1 00:00:12,960 --> 00:00:15,480 Speaker 1: For as long as humans have used computers, we have 2 00:00:15,560 --> 00:00:19,480 Speaker 1: sought to make them faster, smarter, and more capable. But 3 00:00:19,520 --> 00:00:22,440 Speaker 1: what happens when we harness the power of artificial intelligence 4 00:00:22,560 --> 00:00:25,639 Speaker 1: with new PC hardware and software? And what if every 5 00:00:25,640 --> 00:00:29,520 Speaker 1: computer could use AI technology to unleash new capabilities that 6 00:00:29,600 --> 00:00:33,080 Speaker 1: will benefit anyone and everyone who uses a computer. 7 00:00:34,479 --> 00:00:35,360 Speaker 2: What can I help you with? 8 00:00:36,440 --> 00:00:39,320 Speaker 1: This is not a distant dream. Each day it becomes 9 00:00:39,400 --> 00:00:43,080 Speaker 1: more and more of a reality. AI is no longer 10 00:00:43,120 --> 00:00:47,360 Speaker 1: exclusively running in the cloud. It's increasingly finding its way 11 00:00:47,400 --> 00:00:51,680 Speaker 1: into all computers, enabling business users to do more and 12 00:00:51,760 --> 00:00:56,320 Speaker 1: be more with PC technology from Intel. In this episode, 13 00:00:56,640 --> 00:01:00,720 Speaker 1: we'll be focusing on defining the AIPC and what represents 14 00:01:00,960 --> 00:01:05,000 Speaker 1: for the future of this transformative technology. We'll also explore 15 00:01:05,040 --> 00:01:07,280 Speaker 1: what the growth of AI running directly on the PC 16 00:01:07,720 --> 00:01:12,480 Speaker 1: versus the cloud means and how this AIPC revolution changes 17 00:01:12,520 --> 00:01:16,640 Speaker 1: how we work, live and create. Join us as we 18 00:01:16,680 --> 00:01:19,440 Speaker 1: take a journey into the future of AI computing, where 19 00:01:19,480 --> 00:01:23,399 Speaker 1: machines are not just tools but partners in our endeavors. 20 00:01:27,400 --> 00:01:32,160 Speaker 1: Welcome to Technically Speaking, an Intel podcast produced by iHeartMedia's 21 00:01:32,200 --> 00:01:37,240 Speaker 1: Ruby Studio in partnership with Intel. Hey the I'm gram class. 22 00:01:37,840 --> 00:01:40,720 Speaker 1: Joining us today is Robert Hallock, the VP and General 23 00:01:40,720 --> 00:01:44,720 Speaker 1: manager of Client AI and Technical Marketing at Intel. Here 24 00:01:44,720 --> 00:01:51,320 Speaker 1: has a long history with product development and PCs. Welcome 25 00:01:51,360 --> 00:01:52,400 Speaker 1: to the show, Robert. 26 00:01:52,360 --> 00:01:53,120 Speaker 2: Thanks for having me. 27 00:01:53,200 --> 00:01:55,280 Speaker 1: Good to be here. Yeah, I just want to the 28 00:01:55,320 --> 00:01:58,200 Speaker 1: start off by saying, there seems to be a lot 29 00:01:58,240 --> 00:02:01,320 Speaker 1: of talk around AI, and it seems to always revolve 30 00:02:01,360 --> 00:02:05,720 Speaker 1: around cloud based or software as a service type business models. 31 00:02:06,360 --> 00:02:09,919 Speaker 1: How Intel is taking a different route. Well, but can 32 00:02:09,919 --> 00:02:13,520 Speaker 1: you give us an overview of what an AIPC means 33 00:02:13,520 --> 00:02:17,120 Speaker 1: at Intel and can you help us understand some of 34 00:02:17,160 --> 00:02:21,080 Speaker 1: the benefits of running AI directly on a PC versus 35 00:02:21,120 --> 00:02:21,760 Speaker 1: in the cloud. 36 00:02:22,480 --> 00:02:25,000 Speaker 2: Yeah, a lot of people have been exposed to AI 37 00:02:25,440 --> 00:02:30,000 Speaker 2: through the cloud, and these are services like chat, GPT 38 00:02:30,440 --> 00:02:34,440 Speaker 2: or Dolly three where you can create a piece of 39 00:02:34,480 --> 00:02:38,280 Speaker 2: content from a text description. And that's one kind of 40 00:02:38,320 --> 00:02:42,240 Speaker 2: AI that's called generative AI is a category pros and 41 00:02:42,280 --> 00:02:44,560 Speaker 2: cons to that doing it in the cloud. The pro 42 00:02:44,880 --> 00:02:48,760 Speaker 2: is the models are huge, right. They essentially have the 43 00:02:48,760 --> 00:02:53,840 Speaker 2: Internet's collective knowledge of data to produce an answer or 44 00:02:53,840 --> 00:02:57,960 Speaker 2: a picture, and that's highly detailed. But one of the 45 00:02:58,000 --> 00:03:01,720 Speaker 2: cons is that they're not very specific to you what 46 00:03:01,840 --> 00:03:04,720 Speaker 2: you are doing on your PC, the information on your computer, 47 00:03:05,040 --> 00:03:08,200 Speaker 2: or the images that you care about. So that's sort 48 00:03:08,240 --> 00:03:12,960 Speaker 2: of motivating to bring AI capabilities locally to the PC 49 00:03:13,200 --> 00:03:18,040 Speaker 2: where you don't need an Internet connection to use these capabilities. 50 00:03:18,240 --> 00:03:20,880 Speaker 2: And that's sort of the tip of the iceberg, because 51 00:03:21,080 --> 00:03:25,840 Speaker 2: when you move AI to the PC without a cloud connection, 52 00:03:25,919 --> 00:03:29,560 Speaker 2: you can also do new types of AI workloads that 53 00:03:29,600 --> 00:03:33,239 Speaker 2: would be too big or simply can't run in the cloud. 54 00:03:33,360 --> 00:03:37,520 Speaker 2: So a good example of that is in teleconferencing, lots 55 00:03:37,560 --> 00:03:41,840 Speaker 2: of people use background blurring or background pictures, and we 56 00:03:41,880 --> 00:03:45,040 Speaker 2: can actually make that a lot more energy efficient. We 57 00:03:45,080 --> 00:03:48,000 Speaker 2: can actually give you hours of battery life back by 58 00:03:48,080 --> 00:03:50,320 Speaker 2: using AI on that use case. So the whole point 59 00:03:50,480 --> 00:03:53,920 Speaker 2: of AI is that there's a wave of software coming 60 00:03:54,280 --> 00:03:58,920 Speaker 2: that uses AI to improve performance or to reduce power consumption. 61 00:03:59,120 --> 00:04:02,080 Speaker 2: So you'll get better performance from your system and longer 62 00:04:02,120 --> 00:04:04,360 Speaker 2: battery life as well. And this will continue to grow 63 00:04:04,440 --> 00:04:05,880 Speaker 2: over the next five years. 