1 00:00:00,882 --> 00:00:03,632 (exciting music) 2 00:00:10,050 --> 00:00:11,460 - Welcome to "From the Crow's Nest," 3 00:00:11,460 --> 00:00:15,090 a podcast on electromagnetic spectrum operations or EMSO. 4 00:00:15,090 --> 00:00:18,000 I'm your host, Ken Miller, Director of Advocacy and Outreach 5 00:00:18,000 --> 00:00:19,710 for the Association of Old Crows. 6 00:00:19,710 --> 00:00:20,820 Thanks for listening. 7 00:00:20,820 --> 00:00:23,340 In this episode, I sit down with Dr. Joseph Guerci 8 00:00:23,340 --> 00:00:25,230 to discuss the evolution and outlook 9 00:00:25,230 --> 00:00:28,230 for cognitive electronic warfare systems. 10 00:00:28,230 --> 00:00:31,020 Dr. Guerci is an internationally recognized leader 11 00:00:31,020 --> 00:00:32,978 in the research and development and transition 12 00:00:32,978 --> 00:00:36,214 of next generation sensor and cognitive systems. 13 00:00:36,214 --> 00:00:38,190 He has over 30 years of experience 14 00:00:38,190 --> 00:00:40,680 in government, industry, and academia, 15 00:00:40,680 --> 00:00:43,440 including a seven-year term with DARPA, 16 00:00:43,440 --> 00:00:46,320 the Defense Advanced Research Projects Agency, 17 00:00:46,320 --> 00:00:48,000 where he concluded his service 18 00:00:48,000 --> 00:00:51,178 as the Director of the Special Projects Office. 19 00:00:51,178 --> 00:00:53,730 Before I get to him, just a couple quick items. 20 00:00:53,730 --> 00:00:54,930 First, believe it or not, 21 00:00:54,930 --> 00:00:57,513 Congress just recently this weekend 22 00:00:57,513 --> 00:01:00,150 completed its long overdue work 23 00:01:00,150 --> 00:01:03,381 on the Fiscal 2024 federal budget 24 00:01:03,381 --> 00:01:05,203 by passing a funding agreement 25 00:01:05,203 --> 00:01:08,520 before embarking on a two-week recess, 26 00:01:08,520 --> 00:01:10,311 or what they call district work period. 27 00:01:10,311 --> 00:01:13,421 The spending package is $1.2 trillion, 28 00:01:13,421 --> 00:01:17,160 and includes 824.3 billion 29 00:01:17,160 --> 00:01:19,722 for the Pentagon and defense related activities. 30 00:01:19,722 --> 00:01:23,858 That's about 26.8 billion more than Fiscal Year 2023, 31 00:01:23,858 --> 00:01:25,800 or a 3% increase. 32 00:01:25,800 --> 00:01:27,330 I'm in the process of putting together 33 00:01:27,330 --> 00:01:30,300 some of the summary details that we'll put that out 34 00:01:30,300 --> 00:01:34,710 in the form of an article in AOC's weekly eCrow Newsletter 35 00:01:34,710 --> 00:01:36,362 probably next week, 36 00:01:36,362 --> 00:01:37,860 and so be on the lookout for that. 37 00:01:37,860 --> 00:01:39,540 It'll have a little bit more detail. 38 00:01:39,540 --> 00:01:41,403 If you're a regular listener to the show you'll know, 39 00:01:41,403 --> 00:01:43,832 we've been critical of Congress in the past. 40 00:01:43,832 --> 00:01:48,832 We can talk all day about total funding, spending increases, 41 00:01:48,930 --> 00:01:51,120 percentages greater than last year and so forth, 42 00:01:51,120 --> 00:01:54,090 but the reality is that when you don't pass a budget 43 00:01:54,090 --> 00:01:55,767 for six months into the fiscal year, 44 00:01:55,767 --> 00:01:58,410 it really becomes less about the total money 45 00:01:58,410 --> 00:02:00,689 and more about the delays in program development, 46 00:02:00,689 --> 00:02:02,051 missed milestones, 47 00:02:02,051 --> 00:02:05,130 the loss in testing and training, et cetera. 48 00:02:05,130 --> 00:02:06,480 So it goes without saying that it's important 49 00:02:06,480 --> 00:02:08,130 for Congress to figure out a way 50 00:02:08,130 --> 00:02:11,280 to get at least the defense appropriations done on time. 51 00:02:11,280 --> 00:02:12,639 You would think it shouldn't be that hard. 52 00:02:12,639 --> 00:02:16,090 You know, the total federal budget is $6.3 trillion, 53 00:02:16,090 --> 00:02:19,860 but only about 1.7 trillion of that is discretionary, 54 00:02:19,860 --> 00:02:21,600 and that's the part that Congress funds 55 00:02:21,600 --> 00:02:23,190 through its annual appropriations process. 56 00:02:23,190 --> 00:02:25,836 So that's only about 27% of the federal budget. 57 00:02:25,836 --> 00:02:29,718 And of that, about 50% of that is the defense bill. 58 00:02:29,718 --> 00:02:33,060 So if Congress figures out a way 59 00:02:33,060 --> 00:02:34,777 to just pass the defense bill, 60 00:02:34,777 --> 00:02:39,777 you're talking close to 87%, 90% of the federal budget 61 00:02:40,440 --> 00:02:42,420 basically done and running on time. 62 00:02:42,420 --> 00:02:45,210 So, of course, you know, defense being the largest pot 63 00:02:45,210 --> 00:02:47,100 of discretionary funding opens it up 64 00:02:47,100 --> 00:02:49,710 for being used for a lot of political leverage, 65 00:02:49,710 --> 00:02:52,860 but, you know, this is a critical piece of legislation 66 00:02:52,860 --> 00:02:55,200 that Congress has to focus on each year. 67 00:02:55,200 --> 00:02:57,030 That being said, we already know 68 00:02:57,030 --> 00:02:59,839 that don't expect anything in 2025, 69 00:02:59,839 --> 00:03:01,950 This agreement that just passed 70 00:03:01,950 --> 00:03:05,558 will fund the Defense Department through September 30. 71 00:03:05,558 --> 00:03:08,341 This being an election year with Congress and the President, 72 00:03:08,341 --> 00:03:11,580 we can expect CRs to continue, 73 00:03:11,580 --> 00:03:15,030 so Congress will pass CR on September 30 74 00:03:15,030 --> 00:03:17,130 to fund probably until after the election, 75 00:03:17,130 --> 00:03:18,994 and we'll see what happens after that. 76 00:03:18,994 --> 00:03:22,650 But hopefully Congress can figure out a way 77 00:03:22,650 --> 00:03:26,310 to do a better job at getting the defense bill done on time. 78 00:03:26,310 --> 00:03:29,580 I saw an interesting chart in the news the other day, 79 00:03:29,580 --> 00:03:32,580 and it actually highlighted the problem 80 00:03:32,580 --> 00:03:37,320 that we face on this front, because you would think that... 81 00:03:37,320 --> 00:03:39,060 I always was under the impression 82 00:03:39,060 --> 00:03:40,380 that Congress was a little bit better 83 00:03:40,380 --> 00:03:42,720 at getting bills done on time than it actually was, 84 00:03:42,720 --> 00:03:45,878 but if you look at over since 1977, 85 00:03:45,878 --> 00:03:48,960 Congress has only passed all of its appropriations bills 86 00:03:48,960 --> 00:03:53,074 on time four times since 1977, 87 00:03:53,074 --> 00:03:58,074 and that was in '77, '89, '95, and '97. 88 00:03:58,078 --> 00:04:01,932 Other years, various bills have passed and others have not, 89 00:04:01,932 --> 00:04:06,318 the defense bill being the most regularly passed bill, 90 00:04:06,318 --> 00:04:11,318 but still, the trend since about 2007 has not been good. 91 00:04:11,760 --> 00:04:15,510 So for the past 15, 20 years, it's been quite abysmal. 92 00:04:15,510 --> 00:04:17,730 So, you know, Congress has its work cut out for it, 93 00:04:17,730 --> 00:04:19,590 but hopefully, you know, they can figure out a way 94 00:04:19,590 --> 00:04:21,150 to, you know, focus their energies 95 00:04:21,150 --> 00:04:22,980 on that bill moving forward. 96 00:04:22,980 --> 00:04:24,930 The only other quick item I wanted to mention is, 97 00:04:24,930 --> 00:04:27,420 as you know, we've been releasing bonus episodes 98 00:04:27,420 --> 00:04:29,100 every two weeks on the off weeks 99 00:04:29,100 --> 00:04:31,110 of our regular "From the Crow's Nest" episodes. 100 00:04:31,110 --> 00:04:34,272 These are going to form our subscription episodes 101 00:04:34,272 --> 00:04:39,169 available to only paid subscribers and AOC members. 102 00:04:39,169 --> 00:04:40,620 Now, we are in the process 103 00:04:40,620 --> 00:04:43,740 of putting that subscription together behind a paywall, 104 00:04:43,740 --> 00:04:47,250 but until it does, they are free for everyone, 105 00:04:47,250 --> 00:04:48,830 so feel free to download 'em and listen. 106 00:04:48,830 --> 00:04:50,598 The added benefit of this 107 00:04:50,598 --> 00:04:54,011 is that if you are a subscriber or an AOC member, 108 00:04:54,011 --> 00:04:57,150 or until we get it to the pay wall, it's open to everyone, 109 00:04:57,150 --> 00:04:59,790 if you want to listen in on the live recording 110 00:04:59,790 --> 00:05:03,660 of that episode from the audience and actually participate, 111 00:05:03,660 --> 00:05:06,171 you can comment, ask questions, and so forth, 112 00:05:06,171 --> 00:05:07,380 you can do that. 113 00:05:07,380 --> 00:05:08,954 So if you're interested in listening 114 00:05:08,954 --> 00:05:11,880 or participating in the live recording 115 00:05:11,880 --> 00:05:15,330 of those bonus episodes, please go to crows.org, 116 00:05:15,330 --> 00:05:16,710 contact us through that, 117 00:05:16,710 --> 00:05:19,701 and we'll get you the link to participate. 118 00:05:19,701 --> 00:05:24,701 Our next recording for the bonus episode is next Tuesday, 119 00:05:24,900 --> 00:05:29,520 I believe it's April 2 at 1:00 PM Eastern Standard Time. 120 00:05:29,520 --> 00:05:31,620 And with that, I'd like to introduce 121 00:05:31,620 --> 00:05:35,400 or welcome my guest for the show today, Dr. Joseph Guerci. 122 00:05:35,400 --> 00:05:37,590 He is an internationally recognized leader 123 00:05:37,590 --> 00:05:39,840 in research and development and transition 124 00:05:39,840 --> 00:05:42,723 of next generation sensor and cognitive systems. 125 00:05:42,723 --> 00:05:45,308 I had the pleasure of having Dr. Guerci 126 00:05:45,308 --> 00:05:48,569 on AOC webinars sometime last year 127 00:05:48,569 --> 00:05:51,210 to talk about cognitive radar, cognitive EW. 128 00:05:51,210 --> 00:05:54,300 It was a fascinating webinar. 129 00:05:54,300 --> 00:05:57,780 As we were going through it, I realized that this is a topic 130 00:05:57,780 --> 00:05:59,400 that for those of you 131 00:05:59,400 --> 00:06:01,440 who are familiar with our AOC webinars, 132 00:06:01,440 --> 00:06:03,570 it's a little bit more of a technical discussion. 