WEBVTT 1 00:00:03.399 --> 00:00:07.719 Welcome to Bedtime Astronomy. Explore the wonders of the cosmos 2 00:00:07.759 --> 00:00:12.279 with our soothing Bedtime Astronomie podcast. Each episode offers a 3 00:00:12.359 --> 00:00:16.320 gentle journey through the stars, planets, and beyond, perfect for 4 00:00:16.399 --> 00:00:20.239 unwinding after a long day. Let's travel through the mysteries 5 00:00:20.239 --> 00:00:22.440 of the universe as you drift off into a peaceful 6 00:00:22.480 --> 00:00:23.800 slumber under the night sky. 7 00:00:27.000 --> 00:00:29.879 Imagine for a second that you are well millions of 8 00:00:29.879 --> 00:00:33.679 miles away from Earth right just hurtling through the absolute, 9 00:00:33.920 --> 00:00:37.799 pitch black, freezing void of deep space. Everything is going 10 00:00:37.840 --> 00:00:42.039 perfectly according to plan, and then suddenly something completely. 11 00:00:41.640 --> 00:00:44.000 Unexpected happens always does. 12 00:00:43.920 --> 00:00:46.840 Exactly, like maybe a piece of debris strikes the hull, 13 00:00:46.960 --> 00:00:50.200 or an unprecedented solar flare washes over the ship, or 14 00:00:50.520 --> 00:00:53.039 I don't know, a vital thruster just suddenly misfires. 15 00:00:53.079 --> 00:00:53.799 Oh that's bad. 16 00:00:53.960 --> 00:00:56.359 Yeah, it's bad. Alarms are going off. You desperately need 17 00:00:56.399 --> 00:01:00.359 a solution. But here is the terrifying reality. Because of 18 00:01:00.359 --> 00:01:04.599 the sheer, vastness of space, there is this massive communication delay. 19 00:01:04.840 --> 00:01:08.280 Massive Right, a distress signal sent to Earth might take 20 00:01:08.719 --> 00:01:11.480 say twenty minutes just to get there, and then another 21 00:01:11.519 --> 00:01:14.439 twenty minutes for any kind of response to actually come back. 22 00:01:14.560 --> 00:01:16.840 Yeah, and by the time Earth even knows you're in trouble, 23 00:01:17.159 --> 00:01:19.560 I mean, the mission could already be totally. 24 00:01:19.239 --> 00:01:22.760 Lost, completely lost. The spacecraft just has to figure it 25 00:01:22.799 --> 00:01:24.120 out entirely on its own. 26 00:01:24.879 --> 00:01:27.799 Basically the ultimate locked room puzzle. But you know, the 27 00:01:27.879 --> 00:01:30.319 room is traveling at thousands of miles an hour through 28 00:01:30.319 --> 00:01:33.599 the most hostile environment imagined. Yeah, when a mission is 29 00:01:33.640 --> 00:01:37.799 that far out, autonomy isn't just some like futuristic luxury. 30 00:01:38.280 --> 00:01:42.519 It is a baseline requirement for survival. You simply cannot 31 00:01:42.560 --> 00:01:44.400 wait for a human sitting at a desk in a 32 00:01:44.400 --> 00:01:48.000 control room on Earth to just push a button, so. 33 00:01:48.000 --> 00:01:50.439 To figure out how we survive that scenario. Today, we 34 00:01:50.480 --> 00:01:54.480 are talking about a really monumental technological leap from NASA's 35 00:01:54.560 --> 00:01:58.359 Jet Propulsion Laboratory. We're looking at their new high performance 36 00:01:58.400 --> 00:01:59.959 spaceflight computing processor. 37 00:02:00.159 --> 00:02:01.519 Game changer, it really is. 38 00:02:01.799 --> 00:02:05.319 We are talking about a revolutionary AI space chip designed 39 00:02:05.359 --> 00:02:09.439 to let spacecraft literally think for themselves. So, okay, let's 40 00:02:09.479 --> 00:02:12.879 unpack this because to truly appreciate how transformative this piece 41 00:02:12.879 --> 00:02:15.520 of hardware is, we really need to understand the reality 42 00:02:15.520 --> 00:02:18.240 of how spacewhit computing currently operates, and frankly, the current 43 00:02:18.280 --> 00:02:20.280 state of technology up there might surprise. 44 00:02:20.000 --> 00:02:22.840 A lot of people, Oh absolutely, because you know, people 45 00:02:22.960 --> 00:02:26.960 naturally assume that since we are sending these incredibly sophisticated 46 00:02:27.000 --> 00:02:30.240 billion dollar machines out into the cosmos, they must be 47 00:02:30.319 --> 00:02:34.520 running on the absolute bleeding edge of computer processing. 48 00:02:34.120 --> 00:02:35.759 Right, like they've got the best tech available. 