64 00:04:06,440 --> 00:04:09,360 Speaker 1: And Robert I was wondering if you could please describe 65 00:04:09,360 --> 00:04:11,560 Speaker 1: to the audience what the current state of the art 66 00:04:11,600 --> 00:04:15,640 Speaker 1: is with PC architecture. You know, we've got CPUs, which 67 00:04:15,640 --> 00:04:19,279 Speaker 1: are central processing units that make computing possible, and you've 68 00:04:19,279 --> 00:04:23,560 Speaker 1: got GPUs, which are graphical processing units that are instrumental 69 00:04:23,560 --> 00:04:28,840 Speaker 1: for machine learning, gaming applications, video editing. So how does 70 00:04:28,839 --> 00:04:31,920 Speaker 1: that actually differ from the concept of an AIPC. 71 00:04:32,640 --> 00:04:35,039 Speaker 2: That's a great question. So, yeah, users may have heard 72 00:04:35,080 --> 00:04:39,239 Speaker 2: about an AIPC at this point, and at the basic level, 73 00:04:39,440 --> 00:04:44,440 Speaker 2: this is just a new generation computer with hardware inside 74 00:04:44,480 --> 00:04:49,039 Speaker 2: that is capable of accelerating an AI based workload and 75 00:04:49,240 --> 00:04:53,520 Speaker 2: previous hardware, let's say early twenty twenty three computers, you 76 00:04:53,560 --> 00:04:57,880 Speaker 2: can run AI workloads, but they will fall back to 77 00:04:57,960 --> 00:05:02,359 Speaker 2: the CPU cores, which are not as fast and not 78 00:05:02,520 --> 00:05:06,520 Speaker 2: as energy efficient as running that same workload on the 79 00:05:06,560 --> 00:05:10,960 Speaker 2: new accelerators in an AIPC. And the way we're doing 80 00:05:10,960 --> 00:05:14,679 Speaker 2: it at Intel is actually the CPU cores themselves. We've 81 00:05:14,720 --> 00:05:19,760 Speaker 2: added AI accelerating capabilities to those cores. The GPU that's 82 00:05:19,760 --> 00:05:23,159 Speaker 2: built into our processor also has accelerators, and then we've 83 00:05:23,160 --> 00:05:28,040 Speaker 2: also added an entirely new component called the NPU or 84 00:05:28,040 --> 00:05:31,640 Speaker 2: neural processing unit, which sits next to the CPU and 85 00:05:31,640 --> 00:05:34,760 Speaker 2: the GPU and is now its own third category of 86 00:05:34,839 --> 00:05:40,479 Speaker 2: acceleration on the device. And different workloads in AI, or 87 00:05:40,560 --> 00:05:44,400 Speaker 2: different features in AI run best on one of those 88 00:05:44,400 --> 00:05:46,719 Speaker 2: three engines, so you kind of need all three to 89 00:05:46,800 --> 00:05:50,400 Speaker 2: do this well. And our new product, Intel Core Ultra 90 00:05:50,760 --> 00:05:54,200 Speaker 2: is our first AIPC processor, and so if you see 91 00:05:54,200 --> 00:05:56,880 Speaker 2: that name Core Ultra, you know that it has the 92 00:05:57,279 --> 00:06:00,360 Speaker 2: AI accelerators to run these features well. But at its route, 93 00:06:00,800 --> 00:06:04,080 Speaker 2: it is a PC that has AI specific hardware. 94 00:06:04,800 --> 00:06:09,200 Speaker 1: And I'm quite interested in I guess the thinking behind 95 00:06:09,400 --> 00:06:13,359 Speaker 1: at Intel of this shift towards this new architecture. Was 96 00:06:13,400 --> 00:06:17,360 Speaker 1: it something that was internally driven or did you see 97 00:06:17,400 --> 00:06:21,080 Speaker 1: some conversations and discussions with your key customers and clients 98 00:06:21,800 --> 00:06:23,719 Speaker 1: driving that sort of shift. 99 00:06:24,000 --> 00:06:28,280 Speaker 2: This is very much a software industry driven transition, and 100 00:06:28,760 --> 00:06:32,120 Speaker 2: CPU vendors like Intel were innovating to keep pace with 101 00:06:32,160 --> 00:06:34,520 Speaker 2: what's going on in the software environment. And that's kind 102 00:06:34,520 --> 00:06:36,240 Speaker 2: of a key thing I'd want to stress. You know, 103 00:06:36,279 --> 00:06:40,040 Speaker 2: maybe you're on the fence about AIPC or you wonder, 104 00:06:40,120 --> 00:06:42,200 Speaker 2: you know, how long is this thing going to stick around. 105 00:06:42,800 --> 00:06:44,440 Speaker 2: This is one of those cases where it's both the 106 00:06:44,480 --> 00:06:48,359 Speaker 2: hardware industry and the software industry agreeing that this is 107 00:06:48,400 --> 00:06:51,960 Speaker 2: the right thing to do for performance and features and power. 108 00:06:52,600 --> 00:06:55,120 Speaker 2: Because I truly believe that over the next three to 109 00:06:55,200 --> 00:06:59,760 Speaker 2: five years we will reach this point of general acceptance 110 00:07:00,120 --> 00:07:02,760 Speaker 2: in AI, where it's whether you know it or not, 111 00:07:03,040 --> 00:07:06,680 Speaker 2: widely diffused. In most of the applications you're working on 112 00:07:06,920 --> 00:07:09,320 Speaker 2: or working with, many of the features that you value 113 00:07:09,440 --> 00:07:13,520 Speaker 2: will use AI again transparently in the background. But this 114 00:07:13,600 --> 00:07:17,920 Speaker 2: is very much a collaborative, industry wide effort from all 115 00:07:17,960 --> 00:07:21,480 Speaker 2: the hardware makers, all the big software vendors. This is 116 00:07:21,520 --> 00:07:25,720 Speaker 2: here to stay for sure. Let's pause for second here 117 00:07:26,000 --> 00:07:29,680 Speaker 2: to reiterate Robert's point. Right of adoption is always a 118 00:07:29,680 --> 00:07:32,880 Speaker 2: crucial aspect of any new technology, and that's certainly true 119 00:07:32,880 --> 00:07:36,400 Speaker 2: of what we're discussing today with AIPCS. It's important to 120 00:07:36,480 --> 00:07:38,960 Speaker 2: understand that the leaders in the field of computing, both 121 00:07:39,000 --> 00:07:42,680 Speaker 2: in the hardware and software domain, have already begun down 122 00:07:42,720 --> 00:07:45,760 Speaker 2: this revolutionary path and it doesn't seem like there's any 123 00:07:45,880 --> 00:07:48,920 Speaker 2: roadblocks in site. With that in mind, I asked Robert 124 00:07:48,960 --> 00:07:51,840 Speaker 2: about the benefits of moving AI workloads from the cloud 125 00:07:52,160 --> 00:07:54,840 Speaker 2: to the PC, especially when it comes to issues of 126 00:07:54,920 --> 00:08:00,600 Speaker 2: data privacy, latency, and connectivity. You just touched couple that 127 00:08:00,640 --> 00:08:03,320 Speaker 2: are really important to local AI. It's that data privacy 128 00:08:03,400 --> 00:08:07,400 Speaker 2: or data security. We've all read about information going up 129 00:08:07,440 --> 00:08:10,520 Speaker 2: to a cloud resource of some kind, and you don't 130 00:08:10,560 --> 00:08:13,880 Speaker 2: really know what's going to happen with your data or 131 00:08:13,880 --> 00:08:16,400 Speaker 2: your request after that. And that's not to say there's 132 00:08:16,440 --> 00:08:20,440 Speaker 2: anything malicious implied. You just don't know. So I'm sure 133 00:08:20,440 --> 00:08:23,280 Speaker 2: people in corporation, certainly at Intel, when you go to 134 00:08:23,360 --> 00:08:26,240 Speaker 2: a generative AI website, it says, hey, be careful what 135 00:08:26,280 --> 00:08:29,880 Speaker 2: you enter into this textbox. And so moving this stuff 136 00:08:29,920 --> 00:08:32,680 Speaker 2: offline gives you a couple of things. Yes, you get 137 00:08:32,840 --> 00:08:35,840 Speaker 2: chain of custody over the data. It's private, right, it's 138 00:08:35,880 --> 00:08:40,840 Speaker 2: working on your information offline, right, so it doesn't have 139 00:08:40,880 --> 00:08:43,160 Speaker 2: to go to the cloud. That's a big one. Cost 140 00:08:43,280 --> 00:08:46,960 Speaker 2: is another component. A lot of the most powerful AI 141 00:08:47,080 --> 00:08:50,559 Speaker 2: services online, you know, ten to twenty bucks a month, 142 00:08:51,240 --> 00:08:55,400 Speaker 2: and it's not cheap to have several of those. The 143 00:08:55,520 --> 00:09:02,760 Speaker 2: last that is interesting is cloud servers are intrinsically pricey. 144 00:09:03,080 --> 00:09:06,800 Speaker 2: I'm not saying they're expensive, but for AI as a 145 00:09:08,000 --> 00:09:12,360 Speaker 2: genre of software to truly take off and thrive in 146 00:09:12,400 --> 00:09:15,600 Speaker 2: all the way the software vendors want it to, it 147 00:09:15,679 --> 00:09:19,160 Speaker 2: has to reach the local PC. It has to reach 148 00:09:19,640 --> 00:09:22,160 Speaker 2: you know, tens of millions of users, hundreds of millions 149 00:09:22,200 --> 00:09:25,320 Speaker 2: of users on a local device and that's sort of 150 00:09:25,320 --> 00:09:29,800 Speaker 2: that critical mass install base that takes this effort to 151 00:09:29,880 --> 00:09:32,920 Speaker 2: the next level. Cloud was sort of one point zero 152 00:09:32,920 --> 00:09:35,480 Speaker 2: of AI, and now we're trying to, you know, for 153 00:09:35,720 --> 00:09:37,600 Speaker 2: one of a better term, create the two point zero 154 00:09:37,600 --> 00:09:41,000 Speaker 2: where lots of people have access to this and it's 155 00:09:41,000 --> 00:09:44,160 Speaker 2: widely available and in a couple of years Intel alone 156 00:09:44,160 --> 00:09:47,679 Speaker 2: we want to get one hundred million accelerators for AI 157 00:09:47,760 --> 00:09:48,680 Speaker 2: into the hands of people. 158 00:09:49,280 --> 00:09:52,760 Speaker 1: And do you have any examples of at this early 159 00:09:52,800 --> 00:09:57,080 Speaker 1: stage of actually pushing the boundaries of using this sort 160 00:09:57,120 --> 00:10:00,920 Speaker 1: of power in a local PC to do some really 161 00:10:00,920 --> 00:10:01,720 Speaker 1: interesting work. 162 00:10:03,160 --> 00:10:06,520 Speaker 2: One for an enterprise that we've been tinkering with is 163 00:10:07,200 --> 00:10:11,920 Speaker 2: a technology called rag rag, and it's the idea where 164 00:10:12,040 --> 00:10:17,600 Speaker 2: you have a language model running offline on the user's PC. 165 00:10:18,320 --> 00:10:24,679 Speaker 2: But this RAG component can scan your documents, your corporate information, 166 00:10:24,960 --> 00:10:30,400 Speaker 2: and then specialize the LM to be for you, your work, 167 00:10:30,520 --> 00:10:35,120 Speaker 2: your knowledge. So just as an example, we scanned one 168 00:10:35,200 --> 00:10:38,920 Speaker 2: state's DMV manual, which hundreds of pages long for the 169 00:10:38,920 --> 00:10:42,160 Speaker 2: Department of Motor Vehicle manual and now you can ask 170 00:10:42,440 --> 00:10:47,400 Speaker 2: very specific procedural and legal questions about the subject of 171 00:10:47,400 --> 00:10:51,439 Speaker 2: that manual and it'll spit back highly accurate answers for you. 172 00:10:52,160 --> 00:10:56,400 Speaker 2: And if you extend out out words, protecting and promoting 173 00:10:56,559 --> 00:11:00,520 Speaker 2: institutional knowledge is hard. You might have that employ that's 174 00:11:00,520 --> 00:11:04,200 Speaker 2: been there for twenty years and has all that institutional 175 00:11:04,240 --> 00:11:06,520 Speaker 2: knowledge in their head and if they leave, it goes 176 00:11:06,559 --> 00:11:11,920 Speaker 2: with them. But a RAG model could synthesize that knowledge 177 00:11:11,960 --> 00:11:14,920 Speaker 2: for people, so a new employee could just ask a 178 00:11:15,000 --> 00:11:18,320 Speaker 2: question in a text box and get an accurate answer 179 00:11:18,400 --> 00:11:21,600 Speaker 2: about what that company's working on or what this feature does. 180 00:11:22,240 --> 00:11:25,120 Speaker 2: And that's just huge for business. 181 00:11:25,480 --> 00:11:28,959 Speaker 1: Yeah, you know, I'm from a small business sort of background. 