133 00:06:03,570 --> 00:06:04,983 We have presentation slides and so forth 134 00:06:04,983 --> 00:06:07,320 that you can kind of go in deeper, 135 00:06:07,320 --> 00:06:09,510 but I really felt like there was a need 136 00:06:09,510 --> 00:06:13,080 to have this as a topic of conversation on our podcast, 137 00:06:13,080 --> 00:06:15,750 where you can kind of look at it at the 30,000-foot level, 138 00:06:15,750 --> 00:06:18,510 kind of come to a better understanding 139 00:06:18,510 --> 00:06:21,720 of what we mean when we talk about cognitive systems, 140 00:06:21,720 --> 00:06:23,280 because we hear about them a lot, 141 00:06:23,280 --> 00:06:25,411 but I don't think we really dive into what they mean. 142 00:06:25,411 --> 00:06:26,430 I know I don't. 143 00:06:26,430 --> 00:06:29,100 I use the word and I don't really understand the depth 144 00:06:29,100 --> 00:06:30,450 of what those terms mean. 145 00:06:30,450 --> 00:06:32,640 So I reached out to Dr. Guerci, 146 00:06:32,640 --> 00:06:33,824 asked him to be on my podcast, 147 00:06:33,824 --> 00:06:36,810 and Joe, it's great to have you here 148 00:06:36,810 --> 00:06:37,860 on "From the Crows Nest," 149 00:06:37,860 --> 00:06:38,910 and thank you for taking time 150 00:06:38,910 --> 00:06:40,800 out of your schedule to join us. 151 00:06:40,800 --> 00:06:42,263 - Thank you, Ken. It's great to be here. 152 00:06:42,263 --> 00:06:43,980 - All right, so just to get started, 153 00:06:43,980 --> 00:06:45,092 you know, as I mentioned, 154 00:06:45,092 --> 00:06:48,994 this is a topic that mystifies a lot of people. 155 00:06:48,994 --> 00:06:53,817 We use the terms, we think we know what we're talking about, 156 00:06:53,817 --> 00:06:55,998 but oftentimes we're kind of mixing up definitions. 157 00:06:55,998 --> 00:06:58,329 You've been in this field for decades, 158 00:06:58,329 --> 00:07:00,813 and it's really not a new field. 159 00:07:00,813 --> 00:07:03,180 We've thought about cognitive systems 160 00:07:03,180 --> 00:07:06,333 and been working on them for what, 20, 30 years, 161 00:07:07,710 --> 00:07:10,500 but recently, within the last 10, 162 00:07:10,500 --> 00:07:13,380 it's kind of picked up speed. 163 00:07:13,380 --> 00:07:16,770 So could you give us a little bit of insight 164 00:07:16,770 --> 00:07:21,090 into why cognitive systems are picking up pace 165 00:07:21,090 --> 00:07:24,137 in terms of their opportunity being in the spotlight today 166 00:07:24,137 --> 00:07:27,840 and kind of where we came from over the last 20, 30 years? 167 00:07:27,840 --> 00:07:30,843 - Thanks, Ken. That framed the question really well. 168 00:07:32,010 --> 00:07:34,017 I will answer it directly, but let me just first talk 169 00:07:34,017 --> 00:07:36,363 about the definition of cognitive. 170 00:07:37,860 --> 00:07:42,360 Unfortunately, the word cognitive is not that well known, 171 00:07:42,360 --> 00:07:45,210 I would say, or defined by the general populace, 172 00:07:45,210 --> 00:07:46,298 and that's especially true 173 00:07:46,298 --> 00:07:50,100 when interacting with our international colleagues. 174 00:07:50,100 --> 00:07:53,760 You know, translation, many things get lost in translation. 175 00:07:53,760 --> 00:07:55,890 So cognition is actually 176 00:07:55,890 --> 00:07:58,946 a very well-defined scientific term 177 00:07:58,946 --> 00:08:03,946 in the biological and psychiatric sciences. 178 00:08:04,234 --> 00:08:07,739 So it relates to all of the things we think about 179 00:08:07,739 --> 00:08:10,470 of a person who's awake and functional, 180 00:08:10,470 --> 00:08:15,470 like problem solving, memory, things of that sort. 181 00:08:15,870 --> 00:08:19,076 And in fact, the very first page, 182 00:08:19,076 --> 00:08:22,050 the first chapter of my book on cognitive radar, 183 00:08:22,050 --> 00:08:25,110 which was first written back in 2010, 184 00:08:25,110 --> 00:08:26,400 I actually used 185 00:08:26,400 --> 00:08:30,300 the National Institute of Health's definition of cognition, 186 00:08:30,300 --> 00:08:32,940 and then you can very clearly see 187 00:08:32,940 --> 00:08:36,859 how that translates into engineering parlance. 188 00:08:36,859 --> 00:08:38,580 Now, a lot of the confusion 189 00:08:38,580 --> 00:08:41,010 is because people think the word cognitive 190 00:08:41,010 --> 00:08:44,250 is synonymous with consciousness. 191 00:08:44,250 --> 00:08:47,940 Science does not have a definition of consciousness, 192 00:08:47,940 --> 00:08:51,330 but it absolutely has a definition of cognition. 193 00:08:51,330 --> 00:08:54,580 And so to your question, you know, why... 194 00:08:54,580 --> 00:08:57,423 It's been around for a while. That is absolutely true. 195 00:08:58,410 --> 00:09:01,680 We didn't exactly call it cognitive radar and EW 196 00:09:01,680 --> 00:09:04,720 back in the early 2000s, but that's what we were doing, 197 00:09:04,720 --> 00:09:07,830 especially during my time at DARPA. 198 00:09:07,830 --> 00:09:10,710 And why were we doing it and why is it so popular now? 199 00:09:10,710 --> 00:09:14,010 And that's really because of a couple of factors. 200 00:09:14,010 --> 00:09:16,819 One, the electromagnetic spectrum environment 201 00:09:16,819 --> 00:09:21,690 is ever more congested and ever more contested. 202 00:09:21,690 --> 00:09:22,800 Lots going on. 203 00:09:22,800 --> 00:09:24,220 If you turn on your antenna, 204 00:09:24,220 --> 00:09:26,460 you're gonna see all kinds of things. 205 00:09:26,460 --> 00:09:27,892 And if you transmit, 206 00:09:27,892 --> 00:09:31,039 you're gonna elicit all sorts of responses. 207 00:09:31,039 --> 00:09:34,440 And it's becoming ever more increasingly difficult 208 00:09:34,440 --> 00:09:35,433 to sort that out. 209 00:09:36,450 --> 00:09:39,510 The problem is even further exacerbated 210 00:09:39,510 --> 00:09:43,608 by the fact that we have lots of flexibility now globally 211 00:09:43,608 --> 00:09:46,440 to transmit arbitrary waveforms, 212 00:09:46,440 --> 00:09:49,950 to transmit signals that hadn't been seen before. 213 00:09:49,950 --> 00:09:52,990 You know, and back in the old days, 30, 40 years ago, 214 00:09:52,990 --> 00:09:56,130 especially radar transmitters were much more constrained 215 00:09:56,130 --> 00:09:59,310 in what they could do in terms of agility, 216 00:09:59,310 --> 00:10:01,680 and they were much more easily identifiable. 217 00:10:01,680 --> 00:10:03,330 But today, that's all gone, 218 00:10:03,330 --> 00:10:07,170 and that's due to solid state front as replacing tubes 219 00:10:07,170 --> 00:10:11,490 and magnetrons and klystrons and things like that, 220 00:10:11,490 --> 00:10:14,988 and of course, lots of embedded computing, 221 00:10:14,988 --> 00:10:17,940 digital arbitrary waveform generators, 222 00:10:17,940 --> 00:10:22,188 and so you need to actually make decisions on the fly, 223 00:10:22,188 --> 00:10:24,600 and you can only make decisions on the fly 224 00:10:24,600 --> 00:10:28,560 if you understand what you're looking at. 225 00:10:28,560 --> 00:10:32,640 And what really where cognitive really comes in 226 00:10:32,640 --> 00:10:35,070 and really separates itself 227 00:10:35,070 --> 00:10:38,730 is this idea that you don't just sit there passively 228 00:10:38,730 --> 00:10:40,530 and listen and sort things out, 229 00:10:40,530 --> 00:10:44,100 but you actually interact with the environment, 230 00:10:44,100 --> 00:10:45,390 you probe it, 231 00:10:45,390 --> 00:10:48,788 and that probing can give you a lot of information. 232 00:10:48,788 --> 00:10:50,550 And that idea of probing 233 00:10:50,550 --> 00:10:52,040 and enhancing your understanding 234 00:10:52,040 --> 00:10:54,772 of the complete radar or jamming channel 235 00:10:54,772 --> 00:10:58,500 is at the heart of cognitive waveform. 236 00:10:58,500 --> 00:11:00,521 - And so we've seen this, you know, 237 00:11:00,521 --> 00:11:02,490 particularly over the last 10 years. 238 00:11:02,490 --> 00:11:04,410 You know, we talk about the spectrum 239 00:11:04,410 --> 00:11:06,839 being, you know, contested and congested, 240 00:11:06,839 --> 00:11:09,450 and the complexity of trying to figure out 241 00:11:09,450 --> 00:11:12,660 how to transmit and receive the signals. 242 00:11:12,660 --> 00:11:15,600 And so if I'm understanding you correctly, 243 00:11:15,600 --> 00:11:18,720 it's, we oftentimes, I think when we talk about AI, 244 00:11:18,720 --> 00:11:20,400 we talk about it in terms of speed, 245 00:11:20,400 --> 00:11:22,050 but it's not just in terms of speed, 246 00:11:22,050 --> 00:11:23,508 it's in terms of understanding 247 00:11:23,508 --> 00:11:25,770 and breaking down that complexity 248 00:11:25,770 --> 00:11:29,190 and being able to more effectively 249 00:11:29,190 --> 00:11:33,430 or adapt to a certain environment 250 00:11:33,430 --> 00:11:38,280 to transmit or receive or collect that signal. 251 00:11:38,280 --> 00:11:39,690 Is that correct? 252 00:11:39,690 --> 00:11:43,650 - Yeah, so in the old days, you know, not that long ago, 253 00:11:43,650 --> 00:11:46,997 you know, one would go out with a database 254 00:11:46,997 --> 00:11:49,430 of what you expect to see, 255 00:11:49,430 --> 00:11:54,090 and that database would, you know, provide a of prescription 256 00:11:54,090 --> 00:11:57,030 of how you're supposed to behave in response to that. 257 00:11:57,030 --> 00:11:59,640 As I mentioned before, due to all of the advances 258 00:11:59,640 --> 00:12:04,640 that I mentioned and other issues, that just is obsolete. 259 00:12:04,710 --> 00:12:09,210 You know, you literally have to get on station, 260 00:12:09,210 --> 00:12:14,210 see what's going on, and do a lot more of that analysis 261 00:12:14,370 --> 00:12:17,643 that used to be done offline, online. 262 00:12:18,900 --> 00:12:20,100 So it's almost like you need 263 00:12:20,100 --> 00:12:22,410 to take all our subject matter experts 264 00:12:22,410 --> 00:12:25,080 down at the EW program office 265 00:12:25,080 --> 00:12:27,210 and have them with you right in the plane, 266 00:12:27,210 --> 00:12:28,800 making decisions in real time. 267 00:12:28,800 --> 00:12:32,412 But obviously that's not practical for a lot of reasons, 268 00:12:32,412 --> 00:12:34,860 because sometimes the speed involved 269 00:12:34,860 --> 00:12:37,710 needs to be faster than humans can process. 270 00:12:37,710 --> 00:12:40,500 But that idea is basically what the whole point is, right? 271 00:12:40,500 --> 00:12:44,520 It's basically to embed subject matter expertise, 272 00:12:44,520 --> 00:12:47,760 brain power into the embedded computing 273 00:12:47,760 --> 00:12:50,836 so that on the fly you can reprogram 274 00:12:50,836 --> 00:12:53,880 and update your databases and respond. 275 00:12:53,880 --> 00:12:56,178 I mean, that's the best way to look at it. 276 00:12:56,178 --> 00:12:59,586 The old OODA loop, if you will, 277 00:12:59,586 --> 00:13:03,330 was we go up, we fly, we see things, we come down, 278 00:13:03,330 --> 00:13:06,600 we update the database, so next time we go out, we're ready. 