49 00:02:35.960 --> 00:02:40.240 Exactly. You might picture quantum computers or at least processors 50 00:02:40.240 --> 00:02:43.199 that are vastly superior to the laptop sitting on our desks, 51 00:02:43.319 --> 00:02:47.680 But the truth is entirely the opposite. Current space missions 52 00:02:47.719 --> 00:02:51.479 rely on computing cores that are heavily, heavily outdated by 53 00:02:51.520 --> 00:02:53.080 earthly standards. 54 00:02:52.759 --> 00:02:56.240 Which seems completely counterintuitive until you look at the extreme 55 00:02:56.280 --> 00:02:57.879 bottleneck of space exploration. 56 00:02:58.199 --> 00:02:58.439 Right. 57 00:02:58.520 --> 00:03:01.960 The environment beyond our our atmosphere is just utterly brutal 58 00:03:02.080 --> 00:03:06.199 on electronics. I mean, we're talking about intense electromagnetic radiation 59 00:03:06.759 --> 00:03:08.479 constantly bathing the solar. 60 00:03:08.199 --> 00:03:10.240 System, and the temperature swings, yes. 61 00:03:10.080 --> 00:03:13.439 Dramatic temperature fluctuation, swinging from hundreds of degrees above zero 62 00:03:13.599 --> 00:03:16.439 to hundreds of degrees below sometimes in a matter of 63 00:03:16.520 --> 00:03:17.800 like minutes. 64 00:03:18.120 --> 00:03:22.120 Yeah, but you know, the real enemy of modern computing 65 00:03:22.159 --> 00:03:24.240 out there is the high energy particles. 66 00:03:24.479 --> 00:03:26.360 Explain that a bit more. What kind of particles. 67 00:03:26.599 --> 00:03:30.159 Well, those particles are essentially the absolute bane of a 68 00:03:30.199 --> 00:03:34.479 spacelight engineer's existence. We're talking about cosmic rays. They're basically 69 00:03:34.520 --> 00:03:37.759 atomic nuclei traveling at its significant fraction of the speed 70 00:03:37.759 --> 00:03:41.199 of light coming from the Sun and deep space. Okay, 71 00:03:41.439 --> 00:03:44.319 and when one of these particles strikes a delicate modern 72 00:03:44.360 --> 00:03:48.719 computer chip, it does something entirely disruptive. It doesn't necessarily 73 00:03:48.960 --> 00:03:52.120 like leave a physical bullet hole right. Instead, it dumps 74 00:03:52.159 --> 00:03:55.479 a massive amount of kinetic energy directly into the silicon, 75 00:03:55.800 --> 00:03:58.639 and that knocks electrons completely out of their atoms. 76 00:03:59.120 --> 00:04:01.560 Okay, I want to pause here to really visualize this. 77 00:04:02.120 --> 00:04:07.879 You're saying this invisible microscopic collision creates a literal burst 78 00:04:07.879 --> 00:04:10.800 of electrical charge inside the computer processor. 79 00:04:11.000 --> 00:04:15.159 Exactly that sudden burts of stray electrical charge can actually 80 00:04:15.199 --> 00:04:18.480 trick a delicate transistor into changing its state, so it 81 00:04:18.480 --> 00:04:20.439 flips a zero to a one or a one to 82 00:04:20.480 --> 00:04:20.839 a zero. 83 00:04:20.920 --> 00:04:21.360 Oh wow. 84 00:04:21.439 --> 00:04:24.040 Yeah. We call this a single event upset or a bitflip. 85 00:04:24.439 --> 00:04:27.279 And when that happens, the computer might suddenly calculate the 86 00:04:27.319 --> 00:04:30.959 wrong navigational trajectory, or an entire subsystem just gets confused. 87 00:04:31.000 --> 00:04:31.800 That's terrified. 88 00:04:31.920 --> 00:04:35.079 It is. So when a current spacecraft detects this kind 89 00:04:35.120 --> 00:04:39.879 of cascade error, it is programmed to immediately panic and 90 00:04:40.000 --> 00:04:42.199 force itself into what we call safe mode. 91 00:04:42.240 --> 00:04:45.600 And safe mode is basically the spacecraft curling into a 92 00:04:45.680 --> 00:04:47.160 defensive ball to protect itself. 93 00:04:47.199 --> 00:04:48.720 Right, Yeah, that's exactly what it is. Like. 94 00:04:48.800 --> 00:04:52.800 It temporarily shuts down all non essential systems, the science instruments, 95 00:04:52.839 --> 00:04:57.160 turn off, the camera's power down, and the entire multi 96 00:04:57.360 --> 00:04:59.959 billion dollar machine just goes completely dormant. 97 00:05:00.160 --> 00:05:01.279 It just floats there. 98 00:05:01.160 --> 00:05:04.519 Right, just floating, waiting for human engineers back on Earth 99 00:05:04.560 --> 00:05:08.519 to manually analyze the telemetry, find the flipped bit, and 100 00:05:08.560 --> 00:05:10.560 slowly coax the system back to life. 101 00:05:10.600 --> 00:05:14.839 And during that downtime, the spacecraft is effectively blind and paralyzed. 