"redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2022.09.14","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":4001},{"badgeId":"7a9f168ff73447fe856ed6c938b61789-1","templateUuid":"7a9f168ff73447fe856ed6c938b61789","name":"Knowledgeable Investor","description":"Traded more than 10 stocks","bigImgUrl":"https://static.tigerbbs.com/e74cc24115c4fbae6154ec1b1041bf47","smallImgUrl":"https://static.tigerbbs.com/d48265cbfd97c57f9048db29f22227b0","grayImgUrl":"https://static.tigerbbs.com/76c6d6898b073c77e1c537ebe9ac1c57","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2022.06.07","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1102},{"badgeId":"a83d7582f45846ffbccbce770ce65d84-1","templateUuid":"a83d7582f45846ffbccbce770ce65d84","name":"Real Trader","description":"Completed a transaction","bigImgUrl":"https://static.tigerbbs.com/2e08a1cc2087a1de93402c2c290fa65b","smallImgUrl":"https://static.tigerbbs.com/4504a6397ce1137932d56e5f4ce27166","grayImgUrl":"https://static.tigerbbs.com/4b22c79415b4cd6e3d8ebc4a0fa32604","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2021.12.21","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1100}],"userBadgeCount":10,"currentWearingBadge":{"badgeId":"cbe1c45c584340f3bd9ae6c77e0e9981-1","templateUuid":"cbe1c45c584340f3bd9ae6c77e0e9981","name":"Academy Experiencer","description":"5 lessons learned","bigImgUrl":"https://community-static.tradeup.com/news/fb5ae275631fb96a92d475cdc85d2302","smallImgUrl":"https://community-static.tradeup.com/news/c2660a1935bd2105e97c9915619936c3","grayImgUrl":null,"redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":1,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.03.24","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":2006},"individualDisplayBadges":null,"crmLevel":3,"crmLevelSwitch":1,"location":null,"starInvestorFollowerNum":0,"starInvestorFlag":false,"starInvestorOrderShareNum":0,"subscribeStarInvestorNum":35,"ror":null,"winRationPercentage":null,"showRor":false,"investmentPhilosophy":null,"starInvestorSubscribeFlag":false},"page":1,"watchlist":null,"tweetList":[{"id":257349989253312,"gmtCreate":1703863484275,"gmtModify":1703863488970,"author":{"id":"4088247795049920","authorId":"4088247795049920","name":"Yap Wai Hong","avatar":"https://static.tigerbbs.com/e7710b70b68b33532687ac65fc65f96d","crmLevel":3,"crmLevelSwitch":1,"followedFlag":false,"idStr":"4088247795049920","authorIdStr":"4088247795049920"},"themes":[],"htmlText":"2024&nbsp; Beat the market","listText":"2024&nbsp; Beat the market","text":"2024 Beat the market","images":[{"img":"https://community-static.tradeup.com/news/b08e0c6b40afe10bf57de870168021e1","width":"1080","height":"2112"}],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":1,"commentSize":0,"repostSize":0,"link":"https://ttm.financial/post/257349989253312","isVote":1,"tweetType":1,"viewCount":332,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":1,"langContent":"EN","totalScore":0},{"id":9940359750,"gmtCreate":1677715398371,"gmtModify":1677715401959,"author":{"id":"4088247795049920","authorId":"4088247795049920","name":"Yap Wai Hong","avatar":"https://static.tigerbbs.com/e7710b70b68b33532687ac65fc65f96d","crmLevel":3,"crmLevelSwitch":1,"followedFlag":false,"idStr":"4088247795049920","authorIdStr":"4088247795049920"},"themes":[],"htmlText":"<a href=\"https://ttm.financial/S/XLE\">$SPDR能源指数ETF(XLE)$ &nbsp;</a>","listText":"<a href=\"https://ttm.financial/S/XLE\">$SPDR能源指数ETF(XLE)$ &nbsp;</a>","text":"$SPDR能源指数ETF(XLE)$","images":[{"img":"https://community-static.tradeup.com/news/17c9e2b4ad686df7c0cd16c179ef48b9","width":"1080","height":"2762"}],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":3,"commentSize":0,"repostSize":0,"link":"https://ttm.financial/post/9940359750","isVote":1,"tweetType":1,"viewCount":607,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":1,"langContent":"EN","totalScore":0},{"id":9952921955,"gmtCreate":1674366329177,"gmtModify":1676538938481,"author":{"id":"4088247795049920","authorId":"4088247795049920","name":"Yap Wai Hong","avatar":"https://static.tigerbbs.com/e7710b70b68b33532687ac65fc65f96d","crmLevel":3,"crmLevelSwitch":1,"followedFlag":false,"idStr":"4088247795049920","authorIdStr":"4088247795049920"},"themes":[],"htmlText":"<a