279 00:13:06,600 --> 00:13:07,440 That's all gone. 280 00:13:07,440 --> 00:13:10,700 I mean, you have to have that OODA loop on the plane. 281 00:13:10,700 --> 00:13:14,250 Now, I could get into a lot of technical detail 282 00:13:14,250 --> 00:13:16,470 about, well, how on earth do you do that? 283 00:13:16,470 --> 00:13:18,751 And that gets into the weeds obviously, 284 00:13:18,751 --> 00:13:22,110 of the nature of the signals you're observing. 285 00:13:22,110 --> 00:13:24,630 And, you know, at least today 286 00:13:24,630 --> 00:13:27,420 we have to obey the laws of physics as we know it, 287 00:13:27,420 --> 00:13:31,020 and we also have to obey the laws of information theory. 288 00:13:31,020 --> 00:13:33,120 Just because you have a high power transmitter 289 00:13:33,120 --> 00:13:36,810 doesn't mean necessarily that they have the waveform 290 00:13:36,810 --> 00:13:39,662 that would really help them do tracking or detection. 291 00:13:39,662 --> 00:13:42,300 You know, there's physics and information theory, 292 00:13:42,300 --> 00:13:45,208 and those are the anchors that you start from 293 00:13:45,208 --> 00:13:48,744 when trying to do this OODA loop, if you will, on the fly. 294 00:13:48,744 --> 00:13:52,290 - So could you go into that a little bit more, 295 00:13:52,290 --> 00:13:53,550 because I feel like anytime 296 00:13:53,550 --> 00:13:57,968 that there's kind of two forces at play, 297 00:13:57,968 --> 00:14:00,137 we tend to confuse what we're focusing on 298 00:14:00,137 --> 00:14:04,230 when you talk about, you know, information and so forth, 299 00:14:04,230 --> 00:14:06,840 and the waveform and what capability 300 00:14:06,840 --> 00:14:08,518 an adversarial threat it has. 301 00:14:08,518 --> 00:14:13,518 Could you talk a little bit about how that has shown itself 302 00:14:14,160 --> 00:14:17,040 in terms of some of the development that you've done, 303 00:14:17,040 --> 00:14:18,180 how we've adapted to that, 304 00:14:18,180 --> 00:14:20,880 and kind of where that's leading us? 305 00:14:20,880 --> 00:14:22,794 - In the context of electronic warfare, 306 00:14:22,794 --> 00:14:26,280 a very key question is what is the nature 307 00:14:26,280 --> 00:14:28,140 of that signal that you're observing? 308 00:14:28,140 --> 00:14:33,140 So a good example would be, let's say it's a spoof jammer, 309 00:14:33,210 --> 00:14:37,380 like a DRFM, a digital RF memory type of thing, right? 310 00:14:37,380 --> 00:14:40,620 So on a radar scope, you would see, let's say targets, 311 00:14:40,620 --> 00:14:43,406 right, that look like targets you may be interested in, 312 00:14:43,406 --> 00:14:45,107 but are they real? 313 00:14:45,107 --> 00:14:48,060 As you know, smart jamming can spoof you, 314 00:14:48,060 --> 00:14:50,375 leak you fake things that are not there. 315 00:14:50,375 --> 00:14:54,030 And so while I can't get into all the details 316 00:14:54,030 --> 00:14:55,500 on how you might address that 317 00:14:55,500 --> 00:14:57,876 for obvious reasons of sensitivity, 318 00:14:57,876 --> 00:15:02,876 it does lend itself to a proactive probing approach 319 00:15:03,930 --> 00:15:06,660 where you don't just sit there passively 320 00:15:06,660 --> 00:15:08,520 and accept the world as it's presented to you, 321 00:15:08,520 --> 00:15:13,520 but you actually then design probes to answer questions. 322 00:15:13,824 --> 00:15:16,800 Is it a real target or is it a false target 323 00:15:16,800 --> 00:15:19,680 is a really good example 324 00:15:19,680 --> 00:15:24,680 of probably one of the most advanced cognitive EW functions, 325 00:15:25,770 --> 00:15:28,620 for example, you know, in the context 326 00:15:28,620 --> 00:15:30,469 of sorting out jamming signals. 327 00:15:30,469 --> 00:15:32,700 And of course, and if you flip the coin around 328 00:15:32,700 --> 00:15:34,470 and you're the jammer, 329 00:15:34,470 --> 00:15:37,050 then you're also playing Spy v. Spy, right? 330 00:15:37,050 --> 00:15:38,515 You're trying to figure out 331 00:15:38,515 --> 00:15:42,421 whether or not what you are doing is having an effect. 332 00:15:42,421 --> 00:15:44,498 That's a very difficult thing these days, 333 00:15:44,498 --> 00:15:47,160 especially, you know, if you're anything other 334 00:15:47,160 --> 00:15:49,268 than an extremely high powered jammer, 335 00:15:49,268 --> 00:15:54,210 understanding the effect that you're having on a radar 336 00:15:54,210 --> 00:15:56,640 is not easy, generally speaking. 337 00:15:56,640 --> 00:15:59,130 And so you have to get very sophisticated 338 00:15:59,130 --> 00:16:02,737 in the way that you probe and then listen 339 00:16:02,737 --> 00:16:05,700 and then update your knowledge base 340 00:16:05,700 --> 00:16:07,560 as to what you're dealing with. 341 00:16:07,560 --> 00:16:10,158 You know, again, what enables all that, 342 00:16:10,158 --> 00:16:12,510 highly flexible front ends 343 00:16:12,510 --> 00:16:16,530 so that you can transmit arbitrary types of waveforms 344 00:16:16,530 --> 00:16:17,963 that have interesting properties 345 00:16:17,963 --> 00:16:20,627 and elicit interesting responses, 346 00:16:20,627 --> 00:16:25,170 coupled with lots and lots of high power embedded computing, 347 00:16:25,170 --> 00:16:27,950 and that's where the intersection of AI 348 00:16:27,950 --> 00:16:30,990 and cognitive EW takes place, right? 349 00:16:30,990 --> 00:16:33,103 Because that's a lot of heavy lifting, 350 00:16:33,103 --> 00:16:36,330 it's a lot of pattern recognition, for example, 351 00:16:36,330 --> 00:16:37,592 a lot of inference. 352 00:16:37,592 --> 00:16:40,470 And of course, as you know, the deep learning techniques 353 00:16:40,470 --> 00:16:43,281 certainly have a play in that arena. 354 00:16:43,281 --> 00:16:46,410 And that's why I think you're beginning to see more and more 355 00:16:46,410 --> 00:16:49,260 of this intersection of AI and cognitive EW. 356 00:16:49,260 --> 00:16:51,803 But again, I must stress that cognitive systems 357 00:16:51,803 --> 00:16:55,053 have existed, as you pointed out, 358 00:16:55,890 --> 00:16:58,680 long before the modern deep learning AI 359 00:16:58,680 --> 00:17:00,690 has sort of become all the rage. 360 00:17:00,690 --> 00:17:02,880 - I wanted to ask you a little bit more 361 00:17:02,880 --> 00:17:04,110 about the relationship 362 00:17:04,110 --> 00:17:08,400 between AI, deep learning, and cognitive systems. 363 00:17:08,400 --> 00:17:11,550 Before we got on the show here, you had mentioned 364 00:17:11,550 --> 00:17:15,510 that, you know, cognitive systems are a consumer of AI, 365 00:17:15,510 --> 00:17:19,110 and I was wondering if you could go into a little bit more 366 00:17:19,110 --> 00:17:20,700 about discussing how those tie in, 367 00:17:20,700 --> 00:17:24,461 because the two, they're not synonymous. 368 00:17:24,461 --> 00:17:27,120 And so talk a little bit about that relationship 369 00:17:27,120 --> 00:17:28,860 and what we're maybe getting wrong 370 00:17:28,860 --> 00:17:30,931 that we have to rethink or re-understand 371 00:17:30,931 --> 00:17:34,740 to truly grasp what we're dealing with 372 00:17:34,740 --> 00:17:36,813 when we talk about cognitive systems. 373 00:17:37,710 --> 00:17:38,543 - Well, thank you for that question, 374 00:17:38,543 --> 00:17:41,160 because this is really, really important. 375 00:17:41,160 --> 00:17:43,697 A lot of people think cognitive is synonymous with AI, 376 00:17:43,697 --> 00:17:45,690 and it is not. 377 00:17:45,690 --> 00:17:47,490 Again, I went through great pains 378 00:17:47,490 --> 00:17:49,830 to do scientific definitions of cognition. 379 00:17:49,830 --> 00:17:53,400 In Dr. Karen Haigh's book on cognitive EW 380 00:17:53,400 --> 00:17:56,160 she spends a lot of time also trying to clarify that. 381 00:17:56,160 --> 00:18:00,180 But unfortunately, you know, it's kind of easy, 382 00:18:00,180 --> 00:18:02,400 'cause both sound like, you know, 383 00:18:02,400 --> 00:18:06,360 intelligent machine learning cognition, that sounds like AI. 384 00:18:06,360 --> 00:18:08,589 All right, so again, cognition 385 00:18:08,589 --> 00:18:13,589 is basically performing a subject matter expertise function, 386 00:18:13,860 --> 00:18:14,880 but on the fly. 387 00:18:14,880 --> 00:18:19,880 Think of it as having the world's best EW analyst brain 388 00:18:20,040 --> 00:18:21,865 encapsulated on the plane. 389 00:18:21,865 --> 00:18:24,090 And that's a form of expert systems, 390 00:18:24,090 --> 00:18:26,100 you know, that is a form of AI, 391 00:18:26,100 --> 00:18:28,050 and traditionally that's what what we use. 392 00:18:28,050 --> 00:18:30,960 But cognitive systems is definitely, as you said, 393 00:18:30,960 --> 00:18:32,910 a consumer of AI. 394 00:18:32,910 --> 00:18:34,380 It's not synonymous with it. 395 00:18:34,380 --> 00:18:37,200 You can do cognitive radar and EW 396 00:18:37,200 --> 00:18:40,710 without any of the modern deep learning AI techniques. 397 00:18:40,710 --> 00:18:42,180 Absolutely. 398 00:18:42,180 --> 00:18:44,790 But however, incorporating modern AI techniques, 399 00:18:44,790 --> 00:18:46,950 would that make it even better? 400 00:18:46,950 --> 00:18:50,400 Potentially, yes. The answer is probably yes. 401 00:18:50,400 --> 00:18:55,020 A good example would be looking for certain features, right, 402 00:18:55,020 --> 00:18:56,073 in a signal. 403 00:18:57,270 --> 00:18:59,550 That's one thing that we care a lot about. 404 00:18:59,550 --> 00:19:00,974 There's certain features we look for 405 00:19:00,974 --> 00:19:04,110 to be able to sort it out, identify it. 406 00:19:04,110 --> 00:19:06,330 And I think you're probably well aware 407 00:19:06,330 --> 00:19:10,550 that modern AI is very, very good at pattern recognition 408 00:19:10,550 --> 00:19:13,456 when given the appropriate training data 409 00:19:13,456 --> 00:19:15,990 or training environment. 410 00:19:15,990 --> 00:19:20,430 And so rather than writing out thousands of lines 411 00:19:20,430 --> 00:19:25,430 of C++ code to do pattern recognition, feature recognition, 412 00:19:27,570 --> 00:19:29,850 the AI algorithm can implement that 413 00:19:29,850 --> 00:19:31,829 in a deep learning network. 414 00:19:31,829 --> 00:19:34,800 And by the way, blindingly fast, 415 00:19:34,800 --> 00:19:36,242 especially if you implement it 416 00:19:36,242 --> 00:19:39,420 on what are called neuromorphic chips, 417 00:19:39,420 --> 00:19:41,570 that we talked a bit about that previously. 