未登录,请登录后再发表信息
最新评论 (0)
播放视频

AI用新的数据集学习看到的常识

AI Learns Visual Common Sense With New Dataset | Two Minute Papers #169

亲爱的学霸们 这里是由Károly Zsolnai-Fehér带来的两分钟论文
Dear Fellow Scholars, this is Two Minute Papers with Károly Zsolnai-Fehér.
今天我们将会来了解一篇
Today, we are going to talk about a new endeavor
关于让学习算法学会更多常识的新论文
to teach some more common sense to learning algorithms.
不知你是否记得 在之前的一集里
If you remember, in an earlier episode,
我们提到了Andrej Karpathy的一篇杰出论文
we talked about an excellent work by Andrej Karpathy,
他创造了一个算法来通过一句完整详尽的话
who built an algorithm that looked at an input image,
描述输入的图片
and described in a full, well-formed sentence what is depicted there.
顺便说一下 他最近成为了特斯拉的AI主管
By the way, he recently became director of AI at Tesla.
在此之前 他刚博士毕业就在OpenAI工作
Before that, he worked at OpenAI, freshly after graduating with a PhD.
这是我见过的最学院派的生涯
Now that is a scholarly career if I’ve ever seen one!
阅读这篇早期的论文时
Reading about this earlier work was one of those moments
我震惊到得拿稳手里的论文 尽力不让自己从椅子上掉下来
when I really had to hold on to my papers not to fall out of the chair,
但当然了 每一个新的突破背后
but of course, as it should be with every new breakthrough,
很多失败的案例都被深入地研究讨论过了
the failure cases were thoroughly discussed.
这篇新论文的其中一个动机是
One of the the motivations for this new work
我们可以通过创建一个视频数据库来改进结果
is that we could improve the results by creating a video database
这个数据库应包含大量常见的 有助于算法学习的事件
that contains a ton of commonly occurring events that would be useful to learn.
这些事件包括 移动和捡拾 或者握住 戳刺 投掷 倾倒
These events include, moving and picking up, or holding, poking, throwing, pouring, or
或者插入不同的东西 还有其他更多的
plugging in different things, and much more.
目标是让这些神经算法能够接受大量训练数据
The goal is that these neural algorithms would get tons of training data for these,
好让它们能够分辨出
and would be able to distinguish
人类是在给它们看些东西还是说只是在移动东西
whether a human is showing them something,or just moving things about.
现存的视频数据库极其缺少这类信息
The already existing video databases are surprisingly sparse in this sort of information,
而在这个最新公布的数据里
and in this new, freshly published dataset,
我们能够基于10万个标记的视频来促进这个方面的研究
we can learn on a 100.000 labeled videos to accelerate research in this direction.
我喜欢看这些论文是如何交织联系在一起的
I love how many of these works are intertwined
喜欢看后续的研究论文是如何尝试去解决先前技术的缺陷
and how followup research works try to address the weaknesses of previous techniques.
一些最初的学习数据集的结果也被发表在论文中 以此开启这方面的研究
Some initial results with learning on this dataset are also reported to kick things off,
如果你看到结果 它们看起来还挺不错
and they seem quite good if you look at the results here,
不过既然这不是论文的焦点
but since this was not the focus of the paper,
我们不该期待有超常的表现
we shouldn’t expect superhuman performance.
然而 几乎所有的相关研究论文都只是垫脚石
However, as almost all papers in research are stepping stones,
再多两篇后续论文将会开辟一片新天地
two more followup papers down the line, this will be an entirely different discussion.
我很乐意为你报道后续进展
I’d love to report back to you on the progress later.
对此超级激动和期待
Super excited for that.
感谢你的观看和慷慨支持下期再见
Thanks for watching and for your generous support, and I’ll see you next time.

发表评论

译制信息
视频概述

利用视频数据库学习人类常见动作并分辨出来

听录译者

收集自网络

翻译译者

One静茹

审核员

豆子

视频来源

https://www.youtube.com/watch?v=XgB3Xg5st2U

相关推荐