1 00:00:01,459 --> 00:00:04,010 covering the week's top textbook 2 00:00:04,010 --> 00:00:04,020 covering the week's top textbook 3 00:00:04,020 --> 00:00:07,460 covering the week's top textbook slight Linux bias arm has announced the 4 00:00:07,460 --> 00:00:07,470 slight Linux bias arm has announced the 5 00:00:07,470 --> 00:00:11,870 slight Linux bias arm has announced the Evo's yu55 and the cortex m55 for edge 6 00:00:11,870 --> 00:00:11,880 Evo's yu55 and the cortex m55 for edge 7 00:00:11,880 --> 00:00:15,289 Evo's yu55 and the cortex m55 for edge devices bringing an AI neural network 8 00:00:15,289 --> 00:00:15,299 devices bringing an AI neural network 9 00:00:15,299 --> 00:00:17,950 devices bringing an AI neural network accelerator to futures car speakers 10 00:00:17,950 --> 00:00:17,960 accelerator to futures car speakers 11 00:00:17,960 --> 00:00:21,790 accelerator to futures car speakers lightbulbs fridges and other IOT devices 12 00:00:21,790 --> 00:00:21,800 lightbulbs fridges and other IOT devices 13 00:00:21,800 --> 00:00:24,950 lightbulbs fridges and other IOT devices the more powerful of the two known as 14 00:00:24,950 --> 00:00:24,960 the more powerful of the two known as 15 00:00:24,960 --> 00:00:28,390 the more powerful of the two known as the cortex m55 is a general 16 00:00:28,390 --> 00:00:28,400 the cortex m55 is a general 17 00:00:28,400 --> 00:00:31,460 the cortex m55 is a general microcontroller grade CPU blueprint well 18 00:00:31,460 --> 00:00:31,470 microcontroller grade CPU blueprint well 19 00:00:31,470 --> 00:00:34,400 microcontroller grade CPU blueprint well the other names ethos yu55 is 20 00:00:34,400 --> 00:00:34,410 the other names ethos yu55 is 21 00:00:34,410 --> 00:00:38,540 the other names ethos yu55 is essentially an AI accelerator the cortex 22 00:00:38,540 --> 00:00:38,550 essentially an AI accelerator the cortex 23 00:00:38,550 --> 00:00:41,960 essentially an AI accelerator the cortex m55 is based on arms helium technology 24 00:00:41,960 --> 00:00:41,970 m55 is based on arms helium technology 25 00:00:41,970 --> 00:00:45,290 m55 is based on arms helium technology the ethos u55 on the other hand is a 26 00:00:45,290 --> 00:00:45,300 the ethos u55 on the other hand is a 27 00:00:45,300 --> 00:00:47,930 the ethos u55 on the other hand is a novel architecture forearm and has been 28 00:00:47,930 --> 00:00:47,940 novel architecture forearm and has been 29 00:00:47,940 --> 00:00:50,990 novel architecture forearm and has been described as a micro neural processing 30 00:00:50,990 --> 00:00:51,000 described as a micro neural processing 31 00:00:51,000 --> 00:00:54,770 described as a micro neural processing unit or micro NPU for short both 32 00:00:54,770 --> 00:00:54,780 unit or micro NPU for short both 33 00:00:54,780 --> 00:00:57,890 unit or micro NPU for short both processors are available now to license 34 00:00:57,890 --> 00:00:57,900 processors are available now to license 35 00:00:57,900 --> 00:01:00,710 processors are available now to license and are intended to be used together the 36 00:01:00,710 --> 00:01:00,720 and are intended to be used together the 37 00:01:00,720 --> 00:01:04,700 and are intended to be used together the m55 running application code the yu55 38 00:01:04,700 --> 00:01:04,710 m55 running application code the yu55 39 00:01:04,710 --> 00:01:06,950 m55 running application code the yu55 doing all the neural network mathematics 40 00:01:06,950 --> 00:01:06,960 doing all the neural network mathematics 41 00:01:06,960 --> 00:01:09,919 doing all the neural network mathematics in fast Hardware the senior director of 42 00:01:09,919 --> 00:01:09,929 in fast Hardware the senior director of 43 00:01:09,929 --> 00:01:12,830 in fast Hardware the senior director of the IOT and embedded team at arms says 44 00:01:12,830 --> 00:01:12,840 the IOT and embedded team at arms says 45 00:01:12,840 --> 00:01:16,130 the IOT and