418 00:19:42,418 --> 00:19:43,324 I mean, that's... 419 00:19:43,324 --> 00:19:44,607 - We'll get to that in just a moment. 420 00:19:44,607 --> 00:19:46,560 I knew that's where we were heading. 421 00:19:46,560 --> 00:19:48,960 - Yeah. But again, you're right. 422 00:19:48,960 --> 00:19:50,790 Thank you for making the point 423 00:19:50,790 --> 00:19:54,162 that they're not synonymous, cognition and AI, 424 00:19:54,162 --> 00:19:55,920 and thank you for making the point 425 00:19:55,920 --> 00:19:59,070 that cognitive systems are a consumer of AI. 426 00:19:59,070 --> 00:20:00,780 They don't need AI to be cognitive, 427 00:20:00,780 --> 00:20:03,890 but as I just mentioned in that example, 428 00:20:03,890 --> 00:20:06,413 they actually can be quite empowering. 429 00:20:06,413 --> 00:20:10,593 - So with the notion that AI can empower cognitive systems, 430 00:20:10,593 --> 00:20:13,710 and you mentioned that it would probably be better, 431 00:20:13,710 --> 00:20:16,290 and I would tend to agree, 432 00:20:16,290 --> 00:20:21,290 but in what ways might it not improve cognitive systems? 433 00:20:22,405 --> 00:20:25,290 What can you do without AI today 434 00:20:25,290 --> 00:20:26,700 in terms of cognitive systems 435 00:20:26,700 --> 00:20:31,700 that really AI doesn't really help you achieve your goal? 436 00:20:32,013 --> 00:20:33,720 - I think it's best summed up 437 00:20:33,720 --> 00:20:36,637 with the old computer science adage, 438 00:20:36,637 --> 00:20:38,307 "Garbage in, garbage out." 439 00:20:39,360 --> 00:20:40,712 A lot of people don't understand 440 00:20:40,712 --> 00:20:44,523 that this revolution AI is all about deep learning. 441 00:20:45,360 --> 00:20:47,615 These are convolutional neural networks, 442 00:20:47,615 --> 00:20:52,260 and how good they are depends on, 443 00:20:52,260 --> 00:20:54,289 and this can't be emphasized enough, 444 00:20:54,289 --> 00:20:57,183 how well they were trained. 445 00:20:58,620 --> 00:21:00,480 By the way, it shouldn't be surprising 446 00:21:00,480 --> 00:21:02,370 because we are neural networks. 447 00:21:02,370 --> 00:21:03,870 Our brains are neural networks, 448 00:21:03,870 --> 00:21:06,390 and if you weren't brought up properly 449 00:21:06,390 --> 00:21:09,630 and educated properly, well, you get what you get, right? 450 00:21:09,630 --> 00:21:12,423 So the analogy is exactly the same. 451 00:21:13,410 --> 00:21:17,610 And so the danger with AI is, 452 00:21:17,610 --> 00:21:19,964 well, how do you know that that training data 453 00:21:19,964 --> 00:21:22,697 really represents what it's gonna see 454 00:21:22,697 --> 00:21:24,988 when the shooting starts? 455 00:21:24,988 --> 00:21:27,738 (exciting music) 456 00:21:29,220 --> 00:21:31,620 - I wanted to touch on the training aspect, 457 00:21:31,620 --> 00:21:33,030 because we talk a lot in EW 458 00:21:33,030 --> 00:21:35,430 about we need to have realistic threat environments. 459 00:21:35,430 --> 00:21:39,780 We need to, you know, really train like we fight. 460 00:21:39,780 --> 00:21:41,250 And that's becoming increasingly hard. 461 00:21:41,250 --> 00:21:43,110 We opened the show, we talked about, you know, 462 00:21:43,110 --> 00:21:45,693 how crazy the spectrum is these days. 463 00:21:47,470 --> 00:21:52,470 You can't base your AI on realistic, 464 00:21:53,430 --> 00:21:56,141 you can't get maybe that realistic data from training 465 00:21:56,141 --> 00:21:59,693 without kind of rethinking 466 00:21:59,693 --> 00:22:02,100 how you conduct training in general, 467 00:22:02,100 --> 00:22:04,280 and that's where it gets into modeling and simulation. 468 00:22:04,280 --> 00:22:06,390 And I think this touches, 469 00:22:06,390 --> 00:22:09,270 then this gets into the neuromorphic chip concept. 470 00:22:09,270 --> 00:22:11,460 And that's where I wanna kind of go to, 471 00:22:11,460 --> 00:22:14,910 like how do we model and simulate a training environment 472 00:22:14,910 --> 00:22:16,650 that gives us the data that we need, 473 00:22:16,650 --> 00:22:18,990 that we know it's accurate, our systems can learn from, 474 00:22:18,990 --> 00:22:21,390 so that when we do go into the real fight, 475 00:22:21,390 --> 00:22:24,810 we are actually using the knowledge 476 00:22:24,810 --> 00:22:28,260 that we've built through the training. 477 00:22:28,260 --> 00:22:33,060 - Yeah, that is the key to the whole problem. 478 00:22:33,060 --> 00:22:34,470 You put your finger right on it. 479 00:22:34,470 --> 00:22:38,940 So AI is very good at recognizing faces, for example. 480 00:22:38,940 --> 00:22:42,990 We all have iPhones and what have you, and why is that? 481 00:22:42,990 --> 00:22:47,073 Because it's had plenty of controlled training data. 482 00:22:48,120 --> 00:22:50,340 This is a face, this is a nose, these are, you know, 483 00:22:50,340 --> 00:22:54,480 so sometimes we call that annotated training data, whatever. 484 00:22:54,480 --> 00:22:56,490 And the same goes with ChatGPT, right? 485 00:22:56,490 --> 00:22:59,340 It's had access to terabytes of plain text. 486 00:22:59,340 --> 00:23:01,499 So there's copious amounts of training data, 487 00:23:01,499 --> 00:23:05,880 and the quality of the training data was good enough, 488 00:23:05,880 --> 00:23:07,995 as you can see in the results, right? 489 00:23:07,995 --> 00:23:11,070 The conundrum for military applications, 490 00:23:11,070 --> 00:23:13,595 and especially for electronic warfare, 491 00:23:13,595 --> 00:23:17,655 is of course, wait a second, where am I gonna get terabytes 492 00:23:17,655 --> 00:23:22,655 of training data that is absolutely reflective and realistic 493 00:23:22,919 --> 00:23:27,810 of the environment I expect to see when I'm fighting, 494 00:23:27,810 --> 00:23:31,200 including all the potential unpredictable things 495 00:23:31,200 --> 00:23:32,670 that may pop up, new, you know, 496 00:23:32,670 --> 00:23:35,400 they're called war reserve modes, for example. 497 00:23:35,400 --> 00:23:37,323 Well, I can give you an answer. 498 00:23:37,323 --> 00:23:39,899 It's called digital engineering, 499 00:23:39,899 --> 00:23:44,373 and that's both a cop out and an answer at the same time. 500 00:23:46,170 --> 00:23:50,730 Digital engineering is all about replicating digitally, 501 00:23:50,730 --> 00:23:53,130 including modeling and simulation, 502 00:23:53,130 --> 00:23:55,380 all the physics, all the nuances, 503 00:23:55,380 --> 00:23:57,990 all the technical constraints associated with your system. 504 00:23:57,990 --> 00:24:00,041 And as, you know, we've discussed in the past, 505 00:24:00,041 --> 00:24:04,110 B-21 bomber, next generation air dominance, 506 00:24:04,110 --> 00:24:05,940 have taken advantage of digital engineering 507 00:24:05,940 --> 00:24:08,280 to speed up their acquisition cycles. 508 00:24:08,280 --> 00:24:12,960 Well, those same types of tools can be co-opted 509 00:24:12,960 --> 00:24:14,760 to train AI systems, 510 00:24:14,760 --> 00:24:16,740 since if you have such an accurate 511 00:24:16,740 --> 00:24:18,630 synthetic digital environment, 512 00:24:18,630 --> 00:24:22,600 why not immerse the AI into that environment 513 00:24:22,600 --> 00:24:27,000 and let it learn and train it, you know, 514 00:24:27,000 --> 00:24:28,860 to what it would expect to see in the real world. 515 00:24:28,860 --> 00:24:31,380 And I think as we've mentioned previously, 516 00:24:31,380 --> 00:24:34,980 in our previous discussions, you know, we have done that. 517 00:24:34,980 --> 00:24:35,827 In my own research, 518 00:24:35,827 --> 00:24:40,078 we have advanced radio frequency design tools that we use 519 00:24:40,078 --> 00:24:42,390 where we can really replicate... 520 00:24:42,390 --> 00:24:43,890 By the way, there's site-specific too, 521 00:24:43,890 --> 00:24:47,499 so you can pick a part of the world where you wanna operate, 522 00:24:47,499 --> 00:24:50,700 put emitters down, create that complex, 523 00:24:50,700 --> 00:24:53,190 congested contested environment, 524 00:24:53,190 --> 00:24:55,110 and create copious amounts of data. 525 00:24:55,110 --> 00:24:56,916 It is synthetic, admittedly, 526 00:24:56,916 --> 00:25:01,050 but it is definitely next generation modern syn capability. 527 00:25:01,050 --> 00:25:03,930 So if I had to give you an off-the-cuff answer, 528 00:25:03,930 --> 00:25:06,049 were passing each other on the elevator, 529 00:25:06,049 --> 00:25:08,960 how do we solve the training problem 530 00:25:08,960 --> 00:25:10,950 for military applications, 531 00:25:10,950 --> 00:25:13,590 the simple answer would be digital engineering, 532 00:25:13,590 --> 00:25:15,357 as I've sort of outlined a little bit. 533 00:25:15,357 --> 00:25:16,680 - And you mentioned it is being used, 534 00:25:16,680 --> 00:25:18,053 and I've referenced it. 535 00:25:18,053 --> 00:25:19,380 I read an article 536 00:25:19,380 --> 00:25:22,410 about the next generation air dominance fighter, 537 00:25:22,410 --> 00:25:24,090 I think that's what it's called. 538 00:25:24,090 --> 00:25:24,923 - Yes. 539 00:25:24,923 --> 00:25:25,787 - And they were saying 540 00:25:25,787 --> 00:25:28,557 it's gonna start replacing the F-22s around 2030, 541 00:25:28,557 --> 00:25:29,437 and I'm sitting there, I'm like, 542 00:25:29,437 --> 00:25:30,870 "How in the world is that gonna happen?" 543 00:25:30,870 --> 00:25:35,010 And I was not thinking digital engineering at the time. 544 00:25:35,010 --> 00:25:37,200 I was like, "They need at least another 15 years." 545 00:25:37,200 --> 00:25:38,870 I mean, this is DOD. - Yeah. 546 00:25:38,870 --> 00:25:39,810 - And when we were talking, 547 00:25:39,810 --> 00:25:42,180 I'm like, "Oh, that's exactly what they were using." 548 00:25:42,180 --> 00:25:45,753 But digital engineering is not a panacea, 549 00:25:46,650 --> 00:25:50,880 or it hasn't been fully used to its potential 550 00:25:50,880 --> 00:25:52,440 maybe is a better way of doing it, 551 00:25:52,440 --> 00:25:53,829 'cause there are short... 552 00:25:53,829 --> 00:25:55,180 'Cause you mentioned, 553 00:25:55,180 --> 00:25:58,800 you're not able to maybe digitally replicate 554 00:25:58,800 --> 00:26:01,830 every aspect of a design. 555 00:26:01,830 --> 00:26:04,382 And I'm thinking particularly of, you know, 556 00:26:04,382 --> 00:26:07,590 you might be able to do an environment, 557 00:26:07,590 --> 00:26:09,900 but certain components within that environment might not, 558 00:26:09,900 --> 00:26:11,493 like the avionics on a jet, 559 00:26:12,600 --> 00:26:14,130 you might have more of a challenging time 560 00:26:14,130 --> 00:26:18,600 using digital engineering for that purpose. 