In the movies and on TV, robots can talk, think, and follow instructions. The reality, of course, has always been very different. But that may be changing. A new robot, Marge, is part of this change. Reading is one of the many skills that has always been tricky for robots. Sure, they can recognize text, but robot brains don't understand change the way human brains do. Humans can read big and small text, in dim or bright light, at a rock concert or in a soundless room. But for a robot, any change in the surroundings changes its whole activity. A misspelled word was an obstacle. Robots didn't figure out or guess until now. With a software combination of a dictionary and spell-checker, Marge can guess the meanings of misspelled words. She can read a newspaper and use the information in it to answer questions, even if it means dealing with typos or different sized text. Marge's brain is functioning more like a human brain. Marge isn't the only one. Robots are evolving. Scientists are now trying a new technique when designing robots. Instead of trying to create the perfect program for a robot's task, scientists are making robots that learn. This means that scientists input different, shorter programs for different parts of a large task. Then the robot, through trial and error, can evolve to make the best combination of parts. Scientists are using this process to help a robot teach itself to fly. This is very similar to Darwin's process of natural selection. Robots have made a lot of progress in the last 50 years, but they've always been hindered by their difficulty with change. Designing new robots that can adapt, guess, and learn means that huge leaps in robot intelligence might be in our future.
在電影中及電視上，機器人會說話、思考並遵從指示。在現實中，情況當然是大不相同。但這一點或 許正逐漸改變。 新型機器人瑪琪就是這改變的一部分。對機器人來說，閱讀一直是許多棘手的技能之一。機器人當然 能辨識文字，但機器人的腦了解變化的方式，跟人腦不同。人類能讀字體大或小的文字，在昏暗或明亮的 光線下，或在搖滾演唱會、無聲的房間裡都能閱讀。但對機器人來說，周遭環境一改變就會改變它的整個 活動。拼錯的字在過去是個障礙，機器人過去不能「推想」或「猜測」，但現在它們可以了。 瑪琪裝有一套軟體，綜合字典及拼字檢查，因此瑪琪能猜出被拼錯的字的意思。它能讀報紙，並利用 裡頭的資訊來回答問題，這代表它能應付印刷錯誤或不同大小的字體，瑪琪頭腦的運作方式其實是更接近 人腦的。 瑪琪不是唯一的新型機器人，機器人正逐漸進化。目前科學家在設計機器人時嘗試一種新技巧，他 們不再為機器人的某任務創造完美的程式，而是做出會學習的機器人。這表示科學家在大型任務的不同部 分輸入不同且較短的程式，然後，機器人從錯誤中學習，透過反覆試驗進化，做出各部分的最佳組合。科 學家正利用這過程來幫助機器人「學習」飛翔。這過程跟達爾文的物競天擇過程很相近。 機器人在過去50 年中有長足的進步，但它們總是因無法適應改變而受阻。設計能適應、猜測且學習的新機器人，那麼機器人智能在未來就可能會有大躍進。 | 
沒有留言:
張貼留言