embedded team at arms says quote the micro NPU cannot be used on 46 00:01:16,130 --> 00:01:16,140 quote the micro NPU cannot be used on 47 00:01:16,140 --> 00:01:18,230 quote the micro NPU cannot be used on its own it needs to be paired with a CPU 48 00:01:18,230 --> 00:01:18,240 its own it needs to be paired with a CPU 49 00:01:18,240 --> 00:01:21,649 its own it needs to be paired with a CPU like the cortex m55 together the system 50 00:01:21,649 --> 00:01:21,659 like the cortex m55 together the system 51 00:01:21,659 --> 00:01:24,590 like the cortex m55 together the system delivers 480 times the performance 52 00:01:24,590 --> 00:01:24,600 delivers 480 times the performance 53 00:01:24,600 --> 00:01:26,270 delivers 480 times the performance compared to previous cortex-m 54 00:01:26,270 --> 00:01:26,280 compared to previous cortex-m 55 00:01:26,280 --> 00:01:28,789 compared to previous cortex-m generations working on their own end 56 00:01:28,789 --> 00:01:28,799 generations working on their own end 57 00:01:28,799 --> 00:01:29,529 generations working on their own end quote 58 00:01:29,529 --> 00:01:29,539 quote 59 00:01:29,539 --> 00:01:32,480 quote we expect to see chips using these 60 00:01:32,480 --> 00:01:32,490 we expect to see chips using these 61 00:01:32,490 --> 00:01:34,929 we expect to see chips using these blueprints early next year 62 00:01:34,929 --> 00:01:34,939 blueprints early next year 63 00:01:34,939 --> 00:01:37,639 blueprints early next year awesome I can't believe this is real 64 00:01:37,639 --> 00:01:37,649 awesome I can't believe this is real 65 00:01:37,649 --> 00:01:39,950 awesome I can't believe this is real life that neural processing units are 66 00:01:39,950 --> 00:01:39,960 life that neural processing units are 67 00:01:39,960 --> 00:01:43,340 life that neural processing units are now I know heart of an SOC like this is 68 00:01:43,340 --> 00:01:43,350 now I know heart of an SOC like this is 69 00:01:43,350 --> 00:01:47,959 now I know heart of an SOC like this is so it's just signed by - my what exactly 70 00:01:47,959 --> 00:01:47,969 so it's just signed by - my what exactly 71 00:01:47,969 --> 00:01:50,840 so it's just signed by - my what exactly an eighties mind like I remember when 72 00:01:50,840 --> 00:01:50,850 an eighties mind like I remember when 73 00:01:50,850 --> 00:01:54,190 an eighties mind like I remember when which Star Trek was it where they use 74 00:01:54,190 --> 00:01:54,200 which Star Trek was it where they use 75 00:01:54,200 --> 00:01:58,969 which Star Trek was it where they use the neuro packs what are they called 76 00:01:58,969 --> 00:01:58,979 the neuro packs what are they called 77 00:01:58,979 --> 00:02:01,429 the neuro packs what are they called oh my gosh one of the Star Trek's was it 78 00:02:01,429 --> 00:02:01,439 oh my gosh one of the Star Trek's was it 79 00:02:01,439 --> 00:02:04,129 oh my gosh one of the Star Trek's was it enterprise maybe where there's like Oh 80 00:02:04,129 --> 00:02:04,139 enterprise maybe where there's like Oh 81 00:02:04,139 --> 00:02:07,399 enterprise maybe where there's like Oh to have like telepresence yeah within 82 00:02:07,399 --> 00:02:07,409 to have like telepresence yeah within 83 00:02:07,409 --> 00:02:09,109 to have like telepresence yeah within the the ship and like they would blow 84 00:02:09,109 --> 00:02:09,119 the the ship and like they would blow 85 00:02:09,119 --> 00:02:10,609 the the ship and like they would blow out and they'd have to replace them and 86 00:02:10,609 --> 00:02:10,619 out and they'd have to replace them and 87 00:02:10,619 --> 00:02:12,110 out and they'd have to replace them and there were like these gel pack things 88 00:02:12,110 --> 00:02:12,120 there were like these gel pack things 89 00:02:12,120 --> 00:02:15,650 there were like these gel pack things and I'm like that would never happen 90 00:02:15,650 --> 00:02:15,660 and I'm like that