1 00:00:04,480 --> 00:00:12,719 Speaker 1: Welcome to Tech Stuff, a production from iHeartRadio. Hey there, 2 00:00:12,720 --> 00:00:16,080 Speaker 1: and welcome to tech Stuff. I'm your host, Jonathan Strickland. 3 00:00:16,079 --> 00:00:19,360 Speaker 1: I'm an executive producer with iHeart Podcasts and How the 4 00:00:19,400 --> 00:00:22,760 Speaker 1: Tech Are You. So I've got a whole bunch of 5 00:00:22,960 --> 00:00:27,440 Speaker 1: shorter episodes that are focusing on individual, big stories that 6 00:00:27,480 --> 00:00:29,640 Speaker 1: happen in twenty twenty three. I used to do these 7 00:00:30,120 --> 00:00:35,640 Speaker 1: massive tech stuff episodes that spanned maybe two hours of 8 00:00:36,240 --> 00:00:38,640 Speaker 1: content to talk about the big tech stories that happened 9 00:00:38,640 --> 00:00:42,240 Speaker 1: throughout the year. This year, instead, I'm doing individual, shorter 10 00:00:42,320 --> 00:00:44,320 Speaker 1: episodes where I can go into a little more detail 11 00:00:44,800 --> 00:00:47,720 Speaker 1: in those stories and still cover the big stuff that happened. 12 00:00:47,720 --> 00:00:51,040 Speaker 1: So first up, early this year, I said that twenty 13 00:00:51,120 --> 00:00:54,360 Speaker 1: twenty three was going to belong to AI in general 14 00:00:54,480 --> 00:00:58,200 Speaker 1: and generative AI in particular, and I think it's pretty 15 00:00:58,240 --> 00:01:02,560 Speaker 1: safe to say that being a solid prediction, it was 16 00:01:02,600 --> 00:01:04,920 Speaker 1: also an obvious one. I'm not going to pat myself 17 00:01:04,920 --> 00:01:08,240 Speaker 1: on the back for that, but let's be clear, artificial 18 00:01:08,240 --> 00:01:13,960 Speaker 1: intelligence as a discipline stretches back decades, right mid twentieth 19 00:01:14,000 --> 00:01:17,600 Speaker 1: century and even earlier if you're talking about things like 20 00:01:17,959 --> 00:01:22,200 Speaker 1: theory and generative AI has been around for quite some 21 00:01:22,319 --> 00:01:25,920 Speaker 1: time too. Open Ai, which originally started off as a 22 00:01:25,959 --> 00:01:31,080 Speaker 1: nonprofit organization dedicated to advancing artificial intelligence in an ethical 23 00:01:31,280 --> 00:01:36,360 Speaker 1: and safe way, really shook things up in twenty twenty two. 24 00:01:36,400 --> 00:01:40,080 Speaker 1: That's when they released chat GPT chatbot that's built on 25 00:01:40,120 --> 00:01:45,080 Speaker 1: top of the company's Large Language Model or LLM, so 26 00:01:45,200 --> 00:01:48,840 Speaker 1: the chatbot draws from the LM, you can think of 27 00:01:48,880 --> 00:01:52,160 Speaker 1: it that way. But this sparked essentially a race for 28 00:01:52,240 --> 00:01:56,000 Speaker 1: second place, and you had companies like Google and Meta 29 00:01:56,080 --> 00:02:00,120 Speaker 1: and Amazon, even Apple and more started to explore ways 30 00:02:00,280 --> 00:02:04,240 Speaker 1: to develop and integrate generative AI tools. In some cases, 31 00:02:04,600 --> 00:02:09,560 Speaker 1: that race ended up producing a lot of questionable decisions. 