神奇！谷歌的AI技术竟能预测病人何时会死？
<p><span word="AI">AI</span> <span word="knows">knows</span> <span word="when">when</span> <span word="you">you</span>'<span word="re">re</span> <span word="going">going</span> <span word="to">to</span> <span word="die">die</span>. <span word="But">But</span> <span word="unlike">unlike</span> <span word="in">in</span> <span word="sci">sci</span>-<span word="fi">fi</span> <span word="movies">movies</span>, <span word="that">that</span> <span word="information">information</span> <span word="could">could</span> <span word="end">end</span> <span word="up">up</span> <span word="saving">saving</span> <span word="lives">lives</span>.</p><p>人工智能知道你的死亡时间。但和科幻电影不同的是，这些信息最后可能会挽救生命。</p><p><span word="A">A</span> <span word="new">new</span> <span word="paper">paper</span> <span word="published">published</span> <span word="in">in</span> <span word="Nature">Nature</span> <span word="suggests">suggests</span> <span word="that">that</span> <span word="feeding">feeding</span> <span word="electronic">electronic</span> <span word="health">health</span> <span word="record">record</span> <span word="data">data</span> <span word="to">to</span> <span word="a">a</span> <span word="deep">deep</span> <span word="learning">learning</span> <span word="model">model</span> <span word="could">could</span> <span word="substantially">substantially</span> <span word="improve">improve</span> <span word="the">the</span> <span word="accuracy">accuracy</span> <span word="of">of</span> <span word="projected">projected</span> <span word="outcomes">outcomes</span>.</p><p>一篇发表在《自然》杂志上的新论文显示，深度学习模型的进食电子健康数据能够大幅提高预测结果的准确性。</p><p><span word="In">In</span> <span word="trials">trials</span> <span word="using">using</span> <span word="data">data</span> <span word="from">from</span> <span word="two">two</span> <span word="US">US</span> <span word="hospitals">hospitals</span>, <span word="researchers">researchers</span> <span word="were">were</span> <span word="able">able</span> <span word="to">to</span> <span word="show">show</span> <span word="that">that</span> <span word="these">these</span> <span word="algorithms">algorithms</span> <span word="could">could</span> <span word="predict">predict</span> <span word="a">a</span> <span word="patient">patient</span>'<span word="s">s</span> <span word="length">length</span> <span word="of">of</span> <span word="stay">stay</span> <span word="and">and</span> <span word="time">time</span> <span word="of">of</span> <span word="discharge">discharge</span>, <span word="but">but</span> <span word="also">also</span> <span word="the">the</span> <span word="time">time</span> <span word="of">of</span> <span word="death">death</span>.</p><p>在使用了两家美国医院数据的试验中，研究人员可以证明这些算法不仅能够预测病人的住院天数和出院时间，还能预测死亡时间。</p><p><span word="The">The</span> <span word="neural">neural</span> <span word="network">network</span> <span word="described">described</span> <span word="in">in</span> <span word="the">the</span> <span word="study">study</span> <span word="uses">uses</span> <span word="an">an</span> <span word="immense">immense</span> <span word="amount">amount</span> <span word="of">of</span> <span word="data">data</span>, <span word="such">such</span> <span word="as">as</span> <span word="a">a</span> <span word="patient">patient</span>'<span word="s">s</span> <span word="vitals">vitals</span> <span word="and">and</span> <span word="medical">medical</span> <span word="history">history</span>, <span word="to">to</span> <span word="make">make</span> <span word="its">its</span> <span word="predictions">predictions</span>.</p><p>研究中所描述的神经网络使用了大量的数据进行预测，比如病人的生命特征和病史。</p><p><span word="A">A</span> <span word="new">new</span> <span word="algorithm">algorithm</span> <span word="lines">lines</span> <span word="up">up</span> <span word="previous">previous</span> <span word="events">events</span> <span word="of">of</span> <span word="each">each</span> <span word="patient">patient</span>'<span word="s">s</span> <span word="records">records</span> <span word="into">into</span> <span word="a">a</span> <span word="timeline">timeline</span>, <span word="which">which</span> <span word="allowed">allowed</span> <span word="the">the</span> <span word="deep">deep</span> <span word="learning">learning</span> <span word="model">model</span> <span word="to">to</span> <span word="pinpoint">pinpoint</span> <span word="future">future</span> <span word="outcomes">outcomes</span>, <span word="including">including</span> <span word="time">time</span> <span word="of">of</span> <span word="death">death</span>.</p><p>一种新的算法将每个病人所记录的活动经历排列成一个时间轴，这使得深度学习模型能够确定包括死亡时间在内的未来的结果。</p>
页:
[1]