href=\"https://ttm.financial/S/SPYG\">$SPDR Portfolio S&amp;P 500 Growth ETF(SPYG)$ &nbsp;</a>","listText":"<a href=\"https://ttm.financial/S/SPYG\">$SPDR Portfolio S&amp;P 500 Growth ETF(SPYG)$ &nbsp;</a>","text":"$SPDR Portfolio S&P 500 Growth ETF(SPYG)$","images":[{"img":"https://community-static.tradeup.com/news/71c24b6d535ead98905d9b523c4ad077","width":"1080","height":"2182"}],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":1,"commentSize":0,"repostSize":0,"link":"https://ttm.financial/post/9952921955","isVote":1,"tweetType":1,"viewCount":519,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":1,"langContent":"EN","totalScore":0},{"id":9966552311,"gmtCreate":1669598047334,"gmtModify":1676538210985,"author":{"id":"4088247795049920","authorId":"4088247795049920","name":"Yap Wai Hong","avatar":"https://static.tigerbbs.com/e7710b70b68b33532687ac65fc65f96d","crmLevel":3,"crmLevelSwitch":1,"followedFlag":false,"idStr":"4088247795049920","authorIdStr":"4088247795049920"},"themes":[],"htmlText":"<a href=\"https://ttm.financial/S/ABNB\">$爱彼迎(ABNB)$ &nbsp;</a>","listText":"<a href=\"https://ttm.financial/S/ABNB\">$爱彼迎(ABNB)$ &nbsp;</a>","text":"$爱彼迎(ABNB)$","images":[{"img":"https://community-static.tradeup.com/news/9ee23300dbe396ca27a745aafa38474a","width":"1080","height":"2780"}],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":1,"commentSize":0,"repostSize":0,"link":"https://ttm.financial/post/9966552311","isVote":1,"tweetType":1,"viewCount":500,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":1,"langContent":"EN","totalScore":0},{"id":9917357020,"gmtCreate":1665446090460,"gmtModify":1676537606358,"author":{"id":"4088247795049920","authorId":"4088247795049920","name":"Yap Wai Hong","avatar":"https://static.tigerbbs.com/e7710b70b68b33532687ac65fc65f96d","crmLevel":3,"crmLevelSwitch":1,"followedFlag":false,"idStr":"4088247795049920","authorIdStr":"4088247795049920"},"themes":[],"htmlText":"<a href=\"https://ttm.financial/S/OPEN\">$Opendoor Technologies Inc(OPEN)$</a>[得意]&nbsp;","listText":"<a href=\"https://ttm.financial/S/OPEN\">$Opendoor Technologies Inc(OPEN)$</a>[得意]&nbsp;","text":"$Opendoor Technologies Inc(OPEN)$[得意]","images":[{"img":"https://community-static.tradeup.com/news/5528e49b238344af874c3006daf1b524","width":"1080","height":"2874"}],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":4,"commentSize":4,"repostSize":0,"link":"https://ttm.financial/post/9917357020","isVote":1,"tweetType":1,"viewCount":848,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":1,"langContent":"EN","totalScore":0},{"id":9914209471,"gmtCreate":1665281408793,"gmtModify":1676537581332,"author":{"id":"4088247795049920","authorId":"4088247795049920","name":"Yap Wai Hong","avatar":"https://static.tigerbbs.com/e7710b70b68b33532687ac65fc65f96d","crmLevel":3,"crmLevelSwitch":1,"followedFlag":false,"idStr":"4088247795049920","authorIdStr":"4088247795049920"},"themes":[],"htmlText":"<a href=\"https://ttm.financial/S/DKNG\">$DraftKings Inc.(DKNG)$</a>","listText":"<a href=\"https://ttm.financial/S/DKNG\">$DraftKings Inc.(DKNG)$</a>","text":"$DraftKings Inc.(DKNG)$","images":[{"img":"https://community-static.tradeup.com/news/7135acb6637fbfd5ee0d856df2b979e4","width":"1080","height":"2865"}],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":1,"repostSize":0,"link":"https://ttm.financial/post/9914209471","isVote":1,"tweetType":1,"viewCount":563,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":1,"langContent":"EN","totalScore":0},{"id":9916783890,"gmtCreate":1664679809010,"gmtModify":1676537493393,"author":{"id":"4088247795049920","authorId":"4088247795049920","name":"Yap Wai Hong","avatar":"https://static.tigerbbs.com/e7710b70b68b33532687ac65fc65f96d","crmLevel":3,"crmLevelSwitch":1,"followedFlag":false,"idStr":"4088247795049920","authorIdStr":"4088247795049920"},"themes":[],"htmlText":"Palo Alto","listText":"Palo Alto","text":"Palo Alto","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"link":"https://ttm.financial/post/9916783890","repostId":"1150155136","repostType":4,"repost":{"id":"1150155136","pubTimestamp":1664616181,"share":"https://ttm.financial/m/news/1150155136?lang=&edition=fundamental","pubTime":"2022-10-01 17:23","market":"us","language":"zh","title":"特斯拉AI Day 2022全解读：能走会动的Tesla Bot，DOJO超算明年量产、还有FSD新进展","url":"https://stock-news.laohu8.com/highlight/detail?id=1150155136","media":"电动星球News","summary":"几个小时前，特斯拉正式举办了 2022 AI Day，一场全球汽车、人工智能、信息科技行业翘首以待足足 13 个月的发布会。