182 00:11:29,120 --> 00:11:32,280 Speaker 1: My dad has a small business. And the fact that 183 00:11:32,320 --> 00:11:35,120 Speaker 1: you can bring this power to the small and micro 184 00:11:35,200 --> 00:11:39,280 Speaker 1: businesses as well without having to pay these cloud based 185 00:11:40,120 --> 00:11:42,880 Speaker 1: prices and also in conjunction with using some of the 186 00:11:42,960 --> 00:11:46,280 Speaker 1: open source type software that I know Intel is very 187 00:11:46,280 --> 00:11:50,120 Speaker 1: supportive of. I'm just really excited to see the little guys, 188 00:11:50,200 --> 00:11:52,320 Speaker 1: you know, be able to compete with some of the 189 00:11:52,360 --> 00:11:54,040 Speaker 1: technology that the big boys have. 190 00:11:54,840 --> 00:11:58,480 Speaker 2: Absolutely, AI is, at the end of the day, a 191 00:11:58,520 --> 00:12:02,440 Speaker 2: force multiplier for a person. Right like at the root, 192 00:12:02,679 --> 00:12:06,720 Speaker 2: AI is designed to save time writing meeting minutes or 193 00:12:06,800 --> 00:12:11,440 Speaker 2: email summaries or drafting in outline. These are all just 194 00:12:11,520 --> 00:12:15,360 Speaker 2: like time consuming tasks for people, and they don't require 195 00:12:16,320 --> 00:12:19,960 Speaker 2: skill per se, but it's it's time consuming, and so 196 00:12:20,040 --> 00:12:22,480 Speaker 2: being able to offload that to a digital assistant that 197 00:12:22,480 --> 00:12:25,720 Speaker 2: can just sort of ninety percent or ninety five percent 198 00:12:25,840 --> 00:12:29,040 Speaker 2: do that for you allows you to refocus your efforts 199 00:12:29,040 --> 00:12:31,320 Speaker 2: back to something else that is more productive and more 200 00:12:31,360 --> 00:12:34,880 Speaker 2: worthy of your time. And especially for a small business, 201 00:12:35,120 --> 00:12:39,600 Speaker 2: administrative overhead, bureaucratic overhead is hard, it's time consuming. I 202 00:12:39,640 --> 00:12:44,960 Speaker 2: myself own a single member LLC and administrative stuff takes 203 00:12:45,040 --> 00:12:47,000 Speaker 2: up a ton of my time. Like I'd love to 204 00:12:47,000 --> 00:12:48,079 Speaker 2: outsource that to AI. 205 00:12:49,559 --> 00:12:53,000 Speaker 1: That's right. In terms of Intel's history with past, you know, 206 00:12:53,040 --> 00:12:58,120 Speaker 1: technological revolutions, I'm reminded of Intel's initiative to get Wi 207 00:12:58,120 --> 00:13:02,280 Speaker 1: Fi into every your laptop and that was code named Centrino, 208 00:13:03,160 --> 00:13:08,960 Speaker 1: and it's interesting to hear that again. Intel are trying 209 00:13:09,000 --> 00:13:13,000 Speaker 1: to push new technology so that it's ubiquitous, and that's 210 00:13:13,040 --> 00:13:17,959 Speaker 1: exactly what they're doing with these aipcs. And in ten 211 00:13:18,040 --> 00:13:22,000 Speaker 1: twenty years time, I think that aipowered PCs will be 212 00:13:22,080 --> 00:13:26,000 Speaker 1: so ubiquitous that we won't think anything of it. I'd 213 00:13:26,040 --> 00:13:30,160 Speaker 1: like Robert your thoughts on that and more insights into 214 00:13:30,200 --> 00:13:34,200 Speaker 1: the way Intel is evolving their strategy and Intel's role 215 00:13:34,240 --> 00:13:35,480 Speaker 1: in this. Yeah. 216 00:13:35,520 --> 00:13:38,840 Speaker 2: Actually, Centrino is a really nice analogy because that's sort 217 00:13:38,880 --> 00:13:42,160 Speaker 2: of what we're trying to do with these AI accelerators. 218 00:13:42,440 --> 00:13:45,840 Speaker 2: It's not a huge tweak to the configuration of a 219 00:13:45,880 --> 00:13:51,640 Speaker 2: system design because most of the work happens inside the CPU, 220 00:13:51,760 --> 00:13:56,720 Speaker 2: and there are very few external requirements that would change 221 00:13:56,760 --> 00:14:00,000 Speaker 2: a system designed to make this possible. Right, So it's 222 00:13:59,720 --> 00:14:03,480 Speaker 2: a system vendor could theoretically update last year's chassis to 223 00:14:03,520 --> 00:14:07,840 Speaker 2: have a new CPU and that would confer the benefits 224 00:14:07,880 --> 00:14:12,320 Speaker 2: from Core Ultra and an AI workloads and Centrino not 225 00:14:12,440 --> 00:14:15,079 Speaker 2: much different, right. You're adding a Wi Fi chip and 226 00:14:15,200 --> 00:14:16,880 Speaker 2: an antenna to the system. 227 00:14:17,080 --> 00:14:19,040 Speaker 1: Yes, But for those. 228 00:14:18,960 --> 00:14:21,720 Speaker 2: Of you who weren't around during the Centrino days, it 229 00:14:21,880 --> 00:14:24,400 Speaker 2: used to be very common that a laptop would not 230 00:14:24,520 --> 00:14:28,640 Speaker 2: have Wi Fi, which seems sort of unimaginable now, but 231 00:14:29,000 --> 00:14:32,320 Speaker 2: that's because Intel made this massive effort to make it 232 00:14:32,360 --> 00:14:35,480 Speaker 2: common and we all sort of take it for granted now. 233 00:14:36,000 --> 00:14:39,480 Speaker 2: Another analogous moment for me is the addition of graphics 234 00:14:39,680 --> 00:14:43,000 Speaker 2: in the processor. I was a hardware reviewer when that 235 00:14:43,200 --> 00:14:47,200 Speaker 2: started happening, and I remember the conversations back then like 236 00:14:48,200 --> 00:14:50,360 Speaker 2: why are we doing this? You can't even play a 237 00:14:50,360 --> 00:14:52,720 Speaker 2: game on it, what's the point? This is just going 238 00:14:52,760 --> 00:14:56,280 Speaker 2: to make CPUs more expensive, YadA, YadA, YadA. Today web 239 00:14:56,360 --> 00:15:00,880 Speaker 2: pages are rendered by your graphics card orphics accelerator in 240 00:15:00,920 --> 00:15:04,920 Speaker 2: the processor, your browser uses it. It's everywhere. So both of 241 00:15:04,960 --> 00:15:07,880 Speaker 2: those are very foundational examples for me, because I see 242 00:15:08,000 --> 00:15:12,200 Speaker 2: AI as being quite analogous in both respects. But I 243 00:15:12,240 --> 00:15:15,800 Speaker 2: think history will bear out that this was a pivotal 244 00:15:15,840 --> 00:15:19,160 Speaker 2: moment and AI will be very, very widespread, just like 245 00:15:19,200 --> 00:15:20,480 Speaker 2: graphics and just like Wi Fi. 246 00:15:23,520 --> 00:15:26,640 Speaker 1: Coming up next on Technically Speaking and Intel podcast. 