32 00:02:09,919 --> 00:02:13,160 Speaker 1: Right in an effort to try and catch up, companies 33 00:02:13,200 --> 00:02:18,600 Speaker 1: were cutting quarters and launching products definitely before they were ready. Meanwhile, 34 00:02:18,639 --> 00:02:23,240 Speaker 1: we've seen critics express concerns about generative AI in particular, 35 00:02:23,320 --> 00:02:28,239 Speaker 1: and of course AI as a broader concept. You've got 36 00:02:28,680 --> 00:02:32,640 Speaker 1: creators such as artists and writers who worry that AI 37 00:02:32,680 --> 00:02:36,160 Speaker 1: companies are using works made by humans to train up 38 00:02:36,639 --> 00:02:40,200 Speaker 1: AI models, all without first securing permission from the original 39 00:02:40,200 --> 00:02:45,000 Speaker 1: creators to do so. Ultimately, these generative AI tools can 40 00:02:45,080 --> 00:02:49,200 Speaker 1: mimic real specific human creators, so you could tell a 41 00:02:49,320 --> 00:02:54,040 Speaker 1: generative AI chatbot to write a story about a washed 42 00:02:54,120 --> 00:02:58,120 Speaker 1: up author in New England who faces some supernatural threat 43 00:02:58,320 --> 00:03:01,519 Speaker 1: with the added component of in the style of Stephen King, 44 00:03:01,919 --> 00:03:04,800 Speaker 1: for example, But for a chatbot to be able to 45 00:03:04,880 --> 00:03:07,760 Speaker 1: do that, it would first need to train on the 46 00:03:07,800 --> 00:03:10,760 Speaker 1: works of Stephen King in order to grasp the elements 47 00:03:11,160 --> 00:03:14,520 Speaker 1: of Stephen King's style and delivery in order to mimic it. 48 00:03:14,960 --> 00:03:18,320 Speaker 1: The creatives say that these AI companies are exploiting works 49 00:03:18,400 --> 00:03:22,400 Speaker 1: of human effort and potentially they're making it much harder 50 00:03:22,440 --> 00:03:25,280 Speaker 1: for genuine creatives to make a living off their art 51 00:03:25,360 --> 00:03:27,959 Speaker 1: in the long run. That is not a small thing. 52 00:03:28,320 --> 00:03:32,440 Speaker 1: And a lot of those works aren't necessarily available for 53 00:03:32,560 --> 00:03:36,560 Speaker 1: just public consumption. They're locked behind paywalls of some sort. 54 00:03:36,640 --> 00:03:39,920 Speaker 1: So how did the AI model get it? Those are 55 00:03:39,960 --> 00:03:44,800 Speaker 1: the sort of questions that these creators are asking. Generative AI, 56 00:03:45,080 --> 00:03:48,840 Speaker 1: artificial intelligence in general, and automation have all played into 57 00:03:49,000 --> 00:03:52,600 Speaker 1: a lot of discussions about companies replacing human staff with 58 00:03:53,040 --> 00:03:57,560 Speaker 1: algorithms and chat pots and language models. And this isn't 59 00:03:57,760 --> 00:04:00,440 Speaker 1: a hypothetical thing. It's not like, oh, in the future, 60 00:04:00,480 --> 00:04:04,720 Speaker 1: we're going to start seeing AI displace human workers, which 61 00:04:04,880 --> 00:04:07,360 Speaker 1: has been a fear for a very long time right. 62 00:04:07,960 --> 00:04:13,200 Speaker 1: But over the summer, IBM's CEO Arvin Krishna indicated that 63 00:04:13,320 --> 00:04:17,080 Speaker 1: in addition to putting on a hiring freeze across the company, 64 00:04:17,440 --> 00:04:20,440 Speaker 1: he was looking at the long term prospect of replacing 65 00:04:20,880 --> 00:04:25,240 Speaker 1: thousands of jobs at IBM with AI, and he indicated 66 00:04:25,279 --> 00:04:27,080 Speaker 1: that the first jobs that would be likely to go 67 00:04:27,120 --> 00:04:30,839 Speaker 1: to the robots would be stuff like human resources positions. 