102 00:05:15.560 --> 00:05:18.600 So historically NASA has deliberately made a really severe trade 103 00:05:18.600 --> 00:05:23.199 off to avoid those constant, terrifying system crashes. They use older, 104 00:05:23.319 --> 00:05:24.680 incredibly durable chips. 105 00:05:24.759 --> 00:05:25.480 Older chips. 106 00:05:25.560 --> 00:05:29.240 Yeah, those legacy chips have physically larger transistors, and because 107 00:05:29.279 --> 00:05:32.360 they are larger, it requires significantly more energy to flip 108 00:05:32.360 --> 00:05:34.720 their state from a zero to a one, so they 109 00:05:34.759 --> 00:05:37.120 are highly resistant to those cosmic ray strikes. 110 00:05:37.399 --> 00:05:40.279 So it sounds like we're sending these incredibly vital missions 111 00:05:40.279 --> 00:05:42.639 into the void running on the equivalent of like a 112 00:05:42.759 --> 00:05:44.399 nineteen nineties brick phone. 113 00:05:44.439 --> 00:05:46.000 Basically, yeah, and we use. 114 00:05:45.920 --> 00:05:48.519 Them just because we know the screen won't crack, even 115 00:05:48.560 --> 00:05:51.519 though the mission really demands a modern smartphone. 116 00:05:51.560 --> 00:05:54.439 That's a perfect analogy, and that trade off has been 117 00:05:54.759 --> 00:05:59.399 completely necessary. I mean, a cutting edge, ridiculously fast modern 118 00:05:59.439 --> 00:06:04.199 computer is utterly useless if a single solar flare fries 119 00:06:04.240 --> 00:06:07.560 its microscopic circuits on day two of a ten year emission. 120 00:06:07.680 --> 00:06:08.120 That's fair. 121 00:06:08.199 --> 00:06:11.839 Surviving the environment is the non negotiable baseline. Everything else, 122 00:06:11.879 --> 00:06:16.720 processing speed, artificial intelligence, complex data analysis that has traditionally 123 00:06:16.759 --> 00:06:18.439 been treated as a secondary luxury. 124 00:06:18.720 --> 00:06:21.600 Okay, but I am trying to reason through this engineering 125 00:06:21.639 --> 00:06:24.360 problem here. Sure, if we need modern computing power, but 126 00:06:24.399 --> 00:06:27.439 the microscopic transistors are too fragile, why not just take 127 00:06:27.480 --> 00:06:29.759 a powerful modern chip out of a high end server, 128 00:06:30.199 --> 00:06:32.920 put it inside a really thick lead box to block 129 00:06:32.959 --> 00:06:35.879 the radiation, and just send that into space. Why do 130 00:06:36.000 --> 00:06:39.160 we need to invent a completely new chip from scratch, well. 131 00:06:39.040 --> 00:06:42.920 That immediately runs into the tyranny of the rocket equation. Wait, exactly, 132 00:06:43.600 --> 00:06:47.759 in spaceflight, waight is your absolute worst enemy. Every single 133 00:06:47.839 --> 00:06:50.639 ounce you send into orbit costs thousands of dollars in 134 00:06:50.720 --> 00:06:54.040 rocket fuel. A thick lead box heavy enough to stop 135 00:06:54.120 --> 00:06:58.639 high energy cosmic rays would be well impossibly massive for 136 00:06:58.680 --> 00:07:02.120 a modern deep space That makes sense. And add to that, 137 00:07:02.920 --> 00:07:06.319 heavy metal healing can sometimes cause secondary radiation scatter. 138 00:07:06.439 --> 00:07:07.360 Where what does that mean? 139 00:07:07.600 --> 00:07:10.279 So the cosmic ray hits the lead and it shatters, 140 00:07:10.480 --> 00:07:13.800 creating this sort of shotgun blast of secondary particles that 141 00:07:13.920 --> 00:07:14.839 still hit the computer. 142 00:07:15.000 --> 00:07:17.240 Oh wow, so the armor actually makes it worse in 143 00:07:17.279 --> 00:07:18.279 a way sometimes. 144 00:07:18.360 --> 00:07:20.759 Yeah, you cannot just armoor a regular chip. You have 145 00:07:20.839 --> 00:07:25.160 to build the resilience directly into the silicon architecture itself. 146 00:07:24.800 --> 00:07:28.399 Which bridges us directly to NASA's game changing development program. 147 00:07:28.399 --> 00:07:31.319 Down at the Langley Research Center, they realized we are 148 00:07:32.160 --> 00:07:34.360 hitting a wall with the brick phone approach. 149 00:07:34.519 --> 00:07:34.720 Right. 150 00:07:35.279 --> 00:07:39.920 If we want autonomous spacecraft or vastly faster onboard scientific analysis, 151 00:07:40.160 --> 00:07:43.519 we need high performance without sacrificing that baseline of survival. 