561 00:26:18,600 --> 00:26:21,150 Could you talk a little bit about some of the limitations, 562 00:26:21,150 --> 00:26:23,220 and then also when you mentioned this, 563 00:26:23,220 --> 00:26:25,410 you said it's both an answer and a cop out. 564 00:26:25,410 --> 00:26:27,690 So how is it a bit of a cop out? 565 00:26:27,690 --> 00:26:30,420 But what are some of the limitations of digital engineering? 566 00:26:30,420 --> 00:26:32,790 - All right, let's unwrap that question there. So... 567 00:26:32,790 --> 00:26:35,820 - Yeah, I like to ask very complex questions 568 00:26:35,820 --> 00:26:38,550 that give you wide range to answer any way you want. 569 00:26:38,550 --> 00:26:39,383 - It's great. 570 00:26:39,383 --> 00:26:42,270 So why was digital engineering so successful 571 00:26:42,270 --> 00:26:45,210 with B-21 and the next generation air balance? 572 00:26:45,210 --> 00:26:46,890 Well, think about it, right? 573 00:26:46,890 --> 00:26:50,910 Aerodynamics codes, all the major airframe manufacturers 574 00:26:50,910 --> 00:26:52,200 have been perfecting 575 00:26:52,200 --> 00:26:56,100 for decades and decades aerodynamic codes 576 00:26:56,100 --> 00:26:59,112 so that when they put a CAD model 577 00:26:59,112 --> 00:27:02,970 into their digital wind tunnel, 578 00:27:02,970 --> 00:27:06,450 it actually very, very accurately represents 579 00:27:06,450 --> 00:27:09,060 what you're gonna see when you actually fly it. 580 00:27:09,060 --> 00:27:10,980 The same is true with stealth codes, 581 00:27:10,980 --> 00:27:15,980 decades of perfecting the digital MNS codes for stealth. 582 00:27:17,520 --> 00:27:22,520 So, and engines, you know, the major components of airframes 583 00:27:23,970 --> 00:27:28,950 have wonderful tools that have been vetted over decades. 584 00:27:28,950 --> 00:27:30,120 And guess what? 585 00:27:30,120 --> 00:27:33,300 When they utilize those tools, make predictions, 586 00:27:33,300 --> 00:27:35,280 those predictions are very accurate. 587 00:27:35,280 --> 00:27:37,561 As you point out, though, and this is very important, 588 00:27:37,561 --> 00:27:39,570 there's a whole lot of stuff 589 00:27:39,570 --> 00:27:43,860 that gets stuffed inside those airframes, 590 00:27:43,860 --> 00:27:48,030 call 'em avionics, radar is all communication systems, 591 00:27:48,030 --> 00:27:50,820 and then there's antennas stuck on there. 592 00:27:50,820 --> 00:27:52,278 That is the Wild West. 593 00:27:52,278 --> 00:27:56,197 There is no uniform level of quality 594 00:27:56,197 --> 00:28:00,060 in the digital engineering tools for avionics, for example. 595 00:28:00,060 --> 00:28:02,022 There are certainly pieces, 596 00:28:02,022 --> 00:28:04,542 and some companies are better than others, 597 00:28:04,542 --> 00:28:09,542 but it's nowhere near the level of maturity and uniformity 598 00:28:11,354 --> 00:28:14,160 as there is for, you know, aerodynamics 599 00:28:14,160 --> 00:28:15,720 and things of that sort. 600 00:28:15,720 --> 00:28:17,130 So it's a very important point. 601 00:28:17,130 --> 00:28:20,370 So that's why it's sort of an answer and a cop out, right? 602 00:28:20,370 --> 00:28:23,850 If you have digital engineering tools that are accurate, 603 00:28:23,850 --> 00:28:25,380 it's an answer. 604 00:28:25,380 --> 00:28:28,470 If you don't have them, it's a cop out. (chuckles) 605 00:28:28,470 --> 00:28:31,470 So, you know, I know that's the first part of your question. 606 00:28:31,470 --> 00:28:33,030 I think I forgot the second part of your question. 607 00:28:33,030 --> 00:28:34,590 - Well, I mean, I think you actually did answer 608 00:28:34,590 --> 00:28:35,460 both parts of the question, 609 00:28:35,460 --> 00:28:36,738 'cause I was talking about, 610 00:28:36,738 --> 00:28:37,860 you know, why it would be a cop out. 611 00:28:37,860 --> 00:28:42,860 But, you know, with regard to the challenges of avionics, 612 00:28:43,380 --> 00:28:46,590 do we embrace that shortfall, 613 00:28:46,590 --> 00:28:50,340 that challenge of digital engineering 614 00:28:50,340 --> 00:28:51,180 when it comes to avionics? 615 00:28:51,180 --> 00:28:52,410 Do we really truly understand 616 00:28:52,410 --> 00:28:56,910 that when we embark on the overall design 617 00:28:56,910 --> 00:29:00,720 of a system, an aircraft or what have you, 618 00:29:00,720 --> 00:29:04,590 it would seem to me that, you know, in today's fighting age 619 00:29:04,590 --> 00:29:09,590 where spectrum dominance is paramount to success, 620 00:29:11,220 --> 00:29:13,650 avionics, what you have on whatever system, 621 00:29:13,650 --> 00:29:16,093 whether it be an antenna, a radar, a jammer, 622 00:29:16,093 --> 00:29:18,720 that's gotta be your starting point 623 00:29:18,720 --> 00:29:23,160 in terms of your final system. 624 00:29:23,160 --> 00:29:27,780 Is it a matter of simply needing to spend a lot more time 625 00:29:27,780 --> 00:29:30,691 maybe adapting digital engineering to this field, 626 00:29:30,691 --> 00:29:32,970 or is there something that we can do 627 00:29:32,970 --> 00:29:36,634 to just kind of make progress in this area? 628 00:29:36,634 --> 00:29:39,720 - I would like what you just said etched in stone somewhere, 629 00:29:39,720 --> 00:29:41,670 because that's it. 630 00:29:41,670 --> 00:29:43,516 We have to embrace digital engineering 631 00:29:43,516 --> 00:29:46,440 in the electronic warfare world. 632 00:29:46,440 --> 00:29:48,240 There's no other answer. 633 00:29:48,240 --> 00:29:52,721 We're never going to be able to go to a test range 634 00:29:52,721 --> 00:29:57,390 and faithfully replicate what we would see 635 00:29:57,390 --> 00:30:01,721 in a true, especially peer-on-peer engagement, 636 00:30:01,721 --> 00:30:04,440 without lots and lots of help, 637 00:30:04,440 --> 00:30:08,010 which means a lot of synthetic signals 638 00:30:08,010 --> 00:30:11,790 injected into the tests, synthetic targets. 639 00:30:11,790 --> 00:30:14,880 Some people call that live over sim, 640 00:30:14,880 --> 00:30:19,880 so accommodation simulation and live, you know, effects. 641 00:30:19,885 --> 00:30:23,520 But we have to, and by the way, I think, you know, 642 00:30:23,520 --> 00:30:25,650 this kind of went unnoticed by a lot of people, 643 00:30:25,650 --> 00:30:28,716 but OSD came out in December, I think it was, 644 00:30:28,716 --> 00:30:32,880 and mandated that all major defense programs 645 00:30:32,880 --> 00:30:36,060 must employ digital engineering unless they get a waiver. 646 00:30:36,060 --> 00:30:40,230 And I thought, "Wow, I mean, that's really quite something." 647 00:30:40,230 --> 00:30:44,139 Now you just pointed out, even on the B-21, 648 00:30:44,139 --> 00:30:46,860 they can point to all the digital engineering 649 00:30:46,860 --> 00:30:49,380 for the avionics and what have you and the stealth, 650 00:30:49,380 --> 00:30:54,300 but they can't do that for all the avionics inside. 651 00:30:54,300 --> 00:30:57,180 So, but the DOD wants to go that way. 652 00:30:57,180 --> 00:30:59,850 I think you know, there's a digital engineering czar 653 00:30:59,850 --> 00:31:02,640 out of the office of Secretary of Defense. 654 00:31:02,640 --> 00:31:06,930 I think when people call it digital engineering, 655 00:31:06,930 --> 00:31:10,260 some people call it model-based systems engineering, 656 00:31:10,260 --> 00:31:13,083 MBSC is basically a synonym. 657 00:31:14,040 --> 00:31:15,423 We don't have a choice. 658 00:31:16,590 --> 00:31:18,720 And let me just add one more wrinkle to it, 659 00:31:18,720 --> 00:31:22,037 which is, well, wait a second, 660 00:31:22,037 --> 00:31:24,510 who's creating the scenarios 661 00:31:24,510 --> 00:31:26,970 that are used for training, right? 662 00:31:26,970 --> 00:31:31,110 If I said before that cognitive systems were invented 663 00:31:31,110 --> 00:31:34,969 because we don't know exactly what we're gonna be facing, 664 00:31:34,969 --> 00:31:37,200 and yet you're saying, 665 00:31:37,200 --> 00:31:40,530 but humans are creating the training data, 666 00:31:40,530 --> 00:31:42,153 then there's a built-in flaw. 667 00:31:43,050 --> 00:31:45,783 And the way that I've seen to get around that, 668 00:31:46,992 --> 00:31:47,880 and we're actually implementing it 669 00:31:47,880 --> 00:31:51,693 on some of our research here that we're doing for AFRL, 670 00:31:52,650 --> 00:31:57,650 is let's let AI help create scenarios. 671 00:31:59,760 --> 00:32:01,320 And why is that so powerful? 672 00:32:01,320 --> 00:32:04,680 Well, DeepMind from Google is the best example, right? 673 00:32:04,680 --> 00:32:07,410 The original chess playing DeepMind 674 00:32:07,410 --> 00:32:10,350 was trained by subject matter experts, 675 00:32:10,350 --> 00:32:12,960 just the way we would do with digital engineering tools 676 00:32:12,960 --> 00:32:15,217 using humans to create the training data. 677 00:32:15,217 --> 00:32:18,386 But when they let it play on its own, 678 00:32:18,386 --> 00:32:22,350 they just created a training environment, not data, 679 00:32:22,350 --> 00:32:24,138 and let it, by trial and error, 680 00:32:24,138 --> 00:32:29,138 making dumb move after dumb move, eventually it learned. 681 00:32:29,700 --> 00:32:32,559 And guess what? There's no human that can beat it now. 682 00:32:32,559 --> 00:32:37,559 So using AI to help with the training portion 683 00:32:39,090 --> 00:32:41,853 of another AI system is ironic, 684 00:32:42,780 --> 00:32:44,790 but it seems to be the answer 685 00:32:44,790 --> 00:32:48,903 to, well, how do you prepare for the unknown unknowns? 686 00:32:50,970 --> 00:32:55,380 - Well, if you can use AI to train another AI 687 00:32:55,380 --> 00:33:00,033 to design an environment to train against, 688 00:33:00,990 --> 00:33:04,260 can you use AI through digital engineering 689 00:33:04,260 --> 00:33:09,260 to really almost design your next system, 690 00:33:10,950 --> 00:33:14,020 even though that's not what, you know, like, 691 00:33:14,020 --> 00:33:18,330 it can anticipate kind of here's what you're gonna need, 692 00:33:18,330 --> 00:33:20,430 here's a design of what you're going to need, 693 00:33:20,430 --> 00:33:24,900 not based on any sort of real human modeling 694 00:33:24,900 --> 00:33:27,420 or human outline. 