would never happen 91 00:02:15,660 --> 00:02:20,090 and I'm like that would never happen I think it was Star Trek I understand 92 00:02:20,090 --> 00:02:20,100 I think it was Star Trek I understand 93 00:02:20,100 --> 00:02:22,370 I think it was Star Trek I understand that but I remember watching that I 94 00:02:22,370 --> 00:02:22,380 that but I remember watching that I 95 00:02:22,380 --> 00:02:23,540 that but I remember watching that I think I Star Trek Enterprise and I was 96 00:02:23,540 --> 00:02:23,550 think I Star Trek Enterprise and I was 97 00:02:23,550 --> 00:02:25,570 think I Star Trek Enterprise and I was like there's no way we're gonna have 98 00:02:25,570 --> 00:02:25,580 like there's no way we're gonna have 99 00:02:25,580 --> 00:02:28,100 like there's no way we're gonna have that kind of advanced technology it's 100 00:02:28,100 --> 00:02:28,110 that kind of advanced technology it's 101 00:02:28,110 --> 00:02:30,080 that kind of advanced technology it's not possible this is like 15 20 years 102 00:02:30,080 --> 00:02:30,090 not possible this is like 15 20 years 103 00:02:30,090 --> 00:02:32,360 not possible this is like 15 20 years ago and now I'm watching stuff like this 104 00:02:32,360 --> 00:02:32,370 ago and now I'm watching stuff like this 105 00:02:32,370 --> 00:02:33,920 ago and now I'm watching stuff like this come it's like neuro networks I'm like 106 00:02:33,920 --> 00:02:33,930 come it's like neuro networks I'm like 107 00:02:33,930 --> 00:02:37,310 come it's like neuro networks I'm like oh my goodness every year there's a new 108 00:02:37,310 --> 00:02:37,320 oh my goodness every year there's a new 109 00:02:37,320 --> 00:02:38,750 oh my goodness every year there's a new possibility and I never would have 110 00:02:38,750 --> 00:02:38,760 possibility and I never would have 111 00:02:38,760 --> 00:02:42,320 possibility and I never would have expected this this is yeah like so we're 112 00:02:42,320 --> 00:02:42,330 expected this this is yeah like so we're 113 00:02:42,330 --> 00:02:44,600 expected this this is yeah like so we're talking like the next gen of yes AI 114 00:02:44,600 --> 00:02:44,610 talking like the next gen of yes AI 115 00:02:44,610 --> 00:02:47,750 talking like the next gen of yes AI which is so crazy that these little 116 00:02:47,750 --> 00:02:47,760 which is so crazy that these little 117 00:02:47,760 --> 00:02:51,410 which is so crazy that these little chips can now process teraflops worth of 118 00:02:51,410 --> 00:02:51,420 chips can now process teraflops worth of 119 00:02:51,420 --> 00:02:53,270 chips can now process teraflops worth of information like it's I don't know what 120 00:02:53,270 --> 00:02:53,280 information like it's I don't know what 121 00:02:53,280 --> 00:02:55,400 information like it's I don't know what it really is but it's like ridiculous 122 00:02:55,400 --> 00:02:55,410 it really is but it's like ridiculous 123 00:02:55,410 --> 00:02:58,760 it really is but it's like ridiculous amounts of data having to do with AI 124 00:02:58,760 --> 00:02:58,770 amounts of data having to do with AI 125 00:02:58,770 --> 00:03:01,700 amounts of data having to do with AI right so that so the next thing is that 126 00:03:01,700 --> 00:03:01,710 right so that so the next thing is that 127 00:03:01,710 --> 00:03:03,280 right so that so the next thing is that the you know coders will be able to make 128 00:03:03,280 --> 00:03:03,290 the you know coders will be able to make 129 00:03:03,290 --> 00:03:07,430 the you know coders will be able to make thinking devices so those smart speakers 130 00:03:07,430 --> 00:03:07,440 thinking devices so those smart speakers 131 00:03:07,440 --> 00:03:09,380 thinking devices so those smart speakers with this and with this technology will 132 00:03:09,380 --> 00:03:09,390 with this and with this technology will 133 00:03:09,390 --> 00:03:13,010 with this and with this technology will be