152 00:07:44.079 --> 00:07:46.120 And the result of that effort is the high Performance 153 00:07:46.160 --> 00:07:50.560 Spaceflight Computing Processor, the HPSC. Yeah, and the physical form 154 00:07:50.600 --> 00:07:53.639 factor of this hardware is surprising. It is a system 155 00:07:53.680 --> 00:07:56.319 on a chip or and so small enough to fit 156 00:07:56.399 --> 00:07:57.959 completely in the palm of your hand. 157 00:07:58.480 --> 00:08:02.000 The system on a chip architecture is really vital here. Traditionally, 158 00:08:02.360 --> 00:08:05.360 a computer uses a large motherboard. You have your central 159 00:08:05.360 --> 00:08:08.480 processing unit one corner, your memory in another, and your 160 00:08:08.519 --> 00:08:10.079 networking interfaces. 161 00:08:09.600 --> 00:08:10.920 Somewhere else right spread out. 162 00:08:11.040 --> 00:08:14.920 Yeah, and data has to physically travel across wires between 163 00:08:14.959 --> 00:08:18.519 those components, which takes time and energy. And so ce 164 00:08:18.720 --> 00:08:22.480 combines all of those essential components into a single highly 165 00:08:22.480 --> 00:08:26.839 integrated unit. You have your CPUs, your computational offloads, advanced networking, 166 00:08:26.879 --> 00:08:30.319 and memory interfaces all stacked together on one tiny piece 167 00:08:30.360 --> 00:08:30.879 of silicon. 168 00:08:31.120 --> 00:08:33.960 Wait, you mentioned computational offloads. I want to clarify that 169 00:08:34.000 --> 00:08:36.360 before I move on. What does an offload actually do 170 00:08:36.519 --> 00:08:37.440 in this context? 171 00:08:37.600 --> 00:08:41.039 Good question. Think of the main CPU as the head 172 00:08:41.120 --> 00:08:43.639 chef in a busy kitchen. If the head chef has 173 00:08:43.639 --> 00:08:46.000 to chop every single onion and wash every single dish, 174 00:08:46.200 --> 00:08:49.519 the actual complex cooking just grinds to a halt. A 175 00:08:49.559 --> 00:08:54.080 computational offload is like hiring specialized prep cooks. It is 176 00:08:54.159 --> 00:08:57.240 a dedicated piece of hardware. On the chip designed to 177 00:08:57.279 --> 00:09:02.240 handle one specific, highly pensive task, like what like crunching 178 00:09:02.240 --> 00:09:05.919 heavy math for navigation or processing images from a camera, 179 00:09:07.080 --> 00:09:10.720 it offloads that burden from the main CPU, allowing the 180 00:09:10.879 --> 00:09:13.879 entire chip to run vastly faster and more efficiently. 181 00:09:14.159 --> 00:09:16.840 I appreciate that clarity, but I have to push back 182 00:09:16.840 --> 00:09:19.360 on the novelty of this hardware for a second. Yeah, 183 00:09:19.360 --> 00:09:22.159 my tablet and my smartphone use and soci right now. 184 00:09:22.360 --> 00:09:24.320 I mean, that is how we get consumer devices to 185 00:09:24.360 --> 00:09:29.039 be so remarkably thin. What makes NASA's version a technological breakthrough? 186 00:09:29.080 --> 00:09:31.679 If I literally have an Aussie sitting in my pocket right. 187 00:09:31.559 --> 00:09:33.720 Now, well, the chip in your pocket operates in an 188 00:09:33.720 --> 00:09:37.720 incredibly cozy environment. It enjoys the luxury of Earth's magnetic 189 00:09:37.759 --> 00:09:41.000 field protecting it from cosmic radiation, and it just sits 190 00:09:41.000 --> 00:09:42.360 at a comfortable room temperature. 191 00:09:42.399 --> 00:09:42.960 That's true. 192 00:09:43.159 --> 00:09:46.399 NASA's version is radiation hardened from the ground up, utilizing 193 00:09:46.440 --> 00:09:48.240 completely different engineering principles. 194 00:09:48.440 --> 00:09:51.080 So how exactly do they harden it without using that 195 00:09:51.120 --> 00:09:52.799 heavy lead box we talked about earlier. 196 00:09:52.960 --> 00:09:56.519 One of the most fascinating mechanisms is called triple modular redundancy. 197 00:09:56.840 --> 00:09:58.559 Triple modular redundancy. 198 00:09:58.679 --> 00:10:02.039 Yes, Instead of having one microscopic circuit do a calculation, 199 00:10:02.440 --> 00:10:06.360 they build three identical circuits directly into the chip to 200 00:10:06.399 --> 00:10:08.759 do the exact same math simultaneously. 201 00:10:08.879 --> 00:10:10.360 Wait all three at the same time. 202 00:10:10.320 --> 00:10:13.879 Simultaneously, and at the end of the calculation, they essentially 203 00:10:13.960 --> 00:10:14.759 vote on the answer. 204 00:10:14.960 --> 00:10:16.600 Oh that's brilliant, right. 