严格意义上 AI Day 不像是「发布会」，而是「交流会」——马斯克本人也在","content":"<html><head></head><body><p>几个小时前，<a href=\"https://laohu8.com/S/TSLA\">特斯拉</a>正式举办了 2022 AI Day，一场全球汽车、人工<a href=\"https://laohu8.com/S/5RE.SI\">智能</a>、信息科技行业翘首以待足足 13 个月的发布会。</p><p><img src=\"https://static.tigerbbs.com/0b8803475646fc32fd19f6657d5112fe\" tg-width=\"1080\" tg-height=\"610\" referrerpolicy=\"no-referrer\"/></p><p>严格意义上 AI Day 不像是「发布会」，而是「交流会」——马斯克本人也在推特上说，<b>「此活动旨在招聘 AI 和<a href=\"https://laohu8.com/S/300024\">机器人</a>工程师，因此技术含量很高」</b>——换句话说，这是马斯克的高山流水，为特斯拉的锺子期而开。</p><p>不过这并不妨碍我们以比较轻松的视角，记录下这场科技狂欢。因为特斯拉团队几乎 100% 实现了去年的承诺，在本届 AI Day 上带来了以下技术成果：</p><p><b>不再需要群演的真·Tesla Bot 机器人原型机；</b></p><p><b>不再停留在 PPT 的 DOJO POD 人工智能超级计算机；</b></p><p><b>FSD 技术新进展，等等。</b></p><p>当然，即使我们会尽力写得简单点，今天的文章依然会相对硬核。趁着国庆假期，建议大家可以慢慢看，下面马上开始。</p><p><b>一、Tesla Bot 原型机</b></p><p>Optimus 它来了！</p><p><img src=\"https://static.tigerbbs.com/dab53fdbd457f788ef39a1d9919389ec\" tg-width=\"1080\" tg-height=\"616\" referrerpolicy=\"no-referrer\"/></p><p>13 个月前还需要群演的 Tesla Bot，今天正式以原型机的形式出现——原型意思是<b>它还没穿衣服（外壳）</b>。</p><p><img src=\"https://static.tigerbbs.com/fcaee2a3239269a6af58a05db6ac8417\" tg-width=\"1080\" tg-height=\"561\" referrerpolicy=\"no-referrer\"/></p><p>原型机的样子比 PPT 里面明显更粗放，线束、促动器等零件堆砌略显凌乱。但好消息是，Tesla Bot 原型机已经可以走路、打招呼，双手可以完整举过头顶。</p><p><img src=\"https://static.tigerbbs.com/7a2b705c19e93a543e6cc3a1c7e0c37c\" tg-width=\"1080\" tg-height=\"607\" referrerpolicy=\"no-referrer\"/></p><p>在特斯拉的演示视频里，Optimus 已经可以做一些简单的工作，比如搬运箱子、浇花等等。</p><p><img src=\"https://static.tigerbbs.com/1f19762e88d3920e9a4ada708cb6d1f6\" tg-width=\"1080\" tg-height=\"611\" referrerpolicy=\"no-referrer\"/><img src=\"https://static.tigerbbs.com/bfb1c944c5b881d14db28c8f4604b0c2\" tg-width=\"1080\" tg-height=\"611\" referrerpolicy=\"no-referrer\"/><img src=\"https://static.tigerbbs.com/5b534c9c811249fa97342b12691d049b\" tg-width=\"1080\" tg-height=\"611\" referrerpolicy=\"no-referrer\"/></p><p>但更重要的可能是这个画面：Optimus 眼中的世界，<b>通过纯视觉发现并分析周边的一切，然后识别出自己的任务对象。</b></p><p><img src=\"https://static.tigerbbs.com/af0b4407ea6333fe710b73444f74d19c\" tg-width=\"1080\" tg-height=\"605\" referrerpolicy=\"no-referrer\"/></p><p>事实上 Optimus 不是不能装上外壳，但出于工程原因，带外壳版本截止到发布会当天还不能自如地走路（原因后面再解释），只能简单挥舞一下手臂。</p><p><img src=\"https://static.tigerbbs.com/d3544fb2099584436950ca53ffce6023\" tg-width=\"1080\" tg-height=\"614\" referrerpolicy=\"no-referrer\"/></p><p>装上外壳之后我们发现，更接近量产版的 Optimus，变得更胖了——现在它重 73 公斤，比去年 PPT 版「增重」超过 20%，整个「人」圆了一大圈。</p><p><img src=\"https://static.tigerbbs.com/0c2e0f15bfe1b7939ed6fae7f16a0619\" tg-width=\"1080\" tg-height=\"612\" referrerpolicy=\"no-referrer\"/></p><p>更接近量产，也意味着 Optimus 更高阶的参数也可以公布了：<b>100W 静坐功耗、500W 快步走功耗、超过 200 档的关节自由度，光手部自由度就有 27 档。</b></p><p><img src=\"https://static.tigerbbs.com/039afba7177cebbdc4ed877be352b7cf\" tg-width=\"1080\" tg-height=\"611\" referrerpolicy=\"no-referrer\"/></p><p>另外，<b>Optimus 的大脑由单块 FSD Chip 组成，意味着算力应该是 HW3.0 的一半（72TOPS）；电池则是 52V 电压、2.3kWh 容量、内置电子电气元件的一体单元。</b></p><p>说完数字，是时候聊聊 Optimus 的研发逻辑了。</p><p><b>1. 汽车化</b></p><p>马斯克说过<b>「当你能解决自动驾驶，你就能解决现实世界中的人工智能」</b>。这句话点破了特斯拉研发 Optimus 的方法论：大量借鉴汽车研发经验。</p><p>比如借鉴汽车碰撞模拟软件，为 Optimus 编写「跌倒测试」软件。</p><p><img src=\"https://static.tigerbbs.com/d71e64fac130c88c519e38c10f7bae68\" tg-width=\"1080\" tg-height=\"614\" referrerpolicy=\"no-referrer\"/></p><p>再比如利用汽车大规模零件的生产经验，为 Optimus 挑选尽可能保证成本+效率的原材料。「我们不会用碳纤维、钛合金这样的原材料。因为它们虽然很优秀，但像肩膀这样的易损部位，制造和维修成本都太贵了」。</p><p><img src=\"https://static.tigerbbs.com/4a952bb18e0a7d3cd21f635e72bd6f55\" tg-width=\"1080\" tg-height=\"613\" referrerpolicy=\"no-referrer\"/></p><p>除此以外，制造 Optimus 的中心思想，也基本和智能汽车相当：减少线束长度、计算和电子控制单元中心化，等等。</p><p><b>2. 仿生学</b></p><p>既然是类人机器人 humanoid，设计自然要借鉴人类仿生学。</p><p><img src=\"https://static.tigerbbs.com/2059ba8af7bd4a74564e19e5a7a0c3fd\" tg-width=\"1080\" tg-height=\"594\" referrerpolicy=\"no-referrer\"/></p><p>特斯拉用了几个例子解释 Optimus 的仿生学，首先是膝关节。