695 00:33:27,420 --> 00:33:32,420 It's just let AI run over and over again, 696 00:33:32,610 --> 00:33:34,170 all the mistakes made in the past, 697 00:33:34,170 --> 00:33:35,640 all the successes made in the past. 698 00:33:35,640 --> 00:33:38,310 Here's exactly how your next system 699 00:33:38,310 --> 00:33:40,350 is gonna have to look in this threat environment 700 00:33:40,350 --> 00:33:41,183 that we're training against, 701 00:33:41,183 --> 00:33:44,250 and almost take it, in some ways remove us a step further 702 00:33:44,250 --> 00:33:47,910 from the design of the next generation systems. 703 00:33:47,910 --> 00:33:48,743 So is that already being done? 704 00:33:48,743 --> 00:33:50,790 Is that how it's so fast today 705 00:33:50,790 --> 00:33:52,980 with the next generation air dominance? 706 00:33:52,980 --> 00:33:55,500 - So once again, I'm gonna give you a mealy answer. 707 00:33:55,500 --> 00:33:56,402 Yes and no. 708 00:33:56,402 --> 00:34:01,402 So yes, we need to let AI, I'll say run amok 709 00:34:03,060 --> 00:34:06,183 and come up with all these scenarios on its own. 710 00:34:07,710 --> 00:34:08,847 A good example might be 711 00:34:08,847 --> 00:34:13,830 instead of coming up with the next generation air dominance, 712 00:34:13,830 --> 00:34:16,174 why don't we just send it a million stupid drones 713 00:34:16,174 --> 00:34:19,200 and just overwhelm our adversary 714 00:34:19,200 --> 00:34:22,180 with dirt cheap millions of drones, right? 715 00:34:22,180 --> 00:34:24,120 That would be something that, 716 00:34:24,120 --> 00:34:26,640 I mean, an AI system wouldn't be constrained, right? 717 00:34:26,640 --> 00:34:28,427 It would just come up with that answer. 718 00:34:28,427 --> 00:34:31,590 So, but here's what I will say about that. 719 00:34:31,590 --> 00:34:35,460 The good news is while we can let AI run amok 720 00:34:35,460 --> 00:34:38,610 and come up with scenarios that we wouldn't have thought of, 721 00:34:38,610 --> 00:34:42,622 we can then look at those scenarios, and we can say, 722 00:34:42,622 --> 00:34:46,590 "Is that really something we need to worry about?" 723 00:34:46,590 --> 00:34:48,120 A good term for that, 724 00:34:48,120 --> 00:34:50,620 a colleague of mine came up with a great, 725 00:34:50,620 --> 00:34:52,530 we call it black swan. 726 00:34:52,530 --> 00:34:56,880 You know, you might recall the whole point 727 00:34:56,880 --> 00:35:00,300 about black swan was there's no such thing as black swans 728 00:35:00,300 --> 00:35:02,643 until one day we actually found a black swan. 729 00:35:04,718 --> 00:35:09,540 And so we like to call this the black swan analysis stage, 730 00:35:09,540 --> 00:35:12,453 where you let AI come up with things 731 00:35:12,453 --> 00:35:14,067 that no human would've thought of, 732 00:35:14,067 --> 00:35:19,067 but are physically allowed and technologically allowed, 733 00:35:19,135 --> 00:35:22,443 and then you look at that. 734 00:35:23,520 --> 00:35:27,090 So now is where humans come back into play. 735 00:35:27,090 --> 00:35:29,400 Because look, we don't have unlimited resources, 736 00:35:29,400 --> 00:35:30,780 we don't have unlimited time. 737 00:35:30,780 --> 00:35:32,670 We can't spend trillions of dollars 738 00:35:32,670 --> 00:35:34,710 on a single aircraft, for example, 739 00:35:34,710 --> 00:35:39,160 so a subject matter has to look at these black swan events 740 00:35:40,140 --> 00:35:43,830 and decide whether we really care about that. 741 00:35:43,830 --> 00:35:48,830 And like I said, launching a million dumb cheap UAVs 742 00:35:51,465 --> 00:35:53,622 is certainly technically possible, 743 00:35:53,622 --> 00:35:56,130 but are we really gonna, 744 00:35:56,130 --> 00:35:58,860 is that how we're gonna base our whole defense posture 745 00:35:58,860 --> 00:35:59,940 is on something like that? 746 00:35:59,940 --> 00:36:03,600 So it's a yes and no answer, right? 747 00:36:03,600 --> 00:36:07,620 We're not letting go of the reins completely, 748 00:36:07,620 --> 00:36:09,371 only we're letting go of the reins 749 00:36:09,371 --> 00:36:12,873 when we're asking AI to use its imagination 750 00:36:12,873 --> 00:36:14,670 to come up with things, 751 00:36:14,670 --> 00:36:16,368 but we pull back on the reins 752 00:36:16,368 --> 00:36:18,840 when we see what they've come up with 753 00:36:18,840 --> 00:36:20,070 and decide whether or not 754 00:36:20,070 --> 00:36:21,780 that's something we should care about. 755 00:36:21,780 --> 00:36:26,780 - So we're talking about how many things we can do with AI. 756 00:36:28,230 --> 00:36:31,020 I wanna talk a little bit more, kind of take a step back, 757 00:36:31,020 --> 00:36:36,020 and continue talking a little bit about how AI works. 758 00:36:36,180 --> 00:36:40,174 And you had a slide in your webinar presentation 759 00:36:40,174 --> 00:36:42,960 that we were talking about the relationship with AI, 760 00:36:42,960 --> 00:36:45,660 and there's an aspect to AI 761 00:36:45,660 --> 00:36:49,830 that's using neuromorphic computing and neuromorphic chips, 762 00:36:49,830 --> 00:36:51,826 and we were talking about this. 763 00:36:51,826 --> 00:36:53,430 This concept just blew my mind, 764 00:36:53,430 --> 00:36:55,410 because I really never heard the term before. 765 00:36:55,410 --> 00:36:56,760 So I wanted to kind of, 766 00:36:56,760 --> 00:36:59,993 I wanna ask you to talk a little bit about this. 767 00:36:59,993 --> 00:37:02,432 What is this piece of the puzzle, 768 00:37:02,432 --> 00:37:06,210 and what does it it hold in terms of the future 769 00:37:06,210 --> 00:37:08,940 for artificial intelligence, 770 00:37:08,940 --> 00:37:11,793 and then feeding into cognitive radar and EW? 771 00:37:12,630 --> 00:37:14,880 - So cognitive radar, EW, 772 00:37:14,880 --> 00:37:19,019 live and die by embedded systems, right? 773 00:37:19,019 --> 00:37:22,920 They don't have the luxury of living in a laboratory 774 00:37:22,920 --> 00:37:25,080 with thousands of acres of computers, right? 775 00:37:25,080 --> 00:37:29,627 They have to take all their resources on a plane or at UAB 776 00:37:29,627 --> 00:37:33,180 or whatever platform and go into battle. 777 00:37:33,180 --> 00:37:37,320 And so to really leverage the power of AI, 778 00:37:37,320 --> 00:37:38,450 you need to implement them 779 00:37:38,450 --> 00:37:41,820 on efficient embedded computing systems. 780 00:37:41,820 --> 00:37:45,633 Right now, that means FPGAs, GPUs, 781 00:37:46,755 --> 00:37:49,950 and those things are, when all is said and done, 782 00:37:49,950 --> 00:37:52,116 you know, all the peripherals required, 783 00:37:52,116 --> 00:37:55,080 the ruggedization, the MIL-SPEC, 784 00:37:55,080 --> 00:37:57,303 you're talking kilograms and kilowatts. 785 00:37:59,743 --> 00:38:01,353 And as I pointed out, 786 00:38:02,610 --> 00:38:07,020 there is a rather quiet revolutionary part to AI 787 00:38:07,020 --> 00:38:09,298 that's perhaps even bigger 788 00:38:09,298 --> 00:38:12,360 than all the hullabaloo about ChatGPT, 789 00:38:12,360 --> 00:38:14,523 and that's neuromorphic chips. 790 00:38:15,810 --> 00:38:18,484 So neuromorphic chips don't implement 791 00:38:18,484 --> 00:38:22,743 traditional digital flip-flop circuits, things like that. 792 00:38:23,730 --> 00:38:26,594 Essentially they actually, in silicon, 793 00:38:26,594 --> 00:38:31,084 create neurons with interconnects. 794 00:38:31,084 --> 00:38:33,510 And the whole point of a neural network 795 00:38:33,510 --> 00:38:38,040 is the weighting that goes onto those interconnects 796 00:38:38,040 --> 00:38:39,800 from layer to layer. 797 00:38:39,800 --> 00:38:42,900 And the interesting thing about that 798 00:38:42,900 --> 00:38:47,190 is you've got companies like BrainChip in Australia, right, 799 00:38:47,190 --> 00:38:50,130 that is not by any stretch 800 00:38:50,130 --> 00:38:53,291 using the most sophisticated foundry 801 00:38:53,291 --> 00:38:55,860 to achieve ridiculous line lists 802 00:38:55,860 --> 00:38:59,760 like conventional FPGAs and GPUs do. 803 00:38:59,760 --> 00:39:02,850 Instead it's just a different architecture. 804 00:39:02,850 --> 00:39:04,680 But why is that such a big deal? 805 00:39:04,680 --> 00:39:09,271 Well, in the case of BrainChip as well as Intel and IBM, 806 00:39:09,271 --> 00:39:12,420 these chips can be the size of a postage stamp. 807 00:39:12,420 --> 00:39:14,970 And because they're implementing 808 00:39:14,970 --> 00:39:19,080 what are called spiking neural networks, or SNNs, 809 00:39:19,080 --> 00:39:22,920 they only draw power when there's a change of state, 810 00:39:22,920 --> 00:39:24,990 and that's a very short amount of time, 811 00:39:24,990 --> 00:39:26,550 and it's relatively low-power. 812 00:39:26,550 --> 00:39:28,170 So at the end of the day, 813 00:39:28,170 --> 00:39:30,441 you have something the size of a postage stamp 814 00:39:30,441 --> 00:39:35,100 that's implementing a very, very sophisticated 815 00:39:35,100 --> 00:39:37,570 convolutional neural network solution 816 00:39:39,150 --> 00:39:41,760 with grams and milliwatts 817 00:39:41,760 --> 00:39:45,570 as opposed to kilograms and kilowatts. 818 00:39:45,570 --> 00:39:48,780 And so to me, this is the revolution. 819 00:39:48,780 --> 00:39:51,750 This is dawning. This is the thing that changes everything. 820 00:39:51,750 --> 00:39:55,023 So now you see this little UAV coming in, 821 00:39:55,890 --> 00:39:59,790 and you don't think for a second that it could do, you know, 822 00:39:59,790 --> 00:40:02,400 the most sophisticated electronic warfare functions, 823 00:40:02,400 --> 00:40:03,330 for example. 824 00:40:03,330 --> 00:40:07,751 Pulse sorting, feature identification, geolocation, 825 00:40:07,751 --> 00:40:10,800 all these things that require, 826 00:40:10,800 --> 00:40:12,660 you know, thousands of lines of code 827 00:40:12,660 --> 00:40:15,690 and lots of high-speed embedded computing, 828 00:40:15,690 --> 00:40:18,690 all of a sudden it's done on a postage stamp. 829 00:40:18,690 --> 00:40:19,980 That's the crazy thing. 830 00:40:19,980 --> 00:40:23,202 And by the way, in my research we've done it. 831 00:40:23,202 --> 00:40:26,037 we've implemented pulse, the interleaving, 832 00:40:26,037 --> 00:40:28,773 we've implemented, you know, ATR, 833 00:40:29,623 --> 00:40:33,180 specifically on the BrainChip from Australia, by the way. 834 00:40:33,180 --> 00:40:34,533 So really quite amazing. 835 00:40:36,240 --> 00:40:38,640 - So where is this technology? 