205 00:10:16.639 --> 00:10:19.000 If a cosmic ray strikes one of the circuits and 206 00:10:19.039 --> 00:10:21.720 flips a bit, causing it to output a wrong answer, 207 00:10:22.120 --> 00:10:25.200 the other two circuits will just outvote it. The chip 208 00:10:25.320 --> 00:10:30.000 instantly recognizes the error, ignores the glitch, and keeps functioning perfectly. 209 00:10:30.120 --> 00:10:30.639 Wow. 210 00:10:30.759 --> 00:10:33.879 They also use insulating layers in the silicon itself, often 211 00:10:33.919 --> 00:10:38.039 called silicon on insulator, to physically prevent that stray electrical 212 00:10:38.159 --> 00:10:41.879 charge from jumping between components and short circuiting the system. 213 00:10:42.039 --> 00:10:45.399 So they are essentially weaving fault tolerance into the actual 214 00:10:45.399 --> 00:10:48.039 logic of the machine exactly. And here is where the 215 00:10:48.120 --> 00:10:52.080 leap in capabilities gets really staggering. The design goal for 216 00:10:52.120 --> 00:10:54.639 this project was to deliver up to one hundred times 217 00:10:54.679 --> 00:10:57.840 the computing power of the spaceflight computers we are currently using. 218 00:10:58.080 --> 00:11:00.000 One hundred times is already huge, yeah, But. 219 00:11:00.080 --> 00:11:02.639 When they started putting the prototypes together and looking at 220 00:11:02.679 --> 00:11:05.840 the early data, they reported it is operating at performance 221 00:11:05.960 --> 00:11:10.600 levels roughly five hundred times greater than current radiation hardened chips. 222 00:11:10.639 --> 00:11:14.720 Of five hundredfold increase just completely alters what a spacecraft 223 00:11:14.720 --> 00:11:15.360 can accomplish. 224 00:11:15.440 --> 00:11:18.240 How do we properly visualize an increase of five hundred 225 00:11:18.240 --> 00:11:21.879 times Because people often use the cliche of trading a 226 00:11:21.879 --> 00:11:24.679 bicycle for a fighter jet, but that doesn't really capture 227 00:11:24.720 --> 00:11:26.000 the mechanics of data. 228 00:11:25.759 --> 00:11:28.080 Processing, No, it doesn't. A better way to look at 229 00:11:28.159 --> 00:11:31.759 it is through parallel processing. Imagine you were trying to 230 00:11:31.799 --> 00:11:37.320 find one specific quote hidden somewhere inside a massive library. Okay, 231 00:11:37.679 --> 00:11:40.480 the current space chips are like having a single librarian 232 00:11:40.720 --> 00:11:43.240 who has to read every single book page by page, 233 00:11:43.320 --> 00:11:47.799 one after the other. It is reliable, but painfully slow. 234 00:11:48.559 --> 00:11:51.360 This new chip is the equivalent of unleashing five hundred 235 00:11:51.399 --> 00:11:56.759 highly trained librarians to scan the entire library simultaneously. Tasks 236 00:11:56.799 --> 00:11:58.879 that use take hours can now be completed in a 237 00:11:58.960 --> 00:11:59.960 fraction of a second. 238 00:12:00.200 --> 00:12:02.360 Of course, you don't just build a five hundred librarian 239 00:12:02.399 --> 00:12:04.559 super chip, look at the paper specs and bolt it 240 00:12:04.600 --> 00:12:07.200 onto a rocket. You have to definitively prove it can 241 00:12:07.240 --> 00:12:10.240 survive the environment, oh absolutely, which brings us to the 242 00:12:10.279 --> 00:12:14.120 grueling testing phase happening right now at NASA's Jet Propulsion 243 00:12:14.240 --> 00:12:16.320 Laboratory JPL down in southern California. 244 00:12:16.399 --> 00:12:19.399 Yeah. Jim Butler is the project manager overseeing this at JPL, 245 00:12:19.600 --> 00:12:24.039 and his team's entire job is to literally torture this hardware. 246 00:12:24.120 --> 00:12:25.600 Torture is the right word. 247 00:12:25.679 --> 00:12:29.600 They began testing in February twenty twenty six, and they 248 00:12:29.600 --> 00:12:33.840 are putting these chips through an absolute ringer. I mean 249 00:12:34.240 --> 00:12:37.679 they are placing them in thermal vacuum chambers that simulate 250 00:12:37.720 --> 00:12:42.159 the freezing, airless void of space, baking them to extreme temperatures, 251 00:12:42.159 --> 00:12:45.840 and bombarding them with heavy radiation to see if that 252 00:12:45.840 --> 00:12:47.879 triple modular redundancy actually holds up. 253 00:12:48.039 --> 00:12:51.879 And beyond the environmental torture, they're simulating the immense computational 254 00:12:51.919 --> 00:12:54.879 stress of an actual emission. Let's