smarter is the fact that there are 134 00:03:13,010 --> 00:03:13,020 be smarter is the fact that there are 135 00:03:13,020 --> 00:03:15,140 be smarter is the fact that there are two separate things like I'm surprised 136 00:03:15,140 --> 00:03:15,150 two separate things like I'm surprised 137 00:03:15,150 --> 00:03:17,000 two separate things like I'm surprised that well you have to have a CPU you 138 00:03:17,000 --> 00:03:17,010 that well you have to have a CPU you 139 00:03:17,010 --> 00:03:18,410 that well you have to have a CPU you have to have the ability to run 140 00:03:18,410 --> 00:03:18,420 have to have the ability to run 141 00:03:18,420 --> 00:03:20,810 have to have the ability to run traditional code yeah otherwise you 142 00:03:20,810 --> 00:03:20,820 traditional code yeah otherwise you 143 00:03:20,820 --> 00:03:22,220 traditional code yeah otherwise you wouldn't be able to like you wouldn't be 144 00:03:22,220 --> 00:03:22,230 wouldn't be able to like you wouldn't be 145 00:03:22,230 --> 00:03:23,960 wouldn't be able to like you wouldn't be able to as a human interact with it 146 00:03:23,960 --> 00:03:23,970 able to as a human interact with it 147 00:03:23,970 --> 00:03:26,870 able to as a human interact with it there has to be a layer for humans to be 148 00:03:26,870 --> 00:03:26,880 there has to be a layer for humans to be 149 00:03:26,880 --> 00:03:29,630 there has to be a layer for humans to be able to access the NPU yeah hmm 150 00:03:29,630 --> 00:03:29,640 able to access the NPU yeah hmm 151 00:03:29,640 --> 00:03:33,440 able to access the NPU yeah hmm otherwise it just blow our minds how 152 00:03:33,440 --> 00:03:33,450 otherwise it just blow our minds how 153 00:03:33,450 --> 00:03:35,060 otherwise it just blow our minds how quickly before this would be adapted for 154 00:03:35,060 --> 00:03:35,070 quickly before this would be adapted for 155 00:03:35,070 --> 00:03:40,220 quickly before this would be adapted for phones and technology in phones right 156 00:03:40,220 --> 00:03:40,230 phones and technology in phones right 157 00:03:40,230 --> 00:03:42,260 phones and technology in phones right but I mean at this level 158 00:03:42,260 --> 00:03:42,270 but I mean at this level 159 00:03:42,270 --> 00:03:44,480 but I mean at this level I don't know like because that could 160 00:03:44,480 --> 00:03:44,490 I don't know like because that could 161 00:03:44,490 --> 00:03:48,020 I don't know like because that could change mobile computing sure it's gonna 162 00:03:48,020 --> 00:03:48,030 change mobile computing sure it's gonna 163 00:03:48,030 --> 00:03:50,330 change mobile computing sure it's gonna change a lot of sauce that's the idea is 164 00:03:50,330 --> 00:03:50,340 change a lot of sauce that's the idea is 165 00:03:50,340 --> 00:03:53,060 change a lot of sauce that's the idea is that it's getting faster smaller yeah 166 00:03:53,060 --> 00:03:53,070 that it's getting faster smaller yeah 167 00:03:53,070 --> 00:03:55,790 that it's getting faster smaller yeah and more integrate that's the way 168 00:03:55,790 --> 00:03:55,800 and more integrate that's the way 169 00:03:55,800 --> 00:03:58,210 and more integrate that's the way technology goes and so we're seeing a 170 00:03:58,210 --> 00:03:58,220 technology goes and so we're seeing a 171 00:03:58,220 --> 00:04:02,650 technology goes and so we're seeing a massive shift in how data is processed 172 00:04:02,650 --> 00:04:02,660 massive shift in how data is processed 173 00:04:02,660 --> 00:04:04,430 massive shift in how data is processed because we're specifically looking 174 00:04:04,430 --> 00:04:04,440 because we're specifically looking 175 00:04:04,440 --> 00:04:07,250 because we're specifically looking toward AI and machine learning which are 176 00:04:07,250 --> 00:04:07,260 toward AI and machine learning which are 177 00:04:07,260 --> 00:04:09,440 toward AI and machine learning which are vez