特斯拉表示 Optimus 的关节希望尽量复刻生物学上的「非线性」逻辑，也就是贴合膝关节直立到完全弯曲时的受力曲线。</p><p><img src=\"https://static.tigerbbs.com/093cc607f4e1764484894b9e63cfba7c\" tg-width=\"1080\" tg-height=\"609\" referrerpolicy=\"no-referrer\"/></p><p>为此，Optimus 的膝关节使用了类似于平面四杆机构的设计，最终发力效果会更接近人类。</p><p><img src=\"https://static.tigerbbs.com/4600fe5bc521375fed886e766ecd1fdb\" tg-width=\"1080\" tg-height=\"613\" referrerpolicy=\"no-referrer\"/></p><p>紧接着，我们创造人类文明的双手，才是 Optimus 类人之路更大的 boss。</p><p><img src=\"https://static.tigerbbs.com/6c75b0de1bfce1e9340c6f66fb8391c9\" tg-width=\"1080\" tg-height=\"611\" referrerpolicy=\"no-referrer\"/></p><p>Optimus 光手掌区域就用了 6 个促动器，具有 11 档的自由度。拥有自适应的抓握角度、20 磅（9 公斤）负荷、工具使用能力、小物件精准抓握能力等等。</p><p><img src=\"https://static.tigerbbs.com/9165625d2ab1d7f020961db4ebe149c7\" tg-width=\"1080\" tg-height=\"615\" referrerpolicy=\"no-referrer\"/></p><p>此外，Optimus 的手掌用的是「non-backdrivable」无法反向驱动的指尖促动器。学术界的看法是，这样的促动器可以提升在「开放环境」下的性能。</p><p>最后是让 Optimus 学着像人类一样走路——这里用到的仿生学设计叫做「运动重心控制」。</p><p><img src=\"https://static.tigerbbs.com/5f52fd071ef0d384851c030a69d683a4\" tg-width=\"1080\" tg-height=\"614\" referrerpolicy=\"no-referrer\"/></p><p>为什么有外壳的  Optimus 还不会走？其中一个原因就是重量变了，运动重心控制算法需要重新调试。</p><p><img src=\"https://static.tigerbbs.com/d7b2784fc70757f11b9e6cc0a0c727c6\" tg-width=\"1080\" tg-height=\"614\" referrerpolicy=\"no-referrer\"/></p><p>事实上，Optimus 不仅要做到会走路，还要做到别摔倒。所以它不仅需要控制走路的重心，还要稳住受到外力（比如推搡）时的随机动态重心。</p><p><img src=\"https://static.tigerbbs.com/d525cbf7fdbcc36b915b30db5eff613d\" tg-width=\"1080\" tg-height=\"611\" referrerpolicy=\"no-referrer\"/></p><p><b>训练 FSD 用到的神经网络和在线仿真模拟</b>，这次在 Optimus 身上大显身手。<b>路径规划、视觉融合、视觉导航</b>等等熟悉的名词都被「灌输」到 Optimus 脑子里。</p><p><img src=\"https://static.tigerbbs.com/9a31fa75cdf8e5c9d09662e0d21b8c51\" tg-width=\"1080\" tg-height=\"611\" referrerpolicy=\"no-referrer\"/></p><p>这样的努力下，Optimus 今年 4 月迈出了它的第一步；7 月份解锁了骨盆活动；8 月走路时可以摆手臂了——发布会前几周，实现了脚趾离地的类人行走动作。</p><p><img src=\"https://static.tigerbbs.com/e7d02ca8c2cd350fe355c188bfb968ec\" tg-width=\"1080\" tg-height=\"614\" referrerpolicy=\"no-referrer\"/></p><p><b>3. 「肌肉」</b></p><p>我们通过结缔组织包裹着的肌肉完成运动，机器人的「肌肉」则叫做促动器 actuator。</p><p><img src=\"https://static.tigerbbs.com/de747111c66e3c5fe67ab63226333b02\" tg-width=\"1080\" tg-height=\"616\" referrerpolicy=\"no-referrer\"/></p><p>如上图所示，橙色部分均为 Optimus 的促动器，这些促动器也都是特斯拉完全自研的。</p><p><img src=\"https://static.tigerbbs.com/d5da9845173ab7eddf1c4f638e646c1f\" tg-width=\"1080\" tg-height=\"612\" referrerpolicy=\"no-referrer\"/></p><p>特斯拉为  Optimus 从力度大小的角度，设计了 6 种各自独特的促动器——这其实是很小的数字，<b>业界平均是 20-30，甚至 50 种，目的是覆盖尽可能多的人类活动细节。</b></p><p>为什么特斯拉的促动器种类这么少？原因还是 FSD 体系。</p><p>特斯拉举了 28 种人类常见活动，比如抬举手臂、弯曲右膝等。通过分析这些活动反馈的云数据，<b>找出各类运动的相对共同点，然后就可以尽量减少专门设计促动器的种类。</b></p><p><img src=\"https://static.tigerbbs.com/5e1a6c9753e3c8ab582c060fb1f3be97\" tg-width=\"1080\" tg-height=\"612\" referrerpolicy=\"no-referrer\"/></p><p>虽然只是轻描淡写的一张 PPT，但我认为促动器从 50 种减少到 6 种，意义实际上远大于借鉴特斯拉电机经验的促动器本体——<b>因为它代表着数据为王的新工业时代。</b></p><p>不过促动器种类大幅度减少，也意味着 Optimus 前期的实际效果可能会没有那么「类人」，当然还是得等最终交付了。</p><p>最后来说一个数字：<b>2 万美元</b>（约 14 万元）。</p><p>这笔钱买不到半台 Model 3，但却是马斯克口中 Optimus 的目标售价。<b>「它会彻底改变人类社会的效率，就像无人交通可以彻底改变运输效率」</b>。</p><p><b>二、DOJO 的终极形态？</b></p><p>本来发布会的第二部分是 FSD，但那部分过于硬核，我决定先让大家看点激动人心的数字。</p><p>去年 DOJO 惊艳全世界，但遗憾的是有太多细节未公布。<b>D1 芯片是怎么组成 EXA POD 超算系统的？理论性能爆炸，能代表实际应用吗？</b></p><p>这部分，特斯拉举了大量的数据，证明自己已经是计算领域的新巨头。</p><p><img src=\"https://static.tigerbbs.com/46049affe465303b965a9eae221fe1e5\" tg-width=\"1080\" tg-height=\"610\" referrerpolicy=\"no-referrer\"/></p><p><b>首先是散热。</b></p><p>先别发问号，超算平台的散热，一直是衡量超算制造者系统工程能力的重要维度。比如<a href=\"https://laohu8.com/S/GOOG\">谷歌</a>、华为、<a href=\"https://laohu8.com/S/NVDA\">英伟达</a>在公布自家方案的时候，都会花大篇幅讲散热。</p><p>DOJO POD 的散热可以用两个词概括：高集成度、高自研率。</p><p><img src=\"https://static.tigerbbs.com/bc01b3d2eb00c916972bb0f463901f65\" tg-width=\"1080\" tg-height=\"613\" referrerpolicy=\"no-referrer\"/></p><p>特斯拉在 DOJO POD 上使用了全自研的 VRM（电压调节模组），单个 VRM 模组可以在不足 25 美分硬币面积的电路上，提供超过 1000A 的电流。</p><p>高集成度带来的问题，是热膨胀系数 CTE。DOJO 堪称极限的体积集成率和发热，意味着 CTE 稍微失控，都会对系统结构造成巨大破坏（也就是会撑爆）。</p><p><img src=\"https://static.tigerbbs.com/1b89e0309fc92979fcc0544e43395e78\" tg-width=\"1080\" tg-height=\"612\" referrerpolicy=\"no-referrer\"/></p><p>为此，这套自研 VRM 在过去两年内迭代了 14 个版本，最终才完全符合特斯拉对 CTE 指标的要求。