Interactive transcript · 151 lines
Good morning, and a very warm welcome to TalkCast, your trusted daily digest of innovation, culture,
早上好,热烈欢迎收看 TalkCast,您值得信赖的创新、文化每日文摘与科学。
and science.
和科学。
Today is October 22nd, 2026, and whether you are sipping your morning coffee or embarking on your
今天是2026年10月22日,无论您是在品尝早晨的咖啡还是开始您的
daily commute, we are delighted to have you with us.
每日通勤,我们都非常高兴有您陪伴。
I am Edward.
我是爱德华。
And I am Amelia.
我是阿米莉亚。
Good morning, everyone.
大家早上好。
We have a fascinating spectrum of stories lined up for you this morning, ranging from high
今天早上我们为您准备了一系列迷人的故事,范围从
-performance supercomputing in the freezing craters of the moon to some wonderfully peculiar
月球冰冷的火山口的高性能超级计算到一些非常古怪的
rebellions happening right inside our living rooms. Indeed.
在我们客厅里发生的叛乱。确实如此。
On our docket today, an ambitious lunar rover is set to introduce autonomous generative AI models to
在我们今天的日程上,一辆雄心勃勃的月球车将把自主生成式人工智能模型引入
off -planet exploration.
星球外探索。
Back on Earth, Apple is stirring the tech world with an upcoming event expected to overturn decades
回到地球,苹果公司即将在即将到来的活动中搅动科技界,预计该活动将推翻几十年的
of design dogma.
设计教条。
While consumers are taking matters into their own hands by using artificial intelligence to make
而消费者则通过使用人工智能将事情掌握在自己手中,让
their smart televisions pleasantly dumb again.
他们的智能电视愉快地再次变笨。
Plus, we will look at Spotify tightening the screws on shared subscriptions and Nintendo breathing
另外,我们将关注 Spotify 收紧共享订阅的螺丝,以及任天堂为
new life into a beloved casual gaming phenomenon.
一个备受喜爱的休闲游戏现象注入新的生命。
Let us start with that lunar dispatch, Edward.
让我们从那份月球快讯开始,爱德华。
When we think of artificial intelligence, we usually picture massive data centers consuming
当我们想到人工智能时,我们通常会想象消耗庞大数据中心的
megawatts of electricity here on Earth.
地球上数兆瓦的电力。
But humanity is now literally launching these models into the cosmos.
但人类现在正字面意义上将这些模型发射到宇宙中。
That is precisely what Hewlett -Packard Enterprise and aerospace startup Astrolab are preparing to do.
这正是惠普企业公司(Hewlett-Packard Enterprise)和航空航天初创公司 Astrolab 准备做的事情。
They have engineered a specialized, fanless supercomputer that will hitch a ride aboard a lunar
他们设计了一款专用的、无风扇的超级计算机,它将搭乘月球
rover christened FLIP.
被命名为 FLIP 的漫游车。
The objective is to run large language models and advanced computer vision directly on the lunar
其目标是在月球表面直接运行大语言模型和先进的计算机视觉,
surface, completely unassisted by Earth -bound servers.
完全不受地球服务器的辅助。
Now, wait a moment.
现在,等一片刻。
For those of us who think of an LLM as something we use to draft emails or summarize articles, what
对于我们这些认为 LLM 是用来起草电子邮件或总结文章的工具的人来说,
conceivable purpose does an onboard AI serve for a solitary rover rolling across the lunar regolith?
一个在月球表岩屑上滚动的孤独漫游车,其车载人工智能究竟有什么可想象的用途?
Why not simply process the calculations back down in Houston?
为什么不直接在休斯顿处理这些计算呢?
It all boils down to bandwidth and latency.
这一切都归结为带宽和延迟。
Radio signals traveling between Earth and the moon experience a fundamental delay of several
在地球和月球之间传播的无线电信号会经历几秒钟的基本延迟
seconds, not to mention the reality of intermittent signal blackouts when rovers enter deep craters
更不用说当漫游车进入深层陨石坑或穿过月球地形背面时,间歇性信号中断的现实了。
or pass behind lunar terrain.
或者穿过月球地形。
What engineers are practicing here is the ultimate frontier of edge computing, a concept where you
工程师们在这里实践的是边缘计算的终极前沿,这是一个
process raw information right at the physical location where it is collected, rather than
在收集它的物理位置直接处理原始信息,而不是
transmitting vast volumes of telemetry across millions of kilometers.
跨越数百万公里传输大量的