247 00:15:28,440 --> 00:15:31,800 Speaker 2: Being familiar with how prompts work, it's going to be 248 00:15:31,840 --> 00:15:34,240 Speaker 2: a key business skill, and I think there'll be a 249 00:15:34,280 --> 00:15:38,000 Speaker 2: real advantage for employees who know how to engineer a 250 00:15:38,040 --> 00:15:40,240 Speaker 2: good prompt to get a great result quickly. 251 00:15:42,000 --> 00:15:44,080 Speaker 1: We'll be right back after a brief message from our 252 00:15:44,080 --> 00:15:58,120 Speaker 1: partners that Intel welcome back to Technically Speaking. I'm here 253 00:15:58,200 --> 00:16:06,400 Speaker 1: now with Robert Hallett getting back to I guess more 254 00:16:06,440 --> 00:16:09,320 Speaker 1: on the business side of things. Maybe if you can 255 00:16:09,360 --> 00:16:11,240 Speaker 1: talk a little bit about our friends in the IT 256 00:16:11,520 --> 00:16:14,400 Speaker 1: department having to manage the security and all these sort 257 00:16:14,440 --> 00:16:17,080 Speaker 1: of things. Yeah, what's some of the benefits they could 258 00:16:17,120 --> 00:16:19,680 Speaker 1: look forward to in terms of managing these types of 259 00:16:19,720 --> 00:16:21,800 Speaker 1: aipcs in their enterprise. 260 00:16:22,480 --> 00:16:25,120 Speaker 2: I think it'll make every ITDM happy that at the 261 00:16:25,200 --> 00:16:27,960 Speaker 2: end of the day, these AI applications are nothing more 262 00:16:28,120 --> 00:16:32,280 Speaker 2: than endpoint applications. You download them, you install them, and 263 00:16:32,320 --> 00:16:35,400 Speaker 2: they're from software vendors, and I'm sure they'll be breakout 264 00:16:35,520 --> 00:16:38,680 Speaker 2: ISVs that come under the scene as a result of 265 00:16:38,720 --> 00:16:43,200 Speaker 2: this AI transformation, but largely speaking, trusted vendors doing new work. 266 00:16:43,760 --> 00:16:48,280 Speaker 2: And because it's entirely offline, your user has custody of 267 00:16:48,360 --> 00:16:51,240 Speaker 2: the data and the information, which is no different from 268 00:16:51,240 --> 00:16:54,600 Speaker 2: any other application today. So it's not like this transformation 269 00:16:55,680 --> 00:16:58,960 Speaker 2: comes part and parcel with like a radical transformation and 270 00:16:59,080 --> 00:17:02,200 Speaker 2: endpoint management, which would make it way harder. From a 271 00:17:02,240 --> 00:17:07,960 Speaker 2: security point of view, AI has some very interesting benefits 272 00:17:08,160 --> 00:17:11,120 Speaker 2: for security models. So now this is the ITDM point 273 00:17:11,119 --> 00:17:15,800 Speaker 2: of view. We recently with Dell and CrowdStrike, and if 274 00:17:15,800 --> 00:17:19,280 Speaker 2: people don't know what CrowdStrike is, it's, amongst many things, 275 00:17:19,320 --> 00:17:24,280 Speaker 2: an endpoint security solution which is specifically designed to help 276 00:17:24,840 --> 00:17:28,720 Speaker 2: prevent threats that don't attack files. Many of the tax 277 00:17:29,000 --> 00:17:32,320 Speaker 2: that go into a system now aren't like a virus 278 00:17:32,400 --> 00:17:36,000 Speaker 2: that infects a file, it's actually resident in memory. They're 279 00:17:36,080 --> 00:17:43,040 Speaker 2: fileless attacks, and these are way harder to detect and prevent. 280 00:17:43,600 --> 00:17:48,320 Speaker 2: So CrowdStrike uses convolutional neural networks, which is a simpler 281 00:17:48,359 --> 00:17:52,560 Speaker 2: form of neural network or AI, to monitor the real 282 00:17:52,560 --> 00:17:57,240 Speaker 2: time conditions of the system and see if something unusual 283 00:17:57,600 --> 00:18:01,959 Speaker 2: is happening. And this experiment moves these convolutional neural network 284 00:18:02,000 --> 00:18:06,040 Speaker 2: models or CNNs to that neural processing unit or the 285 00:18:06,160 --> 00:18:09,040 Speaker 2: NPU in coraltrum. It did a couple things when we 286 00:18:09,080 --> 00:18:13,879 Speaker 2: did that. First, it gave the processor cores twenty percent 287 00:18:13,960 --> 00:18:17,119 Speaker 2: performance back. The second thing it did it made the 288 00:18:17,160 --> 00:18:20,920 Speaker 2: security model slightly smaller, so it had a smaller memory 289 00:18:20,920 --> 00:18:25,280 Speaker 2: footprint now so user gets some RAM back. And it 290 00:18:25,400 --> 00:18:28,880 Speaker 2: also improved the accuracy of the model because they can 291 00:18:28,920 --> 00:18:32,400 Speaker 2: make the model computationally bigger, right because it has its 292 00:18:32,440 --> 00:18:36,800 Speaker 2: own dedicated accelerator now to run on. And so security 293 00:18:36,840 --> 00:18:40,280 Speaker 2: got better and crowdstreg is a very popular solution, and 294 00:18:40,359 --> 00:18:44,560 Speaker 2: there are other solutions in the pipe for phishing detection, 295 00:18:44,840 --> 00:18:48,159 Speaker 2: which is notoriously hard pattern matching problem. And that's just 296 00:18:48,200 --> 00:18:52,520 Speaker 2: two examples of the way security can be enhanced by 297 00:18:52,640 --> 00:18:56,360 Speaker 2: offloading to an AI specific accelerator. 