</p><p>目前 DOJO POD 已经进入负载测试阶段——单机柜 2.2MW 的负载，相当于 6 台 Model Y 双电机全力输出。</p><p><img src=\"https://static.tigerbbs.com/a690ca8342c81d44c53f98da2a5216e5\" tg-width=\"1080\" tg-height=\"618\" referrerpolicy=\"no-referrer\"/></p><p><b>解决了散热，才有资格说集成度。</b></p><p>一个 DOJO POD 机柜由两层计算托盘和存储系统组成。每一层托盘都有 6 个 D1 Tile 计算「瓦片」——两层 12 片 组成的一个机柜，就可以提供 108PFLOPS 算力的深度学习性能。</p><p><img src=\"https://static.tigerbbs.com/fefe4441c10c18928ac171726b35c75c\" tg-width=\"1080\" tg-height=\"614\" referrerpolicy=\"no-referrer\"/></p><p>对了，DOJO POD 的供电模组也是 52V 电压的，Optimus 母亲实锤了。</p><p>每层托盘都连接着超高速存储系统：640GB 运行内存可以提供超过 18TB 每秒的运算带宽，另外还有超过 1TB 每秒的网络交换。</p><p><img src=\"https://static.tigerbbs.com/4c8182aab7ef49bf1807ba3faf6decdd\" tg-width=\"1080\" tg-height=\"612\" referrerpolicy=\"no-referrer\"/></p><p>为了适配训练软件以及运营/维护，每个托盘还配备了专属的管理计算中心。</p><p><img src=\"https://static.tigerbbs.com/fb8ec92bdd51cc4895f9ad44411a3c44\" tg-width=\"1080\" tg-height=\"615\" referrerpolicy=\"no-referrer\"/></p><p>最终，可以提供<b>1.1E 算力、13TB 运存、1.3TB 缓存</b>的 EXA POD，将于 2023 年 Q1，正式量产——<b>这也是今天发布会唯一一个有确定日期的特斯拉产品。</b></p><p><img src=\"https://static.tigerbbs.com/b3130399d3b4808b0e5083b98e1877eb\" tg-width=\"1080\" tg-height=\"609\" referrerpolicy=\"no-referrer\"/></p><p>意大利炮有了，能不能轰下县城？</p><p><img src=\"https://static.tigerbbs.com/cb806e7598b3d28981907243c404eae4\" tg-width=\"1080\" tg-height=\"613\" referrerpolicy=\"no-referrer\"/></p><p>特斯拉表示，配合专属的编译器，DOJO 的训练延迟，最低可以做到同等规模 GPU 的<b>1/50！</b></p><p>最终，特斯拉的目标是到 2023 年 Q1 量产时，<b>DOJO 可以实现相比英伟达 A100，最高 4.4 倍的单芯片训练速度</b>——甚至能耗和成本都更低。</p><p><img src=\"https://static.tigerbbs.com/14920c55136e9bd5b7291378acfecdf0\" tg-width=\"1080\" tg-height=\"614\" referrerpolicy=\"no-referrer\"/></p><p><b>三、FSD 的新进化</b></p><p>文章来到这里，大家的手指应该已经划了很多次屏幕。这也说明，看到这里依然兴致勃勃的你，一定是特斯拉老粉——那就聊点更「无聊」、更硬核的吧。</p><p><img src=\"https://static.tigerbbs.com/df2d0dfc66bf889e9078c98744c779da\" tg-width=\"1080\" tg-height=\"612\" referrerpolicy=\"no-referrer\" width=\"100%\" height=\"auto\"/></p><p>篇幅有限，本届 AI Day 关于 FSD 的进展，我们只聊三个点：<b>Occupancy Network、Training Optimization、Lanes</b>。</p><p><b>1. Occupancy Network</b></p><p>先聊一个概念：矢量图。做设计的朋友一定很熟悉，这是一种精度（分辨率）可以做到无限，但占用存储空间很小的数字绘图。</p><p>Occupancy Network，就是将 3D 向量数据绘制成矢量图的、 2019 年开始兴起的一种三维重建表达方法。</p><p>有意思的是，特斯拉用了最 Occupancy Network 的方式，表达他们对 Occupancy Network 的应用：网格（方块）化的 3D 模拟。</p><p>其实 FSD 眼中的世界并不是这样 Minecraft 化的，但 Occupancy Network 的本质特征，就是用「决策边界」描绘「物体边缘」。</p><p><img src=\"https://static.tigerbbs.com/933a6dff415d2a20aca4fa2aa387f44c\" tg-width=\"1080\" tg-height=\"612\" referrerpolicy=\"no-referrer\"/></p><p>尽管 Occupancy Network 效率很高，但实际训练规模依然足够可观。目前特斯拉公布的数据是超过<b>14.4 亿帧</b>视频数据，需要超过<b>10 万个 GPU 训练小时</b>，实际视频缓存超过<b>30PB</b>——而且全程 90℃ 满负载。</p><p><img src=\"https://static.tigerbbs.com/8c5c41679c5fd2e95dc11d2a51de11ea\" tg-width=\"1080\" tg-height=\"616\" referrerpolicy=\"no-referrer\"/></p><p><b>二、因此，Training Optimization 训练优化尤为重要。</b></p><p>去年 Andrej 公布了特斯拉的千人 in-house 标注团队，今年特斯拉的重点，则在于优化自动标注流程。</p><p><img src=\"https://static.tigerbbs.com/2e04573dd50a78f4179b4a2ad9853e39\" tg-width=\"1080\" tg-height=\"615\" referrerpolicy=\"no-referrer\"/></p><p>大概总结一下就是，优化过后，训练时视频帧选取会更智能，同时大幅度减少选取的视频帧数量——<b>可以提高 30% 的训练速度</b>。</p><p><img src=\"https://static.tigerbbs.com/8ae0aee2c7838d44aedec2ba8e7b4d6d\" tg-width=\"1080\" tg-height=\"614\" referrerpolicy=\"no-referrer\"/></p><p>另外视频模型训练时 smol 异步库文件体积可以缩小 11%，所需的读取次数足足缩小到 1/4...<b>最终这套优化流程让特斯拉的 Occupancy Network 训练效率提升了 2.3 倍。</b></p><p><b>3. 最后聊聊车道线 Lanes。</b></p><p>从 FSD Beta 10.12 开始，几乎每一版更新，车道线和无保护左转，都是更新日志的第一条。</p><p><img src=\"https://static.tigerbbs.com/a0495d4d6b37869a27b1abb40423df11\" tg-width=\"1080\" tg-height=\"613\" referrerpolicy=\"no-referrer\"/></p><p>为了更准确高效应对车道线，特斯拉这次「编」了一套「属于车道的语言」。其中包括车道级别的地理几何学和拓扑几何学、车道导航、公交车道计算、多乘员车辆车道计算等等。</p><p><img src=\"https://static.tigerbbs.com/a5e8f91a38610e5c5e23fe68081b8660\" tg-width=\"1080\" tg-height=\"615\" referrerpolicy=\"no-referrer\"/></p><p><b>最终这套「车道的语言」，可以在小于 10 毫秒的延迟内，思考超过 7500 万个可能影响车辆决策的因素——而且 FSD 硬件「学会」这套语言的代价（功耗），还不足 8W。</b></p><p><img src=\"https://static.tigerbbs.com/a5e8aa95ee1c66859b5906c6ca135d4e\" tg-width=\"1080\" tg-height=\"613\" referrerpolicy=\"no-referrer\"/></p><p><b>四、四十年后，开始圆梦？</b></p><p>写到这里，我真的很头疼。