836 00:40:38,640 --> 00:40:39,900 You said we've already done it. 837 00:40:39,900 --> 00:40:42,510 We have a pretty good understanding of what it can do. 838 00:40:42,510 --> 00:40:45,060 And like you mentioned, you know, a scenario 839 00:40:45,060 --> 00:40:50,060 where whether it's a UAV or whatever system, 840 00:40:50,130 --> 00:40:52,830 I mean, something the size of a postage stamp, 841 00:40:52,830 --> 00:40:55,590 it completely changes size, weight, power, 842 00:40:55,590 --> 00:40:56,922 all those considerations, 843 00:40:56,922 --> 00:41:00,128 and makes almost anything a potential host 844 00:41:00,128 --> 00:41:04,050 for that capability. - Yeah. 845 00:41:04,050 --> 00:41:07,462 - What are some of the next steps in this, 846 00:41:07,462 --> 00:41:12,462 call it a revolution or rapid evolution of technology? 847 00:41:13,110 --> 00:41:15,723 I mean, because we obviously, you know, 848 00:41:15,723 --> 00:41:18,000 a couple years ago there was a CHIPS Act, you know, 849 00:41:18,000 --> 00:41:19,502 trying to make sure that we, 850 00:41:19,502 --> 00:41:24,502 in the development of a domestic chip production capability, 851 00:41:26,673 --> 00:41:28,320 Congress passed a CHIPS Act 852 00:41:28,320 --> 00:41:31,962 to kind of help spur on domestic foundries, 853 00:41:31,962 --> 00:41:34,980 domestic capability to produce chips. 854 00:41:34,980 --> 00:41:38,245 And does this kind of fall into kind of the... 855 00:41:38,245 --> 00:41:41,550 Is this benefiting from that type of activity? 856 00:41:41,550 --> 00:41:44,070 Is this part of the development 857 00:41:44,070 --> 00:41:45,960 that's happened through the CHIPS Act? 858 00:41:45,960 --> 00:41:47,820 Is there something more that we need to be doing 859 00:41:47,820 --> 00:41:49,710 to spur on this innovation? 860 00:41:49,710 --> 00:41:53,700 - Well, the CHIPS Act is a good thing domestically speaking. 861 00:41:53,700 --> 00:41:54,950 And by the way, part of the CHIPS Act, 862 00:41:54,950 --> 00:41:57,860 it is focused on neuromorphic chips, by the way, 863 00:41:57,860 --> 00:41:59,550 so that's good to know. 864 00:41:59,550 --> 00:42:04,550 However, the real culprit is the age-old valley of death, 865 00:42:05,490 --> 00:42:07,560 bridging the valley of death. 866 00:42:07,560 --> 00:42:10,170 And by the way, I spent seven years at DARPA, 867 00:42:10,170 --> 00:42:13,110 and even at DARPA with the funds I had available to me, 868 00:42:13,110 --> 00:42:17,489 bridging the gap between S&T and Programs of Record 869 00:42:17,489 --> 00:42:22,489 is still a herculean maze of biblical proportions. 870 00:42:24,390 --> 00:42:29,390 And so while you'll hear lots of nice-sounding words 871 00:42:29,400 --> 00:42:31,530 coming out of OSD and other places, 872 00:42:31,530 --> 00:42:33,600 saying, you know, "We gotta move things along. 873 00:42:33,600 --> 00:42:36,210 We gotta spur small business. We gotta..." 874 00:42:36,210 --> 00:42:37,973 it's all S&T funding. 875 00:42:37,973 --> 00:42:42,870 There still is an extraordinary impediment 876 00:42:42,870 --> 00:42:45,693 to getting new technologies into Programs of Record. 877 00:42:46,755 --> 00:42:49,740 And I, you know, I'm not the only one saying that, 878 00:42:49,740 --> 00:42:51,190 so don't take my word for it. 879 00:42:52,920 --> 00:42:57,180 I can tell you lots of horror stories, and I've done it. 880 00:42:57,180 --> 00:43:00,270 I was successful while at DARPA. 881 00:43:00,270 --> 00:43:03,300 So my stuff is on the F-35 and F-22, for example, 882 00:43:03,300 --> 00:43:04,560 and other classified systems. 883 00:43:04,560 --> 00:43:07,470 I mean, I know what it takes to get it done. 884 00:43:07,470 --> 00:43:11,460 Unfortunately, though there's a lot of lip service 885 00:43:11,460 --> 00:43:13,110 about overcoming that barrier, 886 00:43:13,110 --> 00:43:15,827 it still has changed very little 887 00:43:15,827 --> 00:43:19,860 in the 20 years since I've been successful at DARPA 888 00:43:19,860 --> 00:43:20,693 in transitioning. 889 00:43:20,693 --> 00:43:24,780 So I'm sorry, but that's biggest impediment. 890 00:43:24,780 --> 00:43:26,580 And I know it's not a technical thing, 891 00:43:26,580 --> 00:43:29,070 and I know there's lots of- 892 00:43:29,070 --> 00:43:31,380 - But here's what concerns me about that, 893 00:43:31,380 --> 00:43:32,940 is, you know, the valley of death, 894 00:43:32,940 --> 00:43:35,730 I mean, that's been in our terminology, 895 00:43:35,730 --> 00:43:37,680 in our lexicon for decades, 896 00:43:37,680 --> 00:43:39,280 like you say, going way back, you know, 897 00:43:39,280 --> 00:43:42,330 even before we even under, you know, 898 00:43:42,330 --> 00:43:43,500 back in the nineties and eighties 899 00:43:43,500 --> 00:43:46,681 when the technology, while advanced at the time, 900 00:43:46,681 --> 00:43:49,362 pales in comparison to what we can do today, 901 00:43:49,362 --> 00:43:52,620 the process hasn't changed. 902 00:43:52,620 --> 00:43:56,160 And so like if we had a valley of death back then, 903 00:43:56,160 --> 00:43:58,590 how are we ever going to bridge it today 904 00:43:58,590 --> 00:44:00,660 with as fast as technology is moving, 905 00:44:00,660 --> 00:44:04,653 as fast as the solutions we need to prototype and field. 906 00:44:05,730 --> 00:44:07,283 I mean, you mentioned it's herculean. 907 00:44:07,283 --> 00:44:11,010 I mean, it's almost beyond that it seems, 908 00:44:11,010 --> 00:44:13,495 because our system hasn't really changed that much 909 00:44:13,495 --> 00:44:16,085 over the past 20, 30 years. 910 00:44:16,085 --> 00:44:19,620 - Yeah, so maybe it's ironic, I don't know the right word, 911 00:44:19,620 --> 00:44:24,620 but on the S&T side, OSD, the service labs, 912 00:44:25,411 --> 00:44:29,222 you know, I would say that they're pretty forward-leaning 913 00:44:29,222 --> 00:44:31,230 and they're making good investments. 914 00:44:31,230 --> 00:44:34,440 The problem is getting into a Program of Record 915 00:44:34,440 --> 00:44:36,154 is where the rubber hits the road, 916 00:44:36,154 --> 00:44:38,790 and where things get fielded. 917 00:44:38,790 --> 00:44:41,850 And so you look at the S&T budgets, 918 00:44:41,850 --> 00:44:44,010 you look at the number of small businesses 919 00:44:44,010 --> 00:44:46,551 getting DOD S&T funds, 920 00:44:46,551 --> 00:44:49,260 and you could almost say, "Well, they're a success," right? 921 00:44:49,260 --> 00:44:50,790 I mean, we're giving small businesses, 922 00:44:50,790 --> 00:44:51,930 they're coming up with great things. 923 00:44:51,930 --> 00:44:54,720 But then look at how much of that 924 00:44:54,720 --> 00:44:56,610 actually ends up in a Program of Record. 925 00:44:56,610 --> 00:44:58,380 And let me just be clear. 926 00:44:58,380 --> 00:45:00,270 I don't blame the Programs of Record, 927 00:45:00,270 --> 00:45:02,662 because the game is stacked against them. 928 00:45:02,662 --> 00:45:07,470 They, very often, especially if it's newer technology, 929 00:45:07,470 --> 00:45:09,240 they are having lots of problems 930 00:45:09,240 --> 00:45:13,081 with getting the baseline system fielded. 931 00:45:13,081 --> 00:45:17,805 There's cost overruns, there's scheduling issues, 932 00:45:17,805 --> 00:45:22,805 and so they're already with 2.95 strikes against them, 933 00:45:24,510 --> 00:45:25,470 and now all of a sudden 934 00:45:25,470 --> 00:45:29,335 you want to on-ramp and entirely new capability 935 00:45:29,335 --> 00:45:32,380 when they're already behind the eight ball. 936 00:45:32,380 --> 00:45:34,440 That's just impossible, 937 00:45:34,440 --> 00:45:38,640 unless the whole culture of Programs of Record changes 938 00:45:38,640 --> 00:45:40,230 where, for example, you structure it 939 00:45:40,230 --> 00:45:44,490 so that every year you have to answer 940 00:45:44,490 --> 00:45:46,410 how are you dealing with obsolescence? 941 00:45:46,410 --> 00:45:49,890 How are you keeping up? Where are the on-ramps? 942 00:45:49,890 --> 00:45:53,370 How successful were you with these on-ramps, these upgrades, 943 00:45:53,370 --> 00:45:55,230 all of these things? 944 00:45:55,230 --> 00:45:57,343 Because until you fix that, 945 00:45:57,343 --> 00:46:00,390 I don't care how much money you spend on S&T, 946 00:46:00,390 --> 00:46:02,370 you're not gonna get fielded. 947 00:46:02,370 --> 00:46:04,140 - From a technology standpoint, 948 00:46:04,140 --> 00:46:06,510 let's just, you know, assume for a second 949 00:46:06,510 --> 00:46:08,160 that we make some progress 950 00:46:08,160 --> 00:46:11,100 in the policy side of the equation 951 00:46:11,100 --> 00:46:15,810 as it pertains to acquisition and the valley of death. 952 00:46:15,810 --> 00:46:17,730 From a technology perspective, 953 00:46:17,730 --> 00:46:19,650 you've been following this for 20 years. 954 00:46:19,650 --> 00:46:23,190 You know, where are some of the opportunities 955 00:46:23,190 --> 00:46:24,914 that are before you that you're like, 956 00:46:24,914 --> 00:46:27,608 this is the direction we need to go in, 957 00:46:27,608 --> 00:46:29,280 this is something that excites you 958 00:46:29,280 --> 00:46:31,380 or keeps you awake at night in a positive way, 959 00:46:31,380 --> 00:46:33,990 of like this is promising 960 00:46:33,990 --> 00:46:36,180 and it's gonna be your next pursuit? 961 00:46:36,180 --> 00:46:37,013 - Well, we definitely 962 00:46:37,013 --> 00:46:41,346 have to embrace cognitive systems for sure. 963 00:46:41,346 --> 00:46:43,770 I mean, I don't think there's anyone out there 964 00:46:43,770 --> 00:46:47,610 that would say we don't need that kind of flexibility 965 00:46:47,610 --> 00:46:49,710 and adaptability on the fly. 966 00:46:49,710 --> 00:46:52,373 Now, we can argue over just how much cognition we need 967 00:46:52,373 --> 00:46:54,600 and the flavors. 968 00:46:54,600 --> 00:46:58,260 That's fine. So there's that, right? 969 00:46:58,260 --> 00:46:59,687 Let's all just accept that. 970 00:46:59,687 --> 00:47:03,584 And then I think you touched on this earlier, 971 00:47:03,584 --> 00:47:08,343 you know, there's a big push across all the services 972 00:47:08,343 --> 00:47:10,770 on what's called the JSE, 973 00:47:10,770 --> 00:47:13,680 which is the Joint Simulation Environment, 974 00:47:13,680 --> 00:47:15,431 which is this grandiose vision 975 00:47:15,431 --> 00:47:20,001 for having multi-user, multiplayer, 976 00:47:20,001 --> 00:47:24,480 high fidelity training environments, synthetic environments, 977 00:47:24,480 --> 00:47:26,730 which, by the way, can include live over sim, 978 00:47:27,660 --> 00:47:31,620 so that our systems become much more realistic 979 00:47:31,620 --> 00:47:33,420 and reflective of what they're really gonna see 980 00:47:33,420 --> 00:47:35,670 when they get out into the real world. 