298 00:18:56,880 --> 00:19:00,239 Speaker 1: Yeah, and we talked a little bit about you know, 299 00:19:00,359 --> 00:19:04,880 Speaker 1: businesses and enterprises adopting these new aipcs. Is anything special 300 00:19:04,920 --> 00:19:08,359 Speaker 1: that IT departments and organizations need to do to prepare 301 00:19:08,359 --> 00:19:08,840 Speaker 1: for this. 302 00:19:09,280 --> 00:19:12,480 Speaker 2: I'll say there's probably three things that an ITDM would 303 00:19:12,520 --> 00:19:16,080 Speaker 2: want to think about if they intend to use large 304 00:19:16,119 --> 00:19:21,280 Speaker 2: language models or just generative AI. Those workloads are pretty 305 00:19:21,320 --> 00:19:25,560 Speaker 2: sensitive to memory bandwidth, so you wouldn't normally think about 306 00:19:25,640 --> 00:19:30,320 Speaker 2: memory bandwidth in a system purchase, but making sure that 307 00:19:30,480 --> 00:19:34,240 Speaker 2: it has two memory sticks over one. For example, right, 308 00:19:34,280 --> 00:19:36,120 Speaker 2: if you want to get sixteen gigs a RAM, make 309 00:19:36,119 --> 00:19:38,320 Speaker 2: sure that's two by eight instead of one by sixteen, 310 00:19:38,640 --> 00:19:42,639 Speaker 2: Just as an example, that will dramatically improve the performance 311 00:19:42,840 --> 00:19:47,919 Speaker 2: of the LLM, because these language models are fundamentally limited 312 00:19:48,240 --> 00:19:52,760 Speaker 2: or enhanced by the performance of the memory subsystem. Outside 313 00:19:52,800 --> 00:19:57,840 Speaker 2: of that, for other forms of AI in the year ahead, 314 00:19:58,680 --> 00:20:03,080 Speaker 2: there is a calculation called TOPS or terra operations per second, 315 00:20:03,080 --> 00:20:06,240 Speaker 2: which is sort of a ballpark for how much AI 316 00:20:06,400 --> 00:20:12,280 Speaker 2: performance a device can give you. Software makes or breaks 317 00:20:12,440 --> 00:20:15,960 Speaker 2: the acquisition of that TOPS figure. So I can give 318 00:20:16,000 --> 00:20:19,720 Speaker 2: you a billion TOPS, and if I had a very 319 00:20:19,760 --> 00:20:23,960 Speaker 2: poor software stack under that you would never see the 320 00:20:24,040 --> 00:20:28,160 Speaker 2: billion tops. So there's a new level of knowledge the 321 00:20:28,200 --> 00:20:32,600 Speaker 2: ITDM needs to develop on sort of software stack robustness 322 00:20:33,560 --> 00:20:37,000 Speaker 2: underneath that rating. Okay, so you have to be familiar 323 00:20:37,080 --> 00:20:39,520 Speaker 2: with what Intel's doing in the software space, what its 324 00:20:39,520 --> 00:20:42,960 Speaker 2: competitors are doing in the software space to really understand 325 00:20:43,640 --> 00:20:45,920 Speaker 2: whether or not you're going to get a good experience 326 00:20:46,280 --> 00:20:48,440 Speaker 2: out of the device you're purchasing. So that's number two, 327 00:20:49,160 --> 00:20:52,600 Speaker 2: and number three I think would be to say that 328 00:20:53,160 --> 00:20:56,600 Speaker 2: itdms will probably be faced with a lot of advocacy 329 00:20:56,800 --> 00:21:01,879 Speaker 2: for the NPU as an AIX, but it's important to 330 00:21:01,960 --> 00:21:07,160 Speaker 2: understand that the software industry broadly also wants to use 331 00:21:07,520 --> 00:21:11,080 Speaker 2: GPU and CPU as well. So if you make an 332 00:21:11,160 --> 00:21:15,959 Speaker 2: upgrade decision that is all in on NPU performance and 333 00:21:16,000 --> 00:21:20,159 Speaker 2: you didn't check on the GPU or the CPU, you 334 00:21:20,280 --> 00:21:22,920 Speaker 2: may be out in the cold on performance or power 335 00:21:22,960 --> 00:21:26,760 Speaker 2: efficiency for these Frankly a large number of workloads that 336 00:21:26,880 --> 00:21:30,399 Speaker 2: use graphics and CPU for AI acceleration. Those are the 337 00:21:30,440 --> 00:21:33,400 Speaker 2: three things that I would say are new or different 338 00:21:33,920 --> 00:21:38,480 Speaker 2: in this era, But overall it is something that you 339 00:21:38,520 --> 00:21:43,119 Speaker 2: can integrate into your upgrade cycle piecemeal. YEP, Newer devices 340 00:21:43,200 --> 00:21:46,200 Speaker 2: will be faster. I mean, they always are, but it's 341 00:21:46,200 --> 00:21:49,480 Speaker 2: not like you're missing out on a new feature, right, 342 00:21:49,520 --> 00:21:51,959 Speaker 2: We're going to make them faster, But the features are 343 00:21:51,960 --> 00:21:54,600 Speaker 2: delivered by the software, are not the hardware. You can 344 00:21:54,680 --> 00:21:56,520 Speaker 2: kind of get in at any point and get the 345 00:21:56,560 --> 00:21:59,120 Speaker 2: goodness of AI, which is pretty cool as well. 346 00:21:59,800 --> 00:22:04,119 Speaker 1: Yeah, yeah, And are you helping the software vendors, I 347 00:22:04,200 --> 00:22:07,400 Speaker 1: guess compile their code so that it will help them 348 00:22:07,760 --> 00:22:09,879 Speaker 1: really utilize that hardware underneath. 349 00:22:10,840 --> 00:22:13,760 Speaker 2: Yeah, that's the secret of AI. And I'll start with 350 00:22:13,800 --> 00:22:16,600 Speaker 2: an analogy of PC gaming, which I think is a 351 00:22:16,640 --> 00:22:19,679 Speaker 2: lot more familiar to people. So at the bottom, you 352 00:22:19,760 --> 00:22:23,480 Speaker 2: have a piece of hardware, a graphics card, and they 353 00:22:23,520 --> 00:22:27,480 Speaker 2: tell you, you know, it's certain terra flops or gigaflops of performance, 354 00:22:28,000 --> 00:22:32,679 Speaker 2: but everybody knows that really depends on how well optimized 355 00:22:32,720 --> 00:22:37,000 Speaker 2: the game engine is, how well optimized the game you're 356 00:22:37,080 --> 00:22:41,520 Speaker 2: running is, how good the graphics drivers are. AI is 357 00:22:42,000 --> 00:22:47,000 Speaker 2: no different. AI accelerators are actually exposed in direct X 358 00:22:47,160 --> 00:22:52,119 Speaker 2: in Windows as a GPU without display outputs, so the 359 00:22:52,160 --> 00:22:57,280 Speaker 2: system sees them as essentially graphics cards, and instead of 360 00:22:57,320 --> 00:23:01,080 Speaker 2: game engines you have AI models will have features and 361 00:23:01,320 --> 00:23:05,080 Speaker 2: apps just like games you even have run times or 362 00:23:05,080 --> 00:23:08,439 Speaker 2: an environment where the code is running in there's a 363 00:23:08,480 --> 00:23:12,240 Speaker 2: DirectX run time for graphics, there is equivalent run times 364 00:23:12,280 --> 00:23:16,280 Speaker 2: for AI. So in many respects, the AI software stack 365 00:23:16,480 --> 00:23:20,959 Speaker 2: looks and works a lot like the gaming software stack. 