Placeholder Image

字幕列表 影片播放

  • If you really want to understand what's on a written page right now we need a human to

  • read it. That's very slow. We will get, I believe technology very soon that

  • can actually do a pretty good job of reading natural language documents. Watson shows that

  • that's feasible. If you consider Watson was actually able to read Wikipedia and understand

  • it well enough to play a game of this Jeopardy, which is a complex, subtle, ambiguous game

  • of language and got a higher score than the best human players put together.

  • And its knowledge was not programmed fact by fact in Lisp or some computer language.

  • It just read Wikipedia and other encyclopedias, 200 million pages of natural language documents

  • and didn't do a perfect job of understanding it and didn't do a perfect job of answering

  • the questions but was better than the best human players put together from having read

  • natural language documents. That's very impressive.

  • That's coming to a search engine near you. There's major search engines like Google are

  • not just going to be using keywords with synonyms, they're going to actually read for understanding

  • the concepts 'cause if you think about searching there's a lot of information now that's ignored,

  • which is the meaning of all these documents which is why they were created.

  • So if you can have a computer even do a job that's very mediocre compared to human but

  • then can apply the scale of computation. I mean, Watson -- if it read one page, it's

  • not as good as you are, but it didn't read one page -- it read 200 million pages. You

  • and I can't begin to do that. Watson's out reading all medical literature, every medical

  • journal article, every medical book, major medical blogs and will be an expert

  • diagnostician and medical consultant that has read everything. No human can do that.

  • So that's where we're headed. Our search engines will actually also know us very well. We will

  • let them listen in on conversations; verbal, written. They'll watch everything we're reading

  • and writing and saying and hearing, and then they'll be like an assistant. It'll say, "Oh,

  • you know, you were talking about how you can get the supplement phosphatidylcholine into

  • the cells yesterday in that conversation with Joe. You know, there's research that came

  • out 13 minutes ago that speaks to that."

  • It'll be an assistant that helps you through the day, will answer your questions before

  • you ask them or even before you realize you have a question, and you'll just get used

  • to this information popping up that you wanted and you'll be frustrated if you're thinking

  • about something and it doesn't immediately pop up without you even having to ask for

  • it. I'm not actually predicting that until 2029 that we will match human intelligence,

  • but we can nonetheless do things that humans can't do. I mean, Watson, if it read one page,

  • as I said, wouldn't be as strong as you or I, but it was able to read hundreds of millions

  • of pages and it's ability to read each page is going to increase. So that's where we're

  • headed. But then a comment on that is it's not an alien invasion of these intelligent

  • machines to displace us. We will use them to make ourselves smarter, which is what we

  • do today.

If you really want to understand what's on a written page right now we need a human to

字幕與單字

單字即點即查 點擊單字可以查詢單字解釋

A2 初級

雷-庫茲韋爾:你的未來機器人助手 (Ray Kurzweil: Your Robot Assistant of the Future)

  • 58 6
    VoiceTube 發佈於 2021 年 01 月 14 日
影片單字