68 00:04:31,400 --> 00:04:34,520 Speaker 1: But throughout the year, we've seen two interconnected narratives play 69 00:04:34,560 --> 00:04:40,680 Speaker 1: out throughout multiple industries. First, due to the global economic situation, 70 00:04:41,040 --> 00:04:44,560 Speaker 1: a lot of companies are scaling back significantly and they 71 00:04:44,560 --> 00:04:47,360 Speaker 1: are laying off workers left, right, and center. We've seen 72 00:04:47,360 --> 00:04:49,719 Speaker 1: it a lot in the tech space, but it's not 73 00:04:49,800 --> 00:04:53,520 Speaker 1: the only industry to have this happen. Secondly, with the 74 00:04:53,600 --> 00:04:59,000 Speaker 1: rise of generative AI, we've seen some company leaders experiment 75 00:04:59,080 --> 00:05:03,640 Speaker 1: with offloading tech asks to AI powered tools. Now, in 76 00:05:03,720 --> 00:05:08,120 Speaker 1: ideal situations, the AI is meant to augment the work 77 00:05:08,440 --> 00:05:11,719 Speaker 1: of human staffers, not replace them, but to make them 78 00:05:11,760 --> 00:05:15,080 Speaker 1: more effective and more efficient, perhaps leading to things like 79 00:05:15,600 --> 00:05:19,640 Speaker 1: a four day workweek. But in at least a few cases, 80 00:05:20,240 --> 00:05:24,159 Speaker 1: including in the field of writing content specifically for the Internet. 81 00:05:24,880 --> 00:05:28,280 Speaker 1: We've seen a few companies assign writing gigs to AI 82 00:05:28,480 --> 00:05:32,880 Speaker 1: powered generative tools and to just eliminate the human element 83 00:05:33,200 --> 00:05:38,599 Speaker 1: almost entirely. Even my old employer, HowStuffWorks dot Com did that, 84 00:05:39,200 --> 00:05:42,080 Speaker 1: and when the editorial staff raised concerns about the move, 85 00:05:42,839 --> 00:05:47,000 Speaker 1: they were let go. Yikes, But there were other concerns 86 00:05:47,000 --> 00:05:49,520 Speaker 1: as well. This year we heard a lot about an 87 00:05:49,560 --> 00:05:54,200 Speaker 1: issue called hallucinations. Now that term is a bit whibly wobbly, 88 00:05:54,800 --> 00:05:57,600 Speaker 1: as the time Lord would say, it leads to a 89 00:05:57,640 --> 00:06:03,120 Speaker 1: potential alternative label call called confabulations, but either way, the 90 00:06:03,160 --> 00:06:08,640 Speaker 1: output is the same. Sometimes generative AI hits a gap 91 00:06:09,080 --> 00:06:12,320 Speaker 1: in its knowledge, but it is still compelled to give 92 00:06:12,360 --> 00:06:16,320 Speaker 1: an answer to a request, and like a stereotypical dad 93 00:06:16,440 --> 00:06:19,920 Speaker 1: in an American sitcom, it appears to be incapable or 94 00:06:20,000 --> 00:06:23,280 Speaker 1: unwilling to say you know what, I don't know that, 95 00:06:23,920 --> 00:06:27,360 Speaker 1: and instead it just offers up information that sounds reliable 96 00:06:27,680 --> 00:06:31,520 Speaker 1: but in fact it's totally made up. To understand why 97 00:06:31,560 --> 00:06:34,160 Speaker 1: this happens, it helps to have a very basic, high 98 00:06:34,279 --> 00:06:38,360 Speaker 1: level concept of how chatbots form sentences in the first place. 