Placeholder Image

字幕列表 影片播放

自動翻譯
  • Today, artificial intelligence helps doctors diagnose patients,

    如今,人工智能幫助醫生診斷病人。

  • pilots fly commercial aircraft, and city planners predict traffic.

    飛行員駕駛商用飛機,城市規劃者預測交通。

  • But no matter what these AIs are doing, the computer scientists who designed them

    但無論這些人工智能在做什麼,設計它們的計算機科學家們

  • likely don't know exactly how they're doing it.

    很可能不知道他們到底是怎麼做的。

  • This is because artificial intelligence is often self-taught,

    這是因為人工智能往往是自學成才。

  • working off a simple set of instructions

    照本宣科

  • to create a unique array of rules and strategies.

    以創造一系列獨特的規則和策略。

  • So how exactly does a machine learn?

    那麼機器到底是如何學習的呢?

  • There are many different ways to build self-teaching programs.

    構建自學計劃有很多不同的方式。

  • But they all rely on the three basic types of machine learning:

    但它們都依賴於三種基本類型的機器學習。

  • unsupervised learning, supervised learning, and reinforcement learning.

    無監督學習、監督學習和強化學習。

  • To see these in action,

    要看到這些行動。

  • let's imagine researchers are trying to pull information

    讓我們想象一下,研究人員正在試圖提取資訊。

  • from a set of medical data containing thousands of patient profiles.

    從一組包含數千份患者資料的醫療數據中提取。

  • First up, unsupervised learning.

    首先是無監督學習。

  • This approach would be ideal for analyzing all the profiles

    這種方法對於分析所有的剖面圖是非常理想的。

  • to find general similarities and useful patterns.

    以找到一般的相似性和有用的模式。

  • Maybe certain patients have similar disease presentations,

    也許某些患者有類似的疾病表現。

  • or perhaps a treatment produces specific sets of side effects.

    或者說一種治療方法會產生特定的副作用。

  • This broad pattern-seeking approach can be used to identify similarities

    這種廣義的模式搜索方法可以用來識別相似性。

  • between patient profiles and find emerging patterns,

    在患者檔案之間尋找新的模式。

  • all without human guidance.

    都沒有人指導。

  • But let's imagine doctors are looking for something more specific.

    但我們想象一下,醫生要找的是更具體的東西。

  • These physicians want to create an algorithm

    這些醫生希望建立一個算法

  • for diagnosing a particular condition.

    用於診斷某一特定病症。

  • They begin by collecting two sets of data

    他們首先收集了兩組數據-

  • medical images and test results from both healthy patients

    健康患者的醫學影像和檢驗結果

  • and those diagnosed with the condition.

    和那些被診斷出的病情。

  • Then, they input this data into a program

    然後,他們將這些數據輸入到一個程序中

  • designed to identify features shared by the sick patients

    設計用於識別病患者的共同特徵

  • but not the healthy patients.

    但健康的病人卻沒有。

  • Based on how frequently it sees certain features,

    根據它看到某些功能的頻率。

  • the program will assign values to those features' diagnostic significance,

    程序將為這些特徵的診斷意義賦值。

  • generating an algorithm for diagnosing future patients.

    產生一個算法,用於診斷未來的病人。

  • However, unlike unsupervised learning,

    然而,與無監督學習不同。

  • doctors and computer scientists have an active role in what happens next.

    醫生和計算機科學家對接下來發生的事情有積極的作用。

  • Doctors will make the final diagnosis

    醫生會做出最終診斷

  • and check the accuracy of the algorithm's prediction.

    並檢查算法預測的準確性。

  • Then computer scientists can use the updated datasets

    然後,計算機科學家可以使用更新的數據集。

  • to adjust the program's parameters and improve its accuracy.

    以調整程序的參數,提高其準確性。

  • This hands-on approach is called supervised learning.

    這種實踐的方法叫做監督學習。

  • Now, let's say these doctors want to design another algorithm

    現在,我們假設這些醫生想設計另一種算法

  • to recommend treatment plans.