366 00:23:21,400 --> 00:23:25,440 Speaker 2: And so if people think about all the times that 367 00:23:25,760 --> 00:23:29,680 Speaker 2: a GPU that's supposed to be faster on paper didn't 368 00:23:29,760 --> 00:23:32,679 Speaker 2: live up to that number because of one software reason 369 00:23:32,800 --> 00:23:36,960 Speaker 2: or another, that is a possible reality for AI as well. 370 00:23:37,000 --> 00:23:40,800 Speaker 2: And that's why it's so important to make your decision 371 00:23:41,720 --> 00:23:45,280 Speaker 2: not just on tops or what the accelerator is, but 372 00:23:45,960 --> 00:23:49,000 Speaker 2: on the robustness of the software underneath, because that's where 373 00:23:49,000 --> 00:23:50,160 Speaker 2: it really happens. 374 00:23:52,240 --> 00:23:55,639 Speaker 1: Regular listeners to the show Mike remember back in season one, 375 00:23:55,720 --> 00:23:58,520 Speaker 1: we divided a whole episode on the skills that workers 376 00:23:58,520 --> 00:24:02,240 Speaker 1: will need in order to take advantage of the kinds 377 00:24:02,240 --> 00:24:05,919 Speaker 1: of AI tools we discussed today. AI technology can only 378 00:24:06,080 --> 00:24:09,840 Speaker 1: expand our expectations of what's possible if we understand the 379 00:24:09,880 --> 00:24:12,760 Speaker 1: most effective ways to use it. So I asked Robert 380 00:24:12,800 --> 00:24:15,119 Speaker 1: for his thoughts and what workers should focus on to 381 00:24:15,160 --> 00:24:17,960 Speaker 1: take advantage of this new wave of technology, and he 382 00:24:18,040 --> 00:24:20,480 Speaker 1: began his answer with something just about everyone does on 383 00:24:20,520 --> 00:24:23,200 Speaker 1: the Internet, every day. 384 00:24:23,320 --> 00:24:26,000 Speaker 2: Ooh, that's a good one. I actually want to start 385 00:24:26,080 --> 00:24:29,800 Speaker 2: briefly at web searching. You know, web searching, a good, 386 00:24:30,080 --> 00:24:34,680 Speaker 2: well composed query is a learned skill. We have all 387 00:24:34,800 --> 00:24:38,800 Speaker 2: encountered people that haven't quite learned that skill, and they 388 00:24:38,840 --> 00:24:41,639 Speaker 2: get bad search results and they're frustrated with a search engine. 389 00:24:41,760 --> 00:24:44,399 Speaker 2: And not a lot of people teach that skill because 390 00:24:44,400 --> 00:24:48,840 Speaker 2: it's its own language of sorts. You know, you want 391 00:24:48,880 --> 00:24:51,720 Speaker 2: to freeze things to a search box differently than how 392 00:24:51,760 --> 00:24:54,639 Speaker 2: I would ask it out loud. So that takes me 393 00:24:54,680 --> 00:24:59,960 Speaker 2: to AI, where you are still engineering a search of sorts. 394 00:25:00,640 --> 00:25:03,880 Speaker 2: It's called a prompt now, but it's a search and 395 00:25:04,760 --> 00:25:07,360 Speaker 2: the skill is in like, how do you craft that 396 00:25:07,600 --> 00:25:11,399 Speaker 2: prompt to get the result that you're looking for? You 397 00:25:11,440 --> 00:25:15,840 Speaker 2: have to be somewhat specific, and you have to understand 398 00:25:15,880 --> 00:25:18,480 Speaker 2: the limitations of the AI model you're working with. So 399 00:25:18,720 --> 00:25:22,560 Speaker 2: let's take an image creation model. Some of them only 400 00:25:22,640 --> 00:25:26,280 Speaker 2: support what's called positive prompts. You have to describe the 401 00:25:26,280 --> 00:25:29,600 Speaker 2: things that you want and it will kind of omit 402 00:25:29,760 --> 00:25:33,640 Speaker 2: anything else that you're not asking for explicitly. Others support 403 00:25:33,720 --> 00:25:37,520 Speaker 2: negative prompts, and if you're working with a positive only 404 00:25:37,960 --> 00:25:43,000 Speaker 2: image service, you could easily have users saying I want this, 405 00:25:43,000 --> 00:25:45,120 Speaker 2: this and this, but not that, that and that. Yeah, 406 00:25:45,520 --> 00:25:49,560 Speaker 2: the not operator the AI engine's not going to understand it, 407 00:25:49,680 --> 00:25:52,199 Speaker 2: so it's going to give you the things in the 408 00:25:52,359 --> 00:25:54,320 Speaker 2: not category. In the picture. 409 00:25:54,800 --> 00:25:57,240 Speaker 1: Yeah, it's like saying I want a picture of a zoo, 410 00:25:57,280 --> 00:25:59,080 Speaker 1: but no elephants. 411 00:25:58,920 --> 00:26:01,680 Speaker 2: Right, and it doesn't nderstand the word no, So in. 412 00:26:01,680 --> 00:26:03,520 Speaker 1: A positive one, it'll just have elephants. 413 00:26:03,840 --> 00:26:07,479 Speaker 2: Yeah, you get all elephants exactly right. So understanding this 414 00:26:07,560 --> 00:26:10,879 Speaker 2: is a real skill. Understanding like what the AI model 415 00:26:11,160 --> 00:26:15,280 Speaker 2: or by distant extension, what the game engine will let 416 00:26:15,320 --> 00:26:20,400 Speaker 2: you do is really important, and how to be specific 417 00:26:20,560 --> 00:26:26,000 Speaker 2: and understanding how to provide follow up commentary to tweak 418 00:26:26,119 --> 00:26:27,680 Speaker 2: the result to get what you're looking for. 419 00:26:28,000 --> 00:26:28,520 Speaker 1: This is a. 420 00:26:28,480 --> 00:26:34,360 Speaker 2: Whole category of skill that is not taught, not widely 421 00:26:34,480 --> 00:26:37,359 Speaker 2: known in business or even amongst users because this is 422 00:26:37,400 --> 00:26:42,000 Speaker 2: so new. But being familiar with how prompts work is 423 00:26:42,080 --> 00:26:44,440 Speaker 2: going to be a key business skill, and I think 424 00:26:44,440 --> 00:26:47,320 Speaker 2: there'll be a real advantage for employees who know how 425 00:26:47,359 --> 00:26:50,919 Speaker 2: to engineer a good prompt to get a great result quickly. 