talk about the specific 255 00:12:54.919 --> 00:12:58.759 high fidelity landing scenarios JPL is using, because what exactly 256 00:12:58.759 --> 00:13:01.480 happens computationally when a machine tries to land on another 257 00:13:01.519 --> 00:13:03.279 planet like Mars, Well. 258 00:13:03.120 --> 00:13:05.759 A planetary landing is often referred to as seven minutes 259 00:13:05.759 --> 00:13:09.639 of terror. Right the spacecraft is hitting the Martian atmosphere 260 00:13:09.679 --> 00:13:12.320 at thousands of miles an hour. In a span of 261 00:13:12.399 --> 00:13:15.799 just a few minutes, it has to deploy supersonic parachutes, 262 00:13:16.080 --> 00:13:21.720 jettison heat shields, fire retro rockets, and simultaneously use optical cameras, 263 00:13:21.799 --> 00:13:25.559 radar and lightar to rapidly scan the terrain below to 264 00:13:25.639 --> 00:13:27.679 avoid crashing into a crater or a boulder. 265 00:13:27.840 --> 00:13:30.759 The sheer volume of raw data flooding into the computer 266 00:13:30.840 --> 00:13:34.159 in those few seconds must be well astronomical. 267 00:13:34.240 --> 00:13:36.639 Oh it is, and if the processor lags for even 268 00:13:36.639 --> 00:13:39.919 a fraction of a second, the mission becomes a crater wow, 269 00:13:40.240 --> 00:13:43.720 which is precisely why JPL is using these specific scenarios. 270 00:13:44.159 --> 00:13:47.879 They are digitally throwing the entire tidal wave of simulated 271 00:13:47.960 --> 00:13:51.360 landing sensor data at this palm size processor, forcing it 272 00:13:51.399 --> 00:13:54.120 to calculate real time hazard avoidance. 273 00:13:53.679 --> 00:13:55.440 While they're torturing it physically too. 274 00:13:55.480 --> 00:13:59.440 Right, while simultaneously shaking it violently on acoustic vibration tables 275 00:13:59.480 --> 00:14:02.360 that mimic the destructive roar of a rocket launch. They 276 00:14:02.399 --> 00:14:04.440 want to see if the chip misses a single clock 277 00:14:04.519 --> 00:14:08.960 cycle under the most extreme physical and computational stress imaginable. 278 00:14:08.519 --> 00:14:12.240 So they're essentially forcing a supercomputer to run a marathon while. 279 00:14:12.240 --> 00:14:13.320 Locked inside a microwave. 280 00:14:13.360 --> 00:14:16.440 Yeah right, a microwave being shaken in a paint mixer, 281 00:14:16.759 --> 00:14:20.000 just to make sure it doesn't blink during a planetary landing, exactly. 282 00:14:20.440 --> 00:14:24.279 And the testing campaign is proving that this chip handles 283 00:14:24.320 --> 00:14:29.440 the load flawlessly. It is calculating precision landing adjustments entirely 284 00:14:29.480 --> 00:14:32.879 on its own, without relying on an earth bound supercomputer. 285 00:14:33.080 --> 00:14:36.360 That's incredible. And when the team at JPL initially kicked 286 00:14:36.399 --> 00:14:40.159 off this rigorous testing protocol, they sent an internal email 287 00:14:40.200 --> 00:14:43.960 to mark the occasion. The subject line was simply, Hello Universe. 288 00:14:44.360 --> 00:14:47.120 I love that. What's fascinating here is that phrase Hello 289 00:14:47.240 --> 00:14:50.600 Universe is a brilliant, subtle nod to the history of 290 00:14:50.639 --> 00:14:51.519 computer science. 291 00:14:51.600 --> 00:14:53.039 Oh like Hello World. 292 00:14:53.240 --> 00:14:56.360 Yes, whenever a student learns to write code, the very 293 00:14:56.399 --> 00:14:59.120 first program they are taught simply prints the words hello 294 00:14:59.200 --> 00:15:01.879 World on the t terminal screen. It is the universal 295 00:15:01.960 --> 00:15:05.039 sign of a system waking up. By expanding it to 296 00:15:05.080 --> 00:15:08.279 Hello Universe, the engineers are acknowledging the gravity of what 297 00:15:08.320 --> 00:15:12.159 they have built. This is the moment robust modern computing 298 00:15:12.279 --> 00:15:15.279 architecture finally scales to the demands of the cosmos. 299 00:15:15.519 --> 00:15:18.799 It is a massive engineering triumph, but NASA is not 300 00:15:18.919 --> 00:15:21.759 doing this in a vacuum. Back in twenty twenty two, 301 00:15:21.960 --> 00:15:26.440 JPL selected an Arizona based company, Microchip Technology, Inc. As 302 00:15:26.480 --> 00:15:27.480 their commercial partner. 303 00:15:27.720 --> 00:15:30.200 Yeah, they actually co funded the research, right. 