    來建議治療方案。

  • Since these plans will be implemented in stages,

    由於這些計劃將分階段實施。

  • and they may change depending on each individual's response to treatments,

    而且它們可能會根據每個人對治療的反應而改變。

  • the doctors decide to use reinforcement learning.

    醫生們決定使用強化學習。

  • This program uses an iterative approach to gather feedback

    該程序使用迭代方法來收集反饋意見。

  • about which medications, dosages and treatments are most effective.

    關於哪些藥物、劑量和治療方法最有效。

  • Then, it compares that data against each patient's profile

    然後,它將這些數據與每個病人的檔案進行比較。

  • to create their unique, optimal treatment plan.

    來制定自己獨特的最佳治療方案。

  • As the treatments progress and the program receives more feedback,

    隨著治療的進展和項目收到更多的反饋。

  • it can constantly update the plan for each patient.

    它可以不斷更新每個病人的計劃。

  • None of these three techniques are inherently smarter than any other.

    這三種技術本質上都不比其他技術聰明。

  • While some require more or less human intervention,

    雖然有些需要或多或少的人為干預。

  • they all have their own strengths and weaknesses

    尺有所短,寸有所長

  • which makes them best suited for certain tasks.

    這使得它們最適合某些任務。

  • However, by using them together,

    然而,通過將它們一起使用。

  • researchers can build complex AI systems,

    研究人員可以構建複雜的人工智能系統。

  • where individual programs can supervise and teach each other.

    在這裡,各個項目可以互相監督和教導。

  • For example, when our unsupervised learning program

    例如,當我們的無監督學習程序

  • finds groups of patients that are similar,

    找到相似的患者群體。

  • it could send that data to a connected supervised learning program.

    它可以將這些數據發送到一個連接的監督學習程序。

  • That program could then incorporate this information into its predictions.

    然後,該程序可以將這些資訊納入其預測中。

  • Or perhaps dozens of reinforcement learning programs

    或者可能是幾十個強化學習程序。

  • might simulate potential patient outcomes

    可能模擬潛在的病人結果

  • to collect feedback about different treatment plans.

    以收集對不同治療方案的反饋意見。

  • There are numerous ways to create these machine-learning systems,

    創建這些機器學習系統的方法有很多。

  • and perhaps the most promising models

    也可能是最有前途的模式

  • are those that mimic the relationship between neurons in the brain.

    是那些模仿大腦神經元之間關係的。

  • These artificial neural networks can use millions of connections

    這些人工神經網絡可以使用數以百萬計的連接。

  • to tackle difficult tasks like image recognition, speech recognition,

    以應對圖像識別、語音識別等困難任務。

  • and even language translation.

    甚至語言翻譯。

  • However, the more self-directed these models become,

    然而,這些模式越是自我導向。

  • the harder it is for computer scientists

    計算機科學家越難

  • to determine how these self-taught algorithms arrive at their solution.

    以確定這些自學算法是如何得出其解決方案的。

  • Researchers are already looking at ways to make machine learning more transparent.

    研究人員已經在研究如何讓機器學習更加透明。

  • But as AI becomes more involved in our everyday lives,

    但隨著人工智能越來越多地參與到我們的日常生活中。

  • these enigmatic decisions have increasingly large impacts

    這些神祕的決定產生了越來越大的影響

  • on our work, health, and safety.

    對我們的工作、健康和安全。

  • So as machines continue learning to investigate, negotiate and communicate,

    所以隨著機器不斷學習調查、協商和交流。

  • we must also consider how to teach them to teach each other to operate ethically.

    我們還必須考慮如何教導他們,讓他們互相教導經營道德。

Today, artificial intelligence helps doctors diagnose patients,

如今,人工智能幫助醫生診斷病人。

字幕與單字
自動翻譯

影片操作 你可以在這邊進行「影片」的調整,以及「字幕」的顯示

B1 中級 中文 TED-Ed 學習 監督 程序 人工 診斷

人工智能如何學習?- Briana Brownell (How does artificial intelligence learn? - Briana Brownell)

  • 4 2
    林宜悉 發佈於 2021 年 03 月 11 日
影片單字