426 00:26:51,920 --> 00:26:53,800 Speaker 1: So going to take a bit of a step back 427 00:26:53,880 --> 00:26:56,639 Speaker 1: and look far in the horizon in the sort of 428 00:26:56,640 --> 00:27:01,679 Speaker 1: the three to five seven year time range. What's Intel's 429 00:27:02,040 --> 00:27:05,600 Speaker 1: plan for this journey into AI. 430 00:27:06,520 --> 00:27:08,720 Speaker 2: I think at top level, i'd want to acknowledge that 431 00:27:08,760 --> 00:27:13,400 Speaker 2: I understand the skepticism or the I don't understand of AI. 432 00:27:13,560 --> 00:27:17,880 Speaker 2: But where we're going is like a complete industry transformation. 433 00:27:18,200 --> 00:27:21,600 Speaker 2: Intel believes by twenty twenty seven or twenty twenty eight 434 00:27:21,720 --> 00:27:25,400 Speaker 2: that eighty percent of all the computer sold will have 435 00:27:25,520 --> 00:27:29,880 Speaker 2: AI accelerators inside. And by the end of this year, 436 00:27:30,280 --> 00:27:34,120 Speaker 2: we want at least one hundred different software developers partnered 437 00:27:34,200 --> 00:27:37,800 Speaker 2: with Intel. We want to deliver three hundred or so 438 00:27:37,960 --> 00:27:42,119 Speaker 2: different AI features into the marketplace through all of those companies, 439 00:27:42,400 --> 00:27:46,240 Speaker 2: help them optimize it, deliver it, market it. We want 440 00:27:46,240 --> 00:27:49,359 Speaker 2: to bring one hundred million accelerators into the space by 441 00:27:49,359 --> 00:27:52,960 Speaker 2: the end of twenty twenty five, and so just between 442 00:27:53,080 --> 00:27:56,320 Speaker 2: now and twenty twenty seven twenty eight, we're talking about 443 00:27:56,800 --> 00:27:59,919 Speaker 2: zero to eighty percent of the market in under four years, 444 00:28:00,520 --> 00:28:05,359 Speaker 2: which is an extraordinary velocity. And as of right now, 445 00:28:05,920 --> 00:28:09,119 Speaker 2: Intel has the largest number of accelerators, the most number 446 00:28:09,119 --> 00:28:12,680 Speaker 2: of applications in the market, and sort of below the scenes, 447 00:28:13,280 --> 00:28:16,359 Speaker 2: all the enabling tools and softwares and frameworks, we also 448 00:28:16,400 --> 00:28:18,680 Speaker 2: have the most of those as well, So we want 449 00:28:18,720 --> 00:28:22,119 Speaker 2: to be the scale provider for AI, you know, the 450 00:28:22,160 --> 00:28:24,520 Speaker 2: biggest install based and we want to help to succeed 451 00:28:24,600 --> 00:28:27,600 Speaker 2: because that's what software vendors are asking us to do. 452 00:28:28,119 --> 00:28:29,560 Speaker 2: So that's the longest short of it. 453 00:28:30,080 --> 00:28:33,000 Speaker 1: That's great. I think I'll leave it there. Thanks Robert. 454 00:28:33,400 --> 00:28:38,840 Speaker 1: Thanks my deepest thanks to Robert Hallock for his invaluable 455 00:28:38,840 --> 00:28:44,680 Speaker 1: contributions to today's episode of Technically Speaking. Robert's passion and 456 00:28:44,840 --> 00:28:47,320 Speaker 1: enthusiasm has convinced me that we are on the brink 457 00:28:47,480 --> 00:28:51,480 Speaker 1: of a revolutionary technology that could transform humanity. We've all 458 00:28:51,560 --> 00:28:55,200 Speaker 1: used PCs both in our personal and professional lives, but 459 00:28:55,240 --> 00:28:58,560 Speaker 1: the leap to aipowered processes represents a once in a 460 00:28:58,680 --> 00:29:03,360 Speaker 1: generation advancement. Imagine having your own personal AIS system that 461 00:29:03,480 --> 00:29:07,280 Speaker 1: continually adapts to your needs and priorities. Admittedly, this is 462 00:29:07,320 --> 00:29:10,840 Speaker 1: a prospect I'm still trying to fully grasp, but I 463 00:29:10,960 --> 00:29:14,640 Speaker 1: especially appreciate that this technology will operate locally on my PC, 464 00:29:15,240 --> 00:29:19,200 Speaker 1: keeping my data private and secure. I'm also particularly interested 465 00:29:19,240 --> 00:29:23,120 Speaker 1: in how technology can empower the underdog to do amazing things. 466 00:29:23,720 --> 00:29:26,720 Speaker 1: The garage based tech entrepreneur building the next world beating 467 00:29:26,720 --> 00:29:30,040 Speaker 1: app the corner cafe owner getting that extra hour per 468 00:29:30,120 --> 00:29:32,760 Speaker 1: day to spend with their family, and the first year 469 00:29:32,800 --> 00:29:37,400 Speaker 1: design student creating beautiful and innovative art. With AI enhanced 470 00:29:37,400 --> 00:29:41,800 Speaker 1: pieces readily accessible, I believe we can greatly advance human prosperity. 471 00:29:42,720 --> 00:29:48,720 Speaker 1: The future looks bright. Ground. Next episode will continue to 472 00:29:48,800 --> 00:29:51,680 Speaker 1: unpact the involvement of AI to create a more accessible 473 00:29:51,760 --> 00:29:55,360 Speaker 1: and livable city for everyone. Join us again on Tuesday, 474 00:29:55,400 --> 00:29:59,360 Speaker 1: April twenty third for another enlightening discussion here on Technically 475 00:29:59,400 --> 00:30:06,160 Speaker 1: Speaking and Intel podcast. Technically Speaking was produced by Ruby 476 00:30:06,200 --> 00:30:10,280 Speaker 1: Studio from iHeartRadio in partnership with Intel and hosted by 477 00:30:10,320 --> 00:30:14,680 Speaker 1: me Graham Class. Our executive producer is Molly Sosher, our 478 00:30:14,720 --> 00:30:18,400 Speaker 1: EP of Post Production is James Foster, and our supervising 479 00:30:18,440 --> 00:30:22,840 Speaker 1: producer is Nikia Swinton. This episode was edited by Sierra 480 00:30:22,880 --> 00:30:26,240 Speaker 1: Spreen and was written by Molly Sosher and Nick Ferschall.