</p><p>一方面是我们大部分人，都不是这届 AI Day 的对象——马斯克眼里只有招聘。另一方面，是现在一家汽车公司的发布会，对知识面要求实在太高了。</p><p><img src=\"https://static.tigerbbs.com/939520f801acf9535d7055a615bf4a47\" tg-width=\"1080\" tg-height=\"576\" referrerpolicy=\"no-referrer\"/></p><p>还是说回马斯克吧，40 年前的他，还是个每天会看 10 个小时科幻小说的小孩子，沉醉于《银河系漫游指南》、《基地》、《严厉的月亮》等等。</p><p>但正是这些科幻小说，培养了马斯克冰冷却又宏大的事业观。他会跟你说人类社会生产力的效率可以扩大到无限，他会跟你说人口是维系文明的最重要因素。</p><p>所以，当我们把 52 岁的马斯克和 12 岁的马斯克放在一起，你会发现他俩依然在本质上是同一个人。</p><p>也正因如此，你看到他如今几乎涉猎了科幻小说所有最热门题材的商业帝国，才会觉得「哦，那很正常」。</p><p>希望明年我们能看到更接近现实的马斯克童梦吧。</p></body></html>","source":"lsy1574414115752","collect":0,"html":"<!DOCTYPE html>\n<html>\n<head>\n<meta http-equiv=\"Content-Type\" content=\"text/html; charset=utf-8\" />\n<meta name=\"viewport\" content=\"width=device-width,initial-scale=1.0,minimum-scale=1.0,maximum-scale=1.0,user-scalable=no\"/>\n<meta name=\"format-detection\" content=\"telephone=no,email=no,address=no\" />\n<title>特斯拉AI Day 2022全解读：能走会动的Tesla Bot，DOJO超算明年量产、还有FSD新进展</title>\n<style type=\"text/css\">\na,abbr,acronym,address,applet,article,aside,audio,b,big,blockquote,body,canvas,caption,center,cite,code,dd,del,details,dfn,div,dl,dt,\nem,embed,fieldset,figcaption,figure,footer,form,h1,h2,h3,h4,h5,h6,header,hgroup,html,i,iframe,img,ins,kbd,label,legend,li,mark,menu,nav,\nobject,ol,output,p,pre,q,ruby,s,samp,section,small,span,strike,strong,sub,summary,sup,table,tbody,td,tfoot,th,thead,time,tr,tt,u,ul,var,video{ font:inherit;margin:0;padding:0;vertical-align:baseline;border:0 }\nbody{ font-size:16px; line-height:1.5; color:#999; background:transparent; }\n.wrapper{ overflow:hidden;word-break:break-all;padding:10px; }\nh1,h2{ font-weight:normal; line-height:1.35; margin-bottom:.6em; }\nh3,h4,h5,h6{ line-height:1.35; margin-bottom:1em; }\nh1{ font-size:24px; }\nh2{ font-size:20px; }\nh3{ font-size:18px; }\nh4{ font-size:16px; }\nh5{ font-size:14px; }\nh6{ font-size:12px; }\np,ul,ol,blockquote,dl,table{ margin:1.2em 0; }\nul,ol{ margin-left:2em; }\nul{ list-style:disc; }\nol{ list-style:decimal; }\nli,li p{ margin:10px 0;}\nimg{ max-width:100%;display:block;margin:0 auto 1em; }\nblockquote{ color:#B5B2B1; border-left:3px solid #aaa; padding:1em; }\nstrong,b{font-weight:bold;}\nem,i{font-style:italic;}\ntable{ width:100%;border-collapse:collapse;border-spacing:1px;margin:1em 0;font-size:.9em; }\nth,td{ padding:5px;text-align:left;border:1px solid #aaa; }\nth{ font-weight:bold;background:#5d5d5d; }\n.symbol-link{font-weight:bold;}\n/* header{ border-bottom:1px solid #494756; } */\n.title{ margin:0 0 8px;line-height:1.3;color:#ddd; }\n.meta {color:#5e5c6d;font-size:13px;margin:0 0 .5em; }\na{text-decoration:none; color:#2a4b87;}\n.meta .head { display: inline-block; overflow: hidden}\n.head .h-thumb { width: 30px; height: 30px; margin: 0; padding: 0; border-radius: 50%; float: left;}\n.head .h-content { margin: 0; padding: 0 0 0 9px; float: left;}\n.head .h-name {font-size: 13px; color: #eee; margin: 0;}\n.head .h-time {font-size: 11px; color: #7E829C; margin: 0;line-height: 11px;}\n.small {font-size: 12.5px; display: inline-block; transform: scale(0.9); -webkit-transform: scale(0.9); transform-origin: left; -webkit-transform-origin: left;}\n.smaller {font-size: 12.5px; display: inline-block; transform: scale(0.8); -webkit-transform: scale(0.8); transform-origin: left; -webkit-transform-origin: left;}\n.bt-text {font-size: 12px;margin: 1.5em 0 0 0}\n.bt-text p {margin: 0}\n</style>\n</head>\n<body>\n<div class=\"wrapper\">\n<header>\n<h2 class=\"title\">\n特斯拉AI Day 2022全解读：能走会动的Tesla Bot，DOJO超算明年量产、还有FSD新进展\n</h2>\n\n<h4 class=\"meta\">\n\n\n2022-10-01 17:23 北京时间&nbsp;&nbsp;&nbsp;<a href=https://mp.weixin.qq.com/s/HRr8_Og1Y2oBBmouv0E1BA><strong>电动星球News</strong></a>\n\n\n</h4>\n\n</header>\n<article>\n<div>\n<p>几个小时前，特斯拉正式举办了 2022 AI Day，一场全球汽车、人工智能、信息科技行业翘首以待足足 13 个月的发布会。严格意义上 AI Day 不像是「发布会」，而是「交流会」——马斯克本人也在推特上说，「此活动旨在招聘 AI 和机器人工程师，因此技术含量很高」——换句话说，这是马斯克的高山流水，为特斯拉的锺子期而开。不过这并不妨碍我们以比较轻松的视角，记录下这场科技狂欢。因为特斯拉团队几乎...