99 00:06:38,720 --> 00:06:42,919 Speaker 1: So deep down, a chatbot follows a statistical model to 100 00:06:43,000 --> 00:06:47,360 Speaker 1: generate responses to queries. Based on the query, the chatbot 101 00:06:47,400 --> 00:06:52,400 Speaker 1: evaluates numerous potential responses, and generally speaking, it picks the 102 00:06:52,440 --> 00:06:55,920 Speaker 1: most likely word to come next in a sentence. If 103 00:06:56,120 --> 00:06:59,920 Speaker 1: the model has access to real world knowledge to fill 104 00:07:00,120 --> 00:07:03,960 Speaker 1: out its response, it'll favor the real world knowledge and 105 00:07:04,000 --> 00:07:08,360 Speaker 1: include that in the answer. But if it doesn't, well, 106 00:07:08,440 --> 00:07:10,840 Speaker 1: it might just fill in the blanks with words that 107 00:07:10,880 --> 00:07:14,160 Speaker 1: are it deems are most likely from a statistical point 108 00:07:14,200 --> 00:07:18,120 Speaker 1: of view, to follow the previous words. The problem is 109 00:07:18,120 --> 00:07:21,920 Speaker 1: that while the words might be correct from a statistical standpoint, 110 00:07:22,120 --> 00:07:26,200 Speaker 1: they may not actually reflect the truth. So, for example, 111 00:07:26,880 --> 00:07:31,800 Speaker 1: according to Gizmodo, Google's own AI tool gave precise instructions 112 00:07:31,840 --> 00:07:35,240 Speaker 1: on how to cook a poisonous mushroom with the scientific 113 00:07:35,320 --> 00:07:41,360 Speaker 1: name amanita O Creata, also known as the Angel of 114 00:07:41,640 --> 00:07:46,760 Speaker 1: Death or destroying Angel. So, in other words, this is 115 00:07:46,800 --> 00:07:52,280 Speaker 1: a seriously toxic mushroom. So the query was asking Google 116 00:07:52,320 --> 00:07:55,120 Speaker 1: to come up with a way to cook the mushroom safely, 117 00:07:55,480 --> 00:07:58,560 Speaker 1: presumably meaning safe enough so that you could eat it afterward. 118 00:07:59,040 --> 00:08:03,200 Speaker 1: And Google's instruction included soaking the mushrooms in water in 119 00:08:03,240 --> 00:08:06,480 Speaker 1: an effort to leach out the toxins. And it did 120 00:08:06,480 --> 00:08:08,280 Speaker 1: say like you need to be super careful and it 121 00:08:08,360 --> 00:08:10,760 Speaker 1: might take a long time to do this. But here's 122 00:08:10,800 --> 00:08:15,400 Speaker 1: the problem. According to Gizmoto, the toxins in amanita okreata 123 00:08:15,440 --> 00:08:19,760 Speaker 1: are not water soluble. It wouldn't work to soak the mushrooms. 124 00:08:19,880 --> 00:08:22,680 Speaker 1: The poison wouldn't leach out in the first place, so 125 00:08:23,720 --> 00:08:25,880 Speaker 1: the user would be left with mushrooms that were just 126 00:08:25,960 --> 00:08:27,840 Speaker 1: as deadly as they were before you put them in 127 00:08:27,840 --> 00:08:31,880 Speaker 1: the water. The AI tool should have recognized this and 128 00:08:32,000 --> 00:08:35,240 Speaker 1: simply responded with something along the lines of this mushroom 129 00:08:35,280 --> 00:08:38,600 Speaker 1: contains deadly toxins with no shrefire method of removing them 130 00:08:38,880 --> 00:08:42,760 Speaker 1: and should never be consumed. In story, it shouldn't have 131 00:08:42,800 --> 00:08:46,840 Speaker 1: come up with maybe this would work, because people could 132 00:08:46,880 --> 00:08:50,040 Speaker 1: die if they actually followed those directions. Now, some of 133 00:08:50,080 --> 00:08:53,679 Speaker 1: the stories about AI were more about not the technology itself, 134 00:08:54,160 --> 00:08:57,400 Speaker 1: but rather our approach to it and our point of 135 00:08:57,480 --> 00:09:00,960 Speaker 1: view of it. For example, Computer Weekly These Cliffs. Saren 136 00:09:01,280 --> 00:09:05,920 Speaker 1: published a piece titled few organizations have a clear strategy 137 