981 00:47:35,670 --> 00:47:39,530 Again, I come back to lots of good things going on 982 00:47:39,530 --> 00:47:40,980 on the S&T side. 983 00:47:40,980 --> 00:47:43,620 You almost can't, you know, you really can't argue with it, 984 00:47:43,620 --> 00:47:47,430 but that transition to field its systems 985 00:47:47,430 --> 00:47:51,420 and Programs of Record is still very much broken, 986 00:47:51,420 --> 00:47:52,950 and that's just a fact. 987 00:47:52,950 --> 00:47:54,480 And it's not just me saying that. 988 00:47:54,480 --> 00:47:57,600 You can ask anyone who's in the business 989 00:47:57,600 --> 00:48:00,090 of trying to transition technology 990 00:48:00,090 --> 00:48:01,320 to the Department of Defense, 991 00:48:01,320 --> 00:48:02,910 and they'll tell you the same thing. 992 00:48:02,910 --> 00:48:06,976 So, you know, again, S&T community, 993 00:48:06,976 --> 00:48:09,630 doing a great job, I think, generally speaking, 994 00:48:09,630 --> 00:48:12,570 your DARPAs, your AFRLs, all of these, 995 00:48:12,570 --> 00:48:16,320 but that transition piece is just continuing. 996 00:48:16,320 --> 00:48:19,350 And by the way, do our adversaries have the same issues? 997 00:48:19,350 --> 00:48:22,173 Some do, some don't, you know? 998 00:48:23,520 --> 00:48:26,489 And this technology I'm talking about, neuromorphic chips, 999 00:48:26,489 --> 00:48:28,320 that's available to the world. 1000 00:48:28,320 --> 00:48:31,500 I mean, BrainChip is an Australian company. 1001 00:48:31,500 --> 00:48:34,260 There's no ITAR restrictions, so. 1002 00:48:34,260 --> 00:48:36,510 - Well, and also I think it speaks 1003 00:48:36,510 --> 00:48:40,080 to the multidisciplinary approach to technology today. 1004 00:48:40,080 --> 00:48:42,060 I mean, the neuromorphic chip, 1005 00:48:42,060 --> 00:48:44,070 I mean, it has military applications 1006 00:48:44,070 --> 00:48:45,150 you can obviously use it for, 1007 00:48:45,150 --> 00:48:46,740 but, I mean, you're gonna find this 1008 00:48:46,740 --> 00:48:50,340 in all various sectors of an economy and society 1009 00:48:50,340 --> 00:48:54,570 and what we use in everyday life, and so, you know- 1010 00:48:54,570 --> 00:48:56,640 - So Ken, let me just say that the neuromorphic chip 1011 00:48:56,640 --> 00:48:58,560 that BrainChips makes from Australia 1012 00:48:58,560 --> 00:49:00,660 had nothing to do with electronic warfare. 1013 00:49:00,660 --> 00:49:02,910 It's designed to do image processing. 1014 00:49:02,910 --> 00:49:06,008 So one of the things we had to overcome 1015 00:49:06,008 --> 00:49:09,960 was take our electronic warfare I/Q data, 1016 00:49:09,960 --> 00:49:12,870 in-phase and quadrature RF measurement data, 1017 00:49:12,870 --> 00:49:16,110 and put it into a format to make it look like an image 1018 00:49:16,110 --> 00:49:18,862 so that the BrainChip could actually digest it 1019 00:49:18,862 --> 00:49:20,400 and do something with it. 1020 00:49:20,400 --> 00:49:21,930 So you're absolutely right. 1021 00:49:21,930 --> 00:49:25,500 I mean, these chips are not being designed for us 1022 00:49:25,500 --> 00:49:27,480 in the electronic warfare community, 1023 00:49:27,480 --> 00:49:29,010 but they're so powerful 1024 00:49:29,010 --> 00:49:32,095 that we were still able to get it to work. 1025 00:49:32,095 --> 00:49:34,650 Imagine if they put a little effort 1026 00:49:34,650 --> 00:49:37,560 into tailoring it to our needs. 1027 00:49:37,560 --> 00:49:39,420 Then you have a revolution. 1028 00:49:39,420 --> 00:49:41,190 So, sorry to interrupt you there, but I just want... 1029 00:49:41,190 --> 00:49:43,382 You made a point and it's very valid, you know. 1030 00:49:43,382 --> 00:49:46,794 - It's valid. It's valid, it's important. 1031 00:49:46,794 --> 00:49:50,910 I mean, it goes to just the possibilities 1032 00:49:50,910 --> 00:49:52,200 that are out there. 1033 00:49:52,200 --> 00:49:54,840 - Well, and to amplify that point, 1034 00:49:54,840 --> 00:49:56,310 all the advanced capabilities 1035 00:49:56,310 --> 00:50:00,993 that we have in our RF systems, radar and EW, 1036 00:50:01,950 --> 00:50:05,329 most of that is driven by the wireless community, 1037 00:50:05,329 --> 00:50:08,310 the trillion-dollar wireless community 1038 00:50:08,310 --> 00:50:13,310 compared to a paltry radar and EW ecosystem. 1039 00:50:14,520 --> 00:50:18,682 So, you know, what's happening in the commercial world 1040 00:50:18,682 --> 00:50:22,260 is where, and leveraging, you know, 1041 00:50:22,260 --> 00:50:23,970 commercial off-the-shelf technology 1042 00:50:23,970 --> 00:50:27,870 is a gargantuan piece of staying up and keeping up, 1043 00:50:27,870 --> 00:50:31,260 and by the way, addressing obsolescence as well, right? 1044 00:50:31,260 --> 00:50:34,533 If you have a piece of proprietary hardware from the 1980s, 1045 00:50:35,730 --> 00:50:38,673 good luck, you know, with obsolescence, right? 1046 00:50:39,510 --> 00:50:42,000 - Well, that, and also hopefully, you know, 1047 00:50:42,000 --> 00:50:44,370 as we move down this path on standards 1048 00:50:44,370 --> 00:50:46,350 and open systems and so forth, 1049 00:50:46,350 --> 00:50:49,264 some of that will work its way in. 1050 00:50:49,264 --> 00:50:50,790 We can adapt some of that 1051 00:50:50,790 --> 00:50:54,750 so that as we struggle less with obsolescence in the future 1052 00:50:54,750 --> 00:50:55,860 than we do now. 1053 00:50:55,860 --> 00:50:58,200 - We hope. - Hopefully, yes. I mean- 1054 00:50:58,200 --> 00:50:59,375 - Again- - We'll see. 1055 00:50:59,375 --> 00:51:02,310 But, I mean, I would think that's the idea. 1056 00:51:02,310 --> 00:51:04,080 - I mean, look at the day-to-day pressures 1057 00:51:04,080 --> 00:51:06,870 that Programs of Record are under. 1058 00:51:06,870 --> 00:51:09,390 So I'm not gonna get into all kinds of details here, 1059 00:51:09,390 --> 00:51:10,710 but we had a capability 1060 00:51:10,710 --> 00:51:14,460 that was vetted by the program offices 1061 00:51:14,460 --> 00:51:17,010 and was developed under SBIRS, 1062 00:51:17,010 --> 00:51:20,100 and went all the way through to a Phase III SBIR. 1063 00:51:20,100 --> 00:51:22,110 We have flight-qualified software 1064 00:51:22,110 --> 00:51:25,950 to bring this much-needed capability to the war fighter. 1065 00:51:25,950 --> 00:51:27,333 This is all a true story. 1066 00:51:28,170 --> 00:51:32,010 And all of a sudden the program ran into scheduling 1067 00:51:32,010 --> 00:51:33,870 and budgetary constraints, 1068 00:51:33,870 --> 00:51:36,153 so they had a jettison the on-ramps, 1069 00:51:37,200 --> 00:51:39,293 and so a capability that was vetted 1070 00:51:39,293 --> 00:51:42,360 was a really important capability, 1071 00:51:42,360 --> 00:51:46,200 just got thrown to the curb because of the everyday problems 1072 00:51:46,200 --> 00:51:48,630 that Programs of Record run into, 1073 00:51:48,630 --> 00:51:50,877 and that's not how they get judged, right? 1074 00:51:50,877 --> 00:51:53,850 They're judged on getting that baseline system over... 1075 00:51:53,850 --> 00:51:58,710 Look, the F-35 was just recently declared operational, 1076 00:51:58,710 --> 00:51:59,643 what, a month ago? 1077 00:52:00,600 --> 00:52:01,800 You gotta be kiddin' me. 1078 00:52:01,800 --> 00:52:04,140 - Well, Joe, I think this is a good spot 1079 00:52:04,140 --> 00:52:06,180 to, I mean, I feel like if we keep talking 1080 00:52:06,180 --> 00:52:08,680 we can keep going in layer and layer and layer, 1081 00:52:08,680 --> 00:52:11,780 and I don't wanna put our listeners through that, 1082 00:52:11,780 --> 00:52:15,060 but I think a good consolation prize 1083 00:52:15,060 --> 00:52:18,030 is to have you back on the show in the future, 1084 00:52:18,030 --> 00:52:20,903 and we can go a little bit deeper into this, 1085 00:52:20,903 --> 00:52:24,240 but I do really appreciate you taking some time 1086 00:52:24,240 --> 00:52:25,073 to talk about this, 1087 00:52:25,073 --> 00:52:28,014 'cause this is a topic as of, you know, really 24 hours ago, 1088 00:52:28,014 --> 00:52:32,220 I realized how often I just use the word, 1089 00:52:32,220 --> 00:52:34,410 and I never really understood the depth of the definition 1090 00:52:34,410 --> 00:52:35,430 of what the words I was using, 1091 00:52:35,430 --> 00:52:37,350 so I really appreciate you coming on the show, 1092 00:52:37,350 --> 00:52:39,690 kind of helping me understand this better, 1093 00:52:39,690 --> 00:52:41,762 and hopefully our listeners as well. 1094 00:52:41,762 --> 00:52:42,600 - Thank you, Ken. 1095 00:52:42,600 --> 00:52:44,730 You had great questions, great interview. 1096 00:52:44,730 --> 00:52:48,956 And let me give a shout out to AOC. Great organization. 1097 00:52:48,956 --> 00:52:53,172 I'm personally, and my company's a big supporter of AOC 1098 00:52:53,172 --> 00:52:55,140 and what you guys are doing, 1099 00:52:55,140 --> 00:52:57,963 so you're part of the solution, not part of the problem. 1100 00:52:58,890 --> 00:53:00,273 - We appreciate that, 1101 00:53:00,273 --> 00:53:02,580 and, you know, appreciate all that you've done for us 1102 00:53:02,580 --> 00:53:06,300 in terms of helping us understand this really complex topic. 1103 00:53:06,300 --> 00:53:09,420 And really I do say this honestly, 1104 00:53:09,420 --> 00:53:11,412 I do hope to have you back on the show here, 1105 00:53:11,412 --> 00:53:16,290 and there's no shortage of topics of conversation for us, 1106 00:53:16,290 --> 00:53:18,300 so I appreciate you joining me. 1107 00:53:18,300 --> 00:53:19,301 - Thanks again, Ken. 1108 00:53:19,301 --> 00:53:22,320 - That will conclude this episode of "From the Crow's Nest." 1109 00:53:22,320 --> 00:53:24,510 I'd like to thank my guest, Dr. Joe Guerci, 1110 00:53:24,510 --> 00:53:26,280 for joining me for this discussion. 1111 00:53:26,280 --> 00:53:28,560 Also, don't forget to review, share, 1112 00:53:28,560 --> 00:53:30,090 and follow this podcast. 1113 00:53:30,090 --> 00:53:32,700 We always enjoy hearing from our listeners, 1114 00:53:32,700 --> 00:53:35,670 so please take a moment to let us know how we're doing. 1115 00:53:35,670 --> 00:53:38,084 That's it for today. Thanks for listening. 1116 00:53:38,084 --> 00:53:40,834 (exciting music) 1117 00:53:52,480 --> 00:53:56,147 (uptempo music) - Voxtopica.