</p>\n\n<a href=\"https://mp.weixin.qq.com/s/HRr8_Og1Y2oBBmouv0E1BA\">Web Link</a>\n\n</div>\n\n\n</article>\n</div>\n</body>\n</html>\n","type":0,"thumbnail":"https://static.tigerbbs.com/0b8803475646fc32fd19f6657d5112fe","relate_stocks":{"BK4543":"AI","BK4534":"瑞士信贷持仓","BK4548":"巴美列捷福持仓","BK4527":"明星科技股","BK4581":"高盛持仓","BK4551":"寇图资本持仓","BK4550":"红杉资本持仓","BK4533":"AQR资本管理(全球第二大对冲基金)","BK4528":"SaaS概念","BK4555":"新能源车","BK4023":"应用软件","BK4511":"特斯拉概念","BK4099":"汽车制造商","BK4574":"无人驾驶","FSD":"First Trust High Income Long/Sho","TSLA":"特斯拉","AI":"C3.ai, Inc."},"source_url":"https://mp.weixin.qq.com/s/HRr8_Og1Y2oBBmouv0E1BA","is_english":false,"share_image_url":"https://static.laohu8.com/e9f99090a1c2ed51c021029395664489","article_id":"1150155136","content_text":"几个小时前，特斯拉正式举办了 2022 AI Day，一场全球汽车、人工智能、信息科技行业翘首以待足足 13 个月的发布会。严格意义上 AI Day 不像是「发布会」，而是「交流会」——马斯克本人也在推特上说，「此活动旨在招聘 AI 和机器人工程师，因此技术含量很高」——换句话说，这是马斯克的高山流水，为特斯拉的锺子期而开。不过这并不妨碍我们以比较轻松的视角，记录下这场科技狂欢。因为特斯拉团队几乎 100% 实现了去年的承诺，在本届 AI Day 上带来了以下技术成果：不再需要群演的真·Tesla Bot 机器人原型机；不再停留在 PPT 的 DOJO POD 人工智能超级计算机；FSD 技术新进展，等等。当然，即使我们会尽力写得简单点，今天的文章依然会相对硬核。趁着国庆假期，建议大家可以慢慢看，下面马上开始。一、Tesla Bot 原型机Optimus 它来了！13 个月前还需要群演的 Tesla Bot，今天正式以原型机的形式出现——原型意思是它还没穿衣服（外壳）。原型机的样子比 PPT 里面明显更粗放，线束、促动器等零件堆砌略显凌乱。但好消息是，Tesla Bot 原型机已经可以走路、打招呼，双手可以完整举过头顶。在特斯拉的演示视频里，Optimus 已经可以做一些简单的工作，比如搬运箱子、浇花等等。但更重要的可能是这个画面：Optimus 眼中的世界，通过纯视觉发现并分析周边的一切，然后识别出自己的任务对象。事实上 Optimus 不是不能装上外壳，但出于工程原因，带外壳版本截止到发布会当天还不能自如地走路（原因后面再解释），只能简单挥舞一下手臂。装上外壳之后我们发现，更接近量产版的 Optimus，变得更胖了——现在它重 73 公斤，比去年 PPT 版「增重」超过 20%，整个「人」圆了一大圈。更接近量产，也意味着 Optimus 更高阶的参数也可以公布了：100W 静坐功耗、500W 快步走功耗、超过 200 档的关节自由度，光手部自由度就有 27 档。另外，Optimus 的大脑由单块 FSD Chip 组成，意味着算力应该是 HW3.0 的一半（72TOPS）；电池则是 52V 电压、2.3kWh 容量、内置电子电气元件的一体单元。说完数字，是时候聊聊 Optimus 的研发逻辑了。1. 汽车化马斯克说过「当你能解决自动驾驶，你就能解决现实世界中的人工智能」。这句话点破了特斯拉研发 Optimus 的方法论：大量借鉴汽车研发经验。比如借鉴汽车碰撞模拟软件，为 Optimus 编写「跌倒测试」软件。再比如利用汽车大规模零件的生产经验，为 Optimus 挑选尽可能保证成本+效率的原材料。「我们不会用碳纤维、钛合金这样的原材料。因为它们虽然很优秀，但像肩膀这样的易损部位，制造和维修成本都太贵了」。除此以外，制造 Optimus 的中心思想，也基本和智能汽车相当：减少线束长度、计算和电子控制单元中心化，等等。2. 仿生学既然是类人机器人 humanoid，设计自然要借鉴人类仿生学。特斯拉用了几个例子解释 Optimus 的仿生学，首先是膝关节。特斯拉表示 Optimus 的关节希望尽量复刻生物学上的「非线性」逻辑，也就是贴合膝关节直立到完全弯曲时的受力曲线。为此，Optimus 的膝关节使用了类似于平面四杆机构的设计，最终发力效果会更接近人类。紧接着，我们创造人类文明的双手，才是 Optimus 类人之路更大的 boss。Optimus 光手掌区域就用了 6 个促动器，具有 11 档的自由度。拥有自适应的抓握角度、20 磅（9 公斤）负荷、工具使用能力、小物件精准抓握能力等等。此外，Optimus 的手掌用的是「non-backdrivable」无法反向驱动的指尖促动器。学术界的看法是，这样的促动器可以提升在「开放环境」下的性能。最后是让 Optimus 学着像人类一样走路——这里用到的仿生学设计叫做「运动重心控制」。为什么有外壳的  Optimus 还不会走？其中一个原因就是重量变了，运动重心控制算法需要重新调试。事实上，Optimus 不仅要做到会走路，还要做到别摔倒。所以它不仅需要控制走路的重心，还要稳住受到外力（比如推搡）时的随机动态重心。训练 FSD 用到的神经网络和在线仿真模拟，这次在 Optimus 身上大显身手。路径规划、视觉融合、视觉导航等等熟悉的名词都被「灌输」到 Optimus 脑子里。这样的努力下，Optimus 今年 4 月迈出了它的第一步；7 月份解锁了骨盆活动；8 月走路时可以摆手臂了——发布会前几周，实现了脚趾离地的类人行走动作。3. 「肌肉」我们通过结缔组织包裹着的肌肉完成运动，机器人的「肌肉」则叫做促动器 actuator。如上图所示，橙色部分均为 Optimus 的促动器，这些促动器也都是特斯拉完全自研的。特斯拉为  Optimus 从力度大小的角度，设计了 6 种各自独特的促动器——这其实是很小的数字，业界平均是 20-30，甚至 50 种，目的是覆盖尽可能多的人类活动细节。为什么特斯拉的促动器种类这么少？原因还是 FSD 体系。特斯拉举了 28 种人类常见活动，比如抬举手臂、弯曲右膝等。通过分析这些活动反馈的云数据，找出各类运动的相对共同点，然后就可以尽量减少专门设计促动器的种类。虽然只是轻描淡写的一张 PPT，但我认为促动器从 50 种减少到 6 种，意义实际上远大于借鉴特斯拉电机经验的促动器本体——因为它代表着数据为王的新工业时代。不过促动器种类大幅度减少，也意味着 Optimus 前期的实际效果可能会没有那么「类人」，当然还是得等最终交付了。最后来说一个数字：2 万美元（约 14 万元）。这笔钱买不到半台 Model 3，但却是马斯克口中 Optimus 的目标售价。「它会彻底改变人类社会的效率，就像无人交通可以彻底改变运输效率」。二、DOJO 的终极形态？本来发布会的第二部分是 FSD，但那部分过于硬核，我决定先让大家看点激动人心的数字。去年 DOJO 惊艳全世界，但遗憾的是有太多细节未公布。D1 芯片是怎么组成 EXA POD 超算系统的？理论性能爆炸，能代表实际应用吗？这部分，特斯拉举了大量的数据，证明自己已经是计算领域的新巨头。首先是散热。先别发问号，超算平台的散热，一直是衡量超算制造者系统工程能力的重要维度。比如谷歌、华为、英伟达在公布自家方案的时候，都会花大篇幅讲散热。DOJO POD 的散热可以用两个词概括：高集成度、高自研率。特斯拉在 DOJO POD 上使用了全自研的 VRM（电压调节模组），单个 VRM 模组可以在不足 25 美分硬币面积的电路上，提供超过 1000A 的电流。高集成度带来的问题，是热膨胀系数 CTE。DOJO 堪称极限的体积集成率和发热，意味着 CTE 稍微失控，都会对系统结构造成巨大破坏（也就是会撑爆）。为此，这套自研 VRM 在过去两年内迭代了 14 个版本，最终才完全符合特斯拉对 CTE 指标的要求。目前 DOJO POD 已经进入负载测试阶段——单机柜 2.2MW 的负载，相当于 6 台 Model Y 双电机全力输出。解决了散热，才有资格说集成度。一个 DOJO POD 机柜由两层计算托盘和存储系统组成。每一层托盘都有 6 个 D1 Tile 计算「瓦片」——两层 12 片 组成的一个机柜，就可以提供 108PFLOPS 算力的深度学习性能。对了，DOJO POD 的供电模组也是 52V 电压的，Optimus 母亲实锤了。每层托盘都连接着超高速存储系统：640GB 运行内存可以提供超过 18TB 每秒的运算带宽，另外还有超过 1TB 每秒的网络交换。为了适配训练软件以及运营/维护，每个托盘还配备了专属的管理计算中心。最终，可以提供1.1E 算力、13TB 运存、1.3TB 缓存的 EXA POD，将于 2023 年 Q1，正式量产——这也是今天发布会唯一一个有确定日期的特斯拉产品。意大利炮有了，能不能轰下县城？特斯拉表示，配合专属的编译器，DOJO 的训练延迟，最低可以做到同等规模 GPU 的1/50！最终，特斯拉的目标是到 2023 年 Q1 量产时，DOJO 可以实现相比英伟达 A100，最高 4.4 倍的单芯片训练速度——甚至能耗和成本都更低。三、FSD 的新进化文章来到这里，大家的手指应该已经划了很多次屏幕。这也说明，看到这里依然兴致勃勃的你，一定是特斯拉老粉——那就聊点更「无聊」、更硬核的吧。篇幅有限，本届 AI Day 关于 FSD 的进展，我们只聊三个点：Occupancy Network、Training Optimization、Lanes。1. Occupancy Network先聊一个概念：矢量图。做设计的朋友一定很熟悉，这是一种精度（分辨率）可以做到无限，但占用存储空间很小的数字绘图。