Hello steemians, I am Tony and I want to share some news about AI.
大家好 我係Tony 想跟大家分享近期一則有關AI的新聞
Few days ago, in the biggest event of Dota2 scene every year, The International 2017. A show match between a pro player "Dendi" versus the openAI developed by a organization which aim to prevent AI to destroy the world. The show match was so one-sided with a total smash in game one. Dendi forfeited the second match when it was half way done. He said the playstyle of the AI was so weird and made him feel strange and inhuman.
在幾天之前 在dota2年度的最大盛事 世界邀請賽2017 有一場特別的表演場 由知名玩家Dendi 對上由非牟利組織研究 其目的是為了防止人工智能統治世界的openAI 比賽的結果是相當的一面倒 Dendi甚至在第二場比賽的中途就表示電腦的玩法很奇怪而不像人類的感覺 所以投降認輸了
Some people may argue that Dendi is not good enough to challenge the openAI. In fact, more pro players like Sumail and Arteezy from Evil Genius, pajkatt etc have already challenged the AI and they got smashed except pajkatt had defeated the AI once. However, the AI has a self learning system and it absorbed the experience of how they got defeat, so they will not do those mistakes again and become perfect.
可能有人會認為Dendi的實力不足以挑戰openAI 但事實上 有好幾個專業玩家 例如EG的sumail和arteezy, pajkatt等等的玩家 都有跟AI對戰過 但都被打敗了 除了pajkatt曾經戰勝了AI一次 但是AI有自我學習系統 能夠從戰敗中吸取教訓 不會再犯同樣的錯誤 所以AI會變得愈來愈完美
The reason this AI is so powerful is that the developer did not input any dota2 god playing guideline into it. Instead, they only input the basic tutorial of dota2 into it. They let two bots play against each other and learn how to become a "perfect player" including some dota2 knowledges and micro mechanics. The self learning system is actually amazing and it only took two weeks to become a dota2 1v1 master who smashed the real pro players, as a AI.
這個AI的可怕之處 是研發者並沒有把高級的dota2技巧輸入到資料中 而是把兩個AI設定為互相對戰 用兩週的時間 它們自己揣摩到如何能把這遊戲的1v1玩到完美的技巧 這個ai的自我學習系統的實在是太過強大啦
Most of the people have already heard about AlphaGo who defeated some best Go game player in the world. And the target of the developer of AlphaGo is a real time strategies game - Starcraft 2. They said Starcraft 2 is way harder than Go game for the AI as the go game has a ten to the power of 100. however starcraft 2 may be got ten to the power of billion possibilities. But ! This is only because they limited the AI to play at a human level, with a human action per minute, human field of view etc. If they unblock those limits, with the full potential of AI, I think they can also smash human easily in the very near future
很多人都有聽說過Alphago吧 他打敗了很多世界頂級的圍棋選手 然後研發AlphaGo的團隊的下個目標 是一款實時操縱的遊戲 星海爭霸2 他們說 星海比起圍棋更加複雜 圍棋的可能性是10的100次方 但是星海2的複雜度,至少在後面加上100個零 前題是研發者們把AI的設定是在人類玩家的遊玩水平 例如控制速度 全景景像等等 假如把這些限制都解除的話 我估計電腦很快就能虐打人類了
AI is strong because they have big memory and fast calculations etc. They can always achieve perfection to help human for now. A single powerful AI can be stopped. But have you ever think if AI can develop self consciousness and they can cooperate with other AI to form a team, or a group , may be bigger like a community, to take revenge on human ? I know at this point there is nothing we can do, but this is what I always think and worry about in our future
人工智能的強大之處在於他們有人類沒有的龐大記憶 還有快速的運算能力等等 都能夠在不同範疇幫到人類的 但是一個AI可以被停止 如果他們進化到有自我意識呢 然後AI跟AI合作 變成一個小隊 一個組織 或者暗地組成一個小社會再向人類反抗呢? 我清楚我們是沒有辦法能阻止到 但是我就是經常會想和擔憂著
Thank you for reading my blog. I hope you appreciate my sharing. Please hit me with a upvote and a follow : )
感謝你閱讀我的blog 希望你喜歡我的分享 如果你喜歡 請點個upvote和follow我吧 : )
love your post.
Thank you so much for your support : )
1 v 1 暂时是电脑赢,5 v 5 现在还是人类强
對 但如果有一天他們也是合作起來
大概我們就完了哈哈