304 00:15:30.120 --> 00:15:33.120 They co funded the research and development work on the processor, 305 00:15:33.600 --> 00:15:37.559 which opens up a vital question, how exactly does this 306 00:15:37.799 --> 00:15:42.960 unprecedented processing power fundamentally change how we explore Well. 307 00:15:42.799 --> 00:15:47.279 The answer really lies in localized artificial intelligence. Until now, 308 00:15:47.480 --> 00:15:51.480 utilizing AI and deep space was nearly impossible because machine 309 00:15:51.559 --> 00:15:55.559 learning algorithms require massive amounts of rapid computation. 310 00:15:55.200 --> 00:15:58.639 The kind of computation that would instantly overwhelm a legacy spaceship. 311 00:15:58.840 --> 00:16:02.960 Exactly, with this new processor, spacecraft can finally run conference 312 00:16:03.000 --> 00:16:04.919 AI models locally on the fly. 313 00:16:05.039 --> 00:16:07.840 Which brings us full circle to the disaster scenario we 314 00:16:07.879 --> 00:16:10.360 discussed at the beginning of the show. You are millions 315 00:16:10.360 --> 00:16:13.039 of miles away, debreeze strikes the hull, and Earth is 316 00:16:13.039 --> 00:16:16.360 a forty minute round trip message away. With onboard AI, 317 00:16:16.919 --> 00:16:21.000 the spacecraft analyzes the sensor data the millisecond the impact happens, 318 00:16:21.440 --> 00:16:25.159 It instantly realizes a thrust is failing, calculates a new 319 00:16:25.200 --> 00:16:29.919 power distribution to compensate, autonomously fires secondary rockets to stabilize 320 00:16:29.919 --> 00:16:32.639 the spin, and keeps the mission alive, all in real time. 321 00:16:32.799 --> 00:16:36.759 It's amazing and beyond just disaster mitigation. Consider what this 322 00:16:36.840 --> 00:16:41.320 means for actual scientific discovery. Deep space probes collect staggering 323 00:16:41.320 --> 00:16:43.840 amounts of raw data. Oh yeah, think of a probe 324 00:16:43.919 --> 00:16:47.639 orbiting Jupiter looking for a specific split second ice geyser 325 00:16:47.759 --> 00:16:50.720 erupting on the moon Europa. Currently, a probe has to 326 00:16:50.759 --> 00:16:54.720 take thousands of high resolution images and slowly trickle all 327 00:16:54.759 --> 00:16:57.080 of that raw data back to Earth over the deep 328 00:16:57.080 --> 00:16:57.720 space network. 329 00:16:57.720 --> 00:17:00.480 And the deep space network is incredibly reliable, but it 330 00:17:00.519 --> 00:17:05.039 has severe bandwidth limitations. Sending gigabytes of raw photos from 331 00:17:05.119 --> 00:17:07.680 Jupiter is like trying to download a four K movie 332 00:17:07.759 --> 00:17:11.160 over a nineteen nineties dial up modem. It just takes forever, 333 00:17:11.200 --> 00:17:14.359 and humans on Earth have to manually sift through thousands 334 00:17:14.359 --> 00:17:16.720 of photos of empty space just to find the one 335 00:17:16.799 --> 00:17:17.799 image of the geyser. 336 00:17:18.079 --> 00:17:22.240 But with this powerful new processor, the onboard AI serves 337 00:17:22.279 --> 00:17:26.240 as an intelligent curator. It pre processes all the data 338 00:17:26.319 --> 00:17:29.240 right there in Jupiter's orbit. Okay, the AI looks at 339 00:17:29.240 --> 00:17:33.400 the thousand pictures instantly identifies the one single image that 340 00:17:33.480 --> 00:17:38.200 captures the anomaly, discards the useless data, and only transmits 341 00:17:38.200 --> 00:17:41.720 that one highly valuable compressed insight back to Earth. 342 00:17:41.759 --> 00:17:42.559 That's incredible. 343 00:17:42.640 --> 00:17:45.119 It turns a dial up connection into a pipeline of 344 00:17:45.160 --> 00:17:48.079 pure scientific discovery. And we are looking at deploying this 345 00:17:48.119 --> 00:17:52.319 in Earth orbiters tracking rapid climate change, planetary rovers, navigating 346 00:17:52.359 --> 00:17:56.440 the chaotic surface of Mars, and critically keeping living astronauts 347 00:17:56.480 --> 00:17:59.400 safe during crude missions to the Moon where split second 348 00:17:59.400 --> 00:18:01.079 life support is decisions are mandatory. 