Occupancy Network，就是将 3D 向量数据绘制成矢量图的、 2019 年开始兴起的一种三维重建表达方法。有意思的是，特斯拉用了最 Occupancy Network 的方式，表达他们对 Occupancy Network 的应用：网格（方块）化的 3D 模拟。其实 FSD 眼中的世界并不是这样 Minecraft 化的，但 Occupancy Network 的本质特征，就是用「决策边界」描绘「物体边缘」。尽管 Occupancy Network 效率很高，但实际训练规模依然足够可观。目前特斯拉公布的数据是超过14.4 亿帧视频数据，需要超过10 万个 GPU 训练小时，实际视频缓存超过30PB——而且全程 90℃ 满负载。二、因此，Training Optimization 训练优化尤为重要。去年 Andrej 公布了特斯拉的千人 in-house 标注团队，今年特斯拉的重点，则在于优化自动标注流程。大概总结一下就是，优化过后，训练时视频帧选取会更智能，同时大幅度减少选取的视频帧数量——可以提高 30% 的训练速度。另外视频模型训练时 smol 异步库文件体积可以缩小 11%，所需的读取次数足足缩小到 1/4...最终这套优化流程让特斯拉的 Occupancy Network 训练效率提升了 2.3 倍。3. 最后聊聊车道线 Lanes。从 FSD Beta 10.12 开始，几乎每一版更新，车道线和无保护左转，都是更新日志的第一条。为了更准确高效应对车道线，特斯拉这次「编」了一套「属于车道的语言」。其中包括车道级别的地理几何学和拓扑几何学、车道导航、公交车道计算、多乘员车辆车道计算等等。最终这套「车道的语言」，可以在小于 10 毫秒的延迟内，思考超过 7500 万个可能影响车辆决策的因素——而且 FSD 硬件「学会」这套语言的代价（功耗），还不足 8W。四、四十年后，开始圆梦？写到这里，我真的很头疼。一方面是我们大部分人，都不是这届 AI Day 的对象——马斯克眼里只有招聘。另一方面，是现在一家汽车公司的发布会，对知识面要求实在太高了。还是说回马斯克吧，40 年前的他，还是个每天会看 10 个小时科幻小说的小孩子，沉醉于《银河系漫游指南》、《基地》、《严厉的月亮》等等。但正是这些科幻小说，培养了马斯克冰冷却又宏大的事业观。他会跟你说人类社会生产力的效率可以扩大到无限，他会跟你说人口是维系文明的最重要因素。所以，当我们把 52 岁的马斯克和 12 岁的马斯克放在一起，你会发现他俩依然在本质上是同一个人。也正因如此，你看到他如今几乎涉猎了科幻小说所有最热门题材的商业帝国，才会觉得「哦，那很正常」。希望明年我们能看到更接近现实的马斯克童梦吧。","news_type":1},"isVote":1,"tweetType":1,"viewCount":631,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"EN","totalScore":0},{"id":9916875410,"gmtCreate":1664581387230,"gmtModify":1676537478891,"author":{"id":"4088247795049920","authorId":"4088247795049920","name":"Yap Wai Hong","avatar":"https://static.tigerbbs.com/e7710b70b68b33532687ac65fc65f96d","crmLevel":3,"crmLevelSwitch":1,"followedFlag":false,"idStr":"4088247795049920","authorIdStr":"4088247795049920"},"themes":[],"htmlText":"<a href=\"https://ttm.financial/S/SE\">$Sea Ltd(SE)$</a>[得意]&nbsp;","listText":"<a href=\"https://ttm.financial/S/SE\">$Sea Ltd(SE)$</a>[得意]&nbsp;","text":"$Sea Ltd(SE)$[得意]","images":[{"img":"https://community-static.tradeup.com/news/40f322622d708a1acd7da562f64cb84a","width":"1080","height":"2874"}],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"link":"https://ttm.financial/post/9916875410","isVote":1,"tweetType":1,"viewCount":485,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":1,"langContent":"EN","totalScore":0},{"id":9913711477,"gmtCreate":1664070617199,"gmtModify":1676537385573,"author":{"id":"4088247795049920","authorId":"4088247795049920","name":"Yap Wai Hong","avatar":"https://static.tigerbbs.com/e7710b70b68b33532687ac65fc65f96d","crmLevel":3,"crmLevelSwitch":1,"followedFlag":false,"idStr":"4088247795049920","authorIdStr":"4088247795049920"},"themes":[],"htmlText":"<a href=\"https://ttm.financial/S/OPEN\">$Opendoor Technologies Inc(OPEN)$</a>[得意]&nbsp;","listText":"<a href=\"https://ttm.financial/S/OPEN\">$Opendoor Technologies Inc(OPEN)$</a>[得意]&nbsp;","text":"$Opendoor Technologies Inc(OPEN)$[得意]","images":[{"img":"https://community-static.tradeup.com/news/970f4753d44ca3db2eea84a63a1721e2","width":"1080","height":"2865"}],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":1,"commentSize":0,"repostSize":0,"link":"https://ttm.financial/post/9913711477","isVote":1,"tweetType":1,"viewCount":614,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":1,"langContent":"EN","totalScore":0},{"id":9910764599,"gmtCreate":1663686469311,"gmtModify":1676537315934,"author":{"id":"4088247795049920","authorId":"4088247795049920","name":"Yap Wai Hong","avatar":"https://static.tigerbbs.com/e7710b70b68b33532687ac65fc65f96d","crmLevel":3,"crmLevelSwitch":1,"followedFlag":false,"idStr":"4088247795049920","authorIdStr":"4088247795049920"},"themes":[],"htmlText":"<a href=\"https://ttm.financial/S/KEN\">$Kenon Holdings Ltd(KEN)$</a>[得意]&nbsp;","listText":"<a href=\"https://ttm.financial/S/KEN\">$Kenon Holdings Ltd(KEN)$</a>[得意]&nbsp;","text":"$Kenon Holdings Ltd(KEN)$[得意]","images":[{"img":"https://community-static.tradeup.com/news/105dbb424913f2e23094b713d061ba12","width":"1080","height":"2766"}],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":1,"commentSize":2,"repostSize":0,"link":"https://ttm.financial/post/9910764599","isVote":1,"tweetType":1,"viewCount":644,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":1,"langContent":"EN","totalScore":0}],"defaultTab":"posts","isTTM":true}