349 00:18:01.440 --> 00:18:03.680 Now I want to pivot here and ask the really 350 00:18:03.759 --> 00:18:06.680 vital grounding question. It is amazing that a probe near 351 00:18:06.759 --> 00:18:08.880 Jupiter will soon be able to think for itself, and 352 00:18:08.960 --> 00:18:11.440 it is incredible that astronauts will have vastly better life 353 00:18:11.440 --> 00:18:14.759 support computers. Right, but how does this commercial partnership with 354 00:18:14.799 --> 00:18:18.759 Microchip technology benefit you? The listener driving to work right now? 355 00:18:19.039 --> 00:18:22.880 What is the practical earthly application of this highly specific 356 00:18:22.960 --> 00:18:23.799 space hardware. 357 00:18:24.119 --> 00:18:26.920 This is where we see the historical pattern of space 358 00:18:27.000 --> 00:18:34.519 exploration accelerating terrestrial technology. Think about the engineering paradox NASA 359 00:18:34.720 --> 00:18:39.000 forced Microchip technology to solve here. Okay, build a processor 360 00:18:39.079 --> 00:18:43.160 capable of running advanced AI and processing immense amounts of 361 00:18:43.200 --> 00:18:47.559 sensor data instantly, but make it incredibly compact, energy efficient, 362 00:18:47.960 --> 00:18:50.799 and essentially indestructible against systemic failure. 363 00:18:51.000 --> 00:18:53.960 So they demanded indestructible AI computing. 364 00:18:54.119 --> 00:18:58.160 Indestructible. Now take that exact same processor architecture and install 365 00:18:58.200 --> 00:19:01.839 it into a commercial airplane. Oh wow, the aviation industry 366 00:19:01.880 --> 00:19:05.440 requires fault tolerant computing where an in flight system failure 367 00:19:05.519 --> 00:19:08.119 is simply not an option. Or to bring it directly 368 00:19:08.160 --> 00:19:11.640 to your morning commute, consider the chips powering autonomous vehicles. 369 00:19:11.720 --> 00:19:14.279 That makes perfect sense. A self driving car is basically 370 00:19:14.279 --> 00:19:16.839 a terrestrial spacecraft navigating dynamic hazards. 371 00:19:16.880 --> 00:19:19.839 Exactly, the chip sitting behind the dashboard of an autonomous 372 00:19:19.880 --> 00:19:22.720 car driving you to work relies on a massive influx 373 00:19:22.720 --> 00:19:26.480 of data from cameras, lightar, and radar to navigate unpredictable 374 00:19:26.519 --> 00:19:30.039 city streets safely. While it doesn't face the extreme cosmic 375 00:19:30.119 --> 00:19:35.519 radiation of deep space, it absolutely requires that same fault tolerance. 376 00:19:36.119 --> 00:19:39.200 If a component glitches while the car is traveling seventy 377 00:19:39.240 --> 00:19:42.319 miles an hour down a highway, it cannot freeze or 378 00:19:42.440 --> 00:19:43.440 enter a safe mode. 379 00:19:43.599 --> 00:19:44.599 No, definitely not. 380 00:19:44.759 --> 00:19:46.759 It needs to make a life or death decision in 381 00:19:46.799 --> 00:19:49.640 a fraction of a second, without waiting for a slow 382 00:19:50.000 --> 00:19:54.480 cellular cloud connection. By partnering with NASA to build the 383 00:19:54.599 --> 00:19:59.359 ultimate autonomous base brain, microchip is inadvertently designed the perfect 384 00:19:59.440 --> 00:20:02.000 processor for high stakes earthly industries. 385 00:20:02.240 --> 00:20:05.240 So the technological foundation that will land the next advanced 386 00:20:05.319 --> 00:20:08.039 rover on Mars is the exact same architecture that could 387 00:20:08.039 --> 00:20:11.359 eventually be making your drive to the grocery store infinitely safer. 388 00:20:11.440 --> 00:20:12.200 It's all connected. 389 00:20:12.519 --> 00:20:15.680 It really showcases the ultimate return on investment for pushing 390 00:20:15.720 --> 00:20:18.799 the boundaries of engineering. We set out today to understand 391 00:20:18.839 --> 00:20:20.920 how this new chip works, and the journey reveals that 392 00:20:20.960 --> 00:20:23.720 we are watching the evolution of our space program happen 393 00:20:23.880 --> 00:20:25.079 right now in real time. 394 00:20:25.200 --> 00:20:25.799 We really are. 395 00:20:26.079 --> 00:20:30.039 We are officially graduating from sending highly durable but ultimately 396 00:20:30.160 --> 00:20:33.640 remote controlled calculators out into the dark. We are moving 397 00:20:33.680 --> 00:20:38.519 toward deploying brilliant palm sized brains, brains that can pilot 398 00:20:38.559 --> 00:20:42.599 themselves through the void flawlessly, land on alien worlds, and 399 00:20:42.720 --> 00:20:46.839 accelerate our understanding of the universe by analyzing the cosmos 400 00:20:47.079 --> 00:20:48.039 entirely on their own. 401 00:20:48.319 --> 00:20:51.400