00:09:05,920 --> 00:09:09,880 Speaker 1: for AI, and Saren cites a study by a company 138 00:09:09,880 --> 00:09:13,680 Speaker 1: called mesh Ai Limited that said only fifteen percent of 139 00:09:13,800 --> 00:09:18,640 Speaker 1: organizations have a clear strategy for integrating AI into their organization. Meanwhile, 140 00:09:18,880 --> 00:09:22,320 Speaker 1: it seems like every organization is actually exploring AI to 141 00:09:22,400 --> 00:09:26,200 Speaker 1: some degree. Obviously, not every organization is going to implement 142 00:09:26,320 --> 00:09:30,240 Speaker 1: fully some half baked AI scheme, but some of them 143 00:09:30,320 --> 00:09:33,400 Speaker 1: certainly seem to be trying. And if there's one lesson 144 00:09:33,440 --> 00:09:35,720 Speaker 1: we can take away from tech in general, it's that's 145 00:09:35,840 --> 00:09:38,680 Speaker 1: raally a good idea to put a new, poorly understood 146 00:09:38,720 --> 00:09:42,040 Speaker 1: technology to use. Still, there's a sense that if a 147 00:09:42,080 --> 00:09:45,560 Speaker 1: company doesn't move in on AI soon, it's going to 148 00:09:45,559 --> 00:09:49,560 Speaker 1: be left behind by its competitors. There's market pressure in 149 00:09:49,640 --> 00:09:52,480 Speaker 1: place here that's at odds with the lack of a 150 00:09:52,480 --> 00:09:55,560 Speaker 1: clear strategy, and the rise in interest in AI also 151 00:09:55,600 --> 00:10:00,000 Speaker 1: fuels other parts of the tech industry. Specifically microchip MANUFAT 152 00:10:00,520 --> 00:10:04,640 Speaker 1: are rushing to meet demand. They're producing high performance processors 153 00:10:04,960 --> 00:10:08,600 Speaker 1: that are best suited for certain AI implementations. In Nvidia 154 00:10:08,800 --> 00:10:12,040 Speaker 1: is the main example here where most people know in 155 00:10:12,200 --> 00:10:17,560 Speaker 1: Vidia as a graphics computer chip manufacturer that largely caters 156 00:10:17,600 --> 00:10:23,120 Speaker 1: to gamers, but in Nvidia has really embraced making chips 157 00:10:23,160 --> 00:10:30,920 Speaker 1: specifically designed to operate in AI implementations. And also companies 158 00:10:30,920 --> 00:10:33,200 Speaker 1: that provide a lot of cloud computing functions are really 159 00:10:33,240 --> 00:10:36,600 Speaker 1: getting into the act too. They're also stepping in because 160 00:10:36,640 --> 00:10:42,040 Speaker 1: it requires so much computing power to run advanced AI operations, 161 00:10:42,520 --> 00:10:44,400 Speaker 1: and so we're seeing a big spike in demand for 162 00:10:44,440 --> 00:10:48,040 Speaker 1: those types of tech solutions. I've got more to say 163 00:10:48,080 --> 00:10:51,440 Speaker 1: about AI in general and generative AI in particular in 164 00:10:51,480 --> 00:10:54,360 Speaker 1: the year twenty twenty three, but before we can get 165 00:10:54,400 --> 00:10:56,640 Speaker 1: to that, we're going to take a quick break to 166 00:10:56,720 --> 00:11:08,319 Speaker 1: thank our sponsors and we're back. So next up, I'd 167 00:11:08,360 --> 00:11:12,439 Speaker 1: like to talk about the various stories around AI and 168 00:11:12,679 --> 00:11:17,679 Speaker 1: perceived emergent capabilities. So that would mean cases where AI 169 00:11:17,800 --> 00:11:19,920 Speaker 1: seems to be able to do more than what it 170 00:11:20,000 --> 00:11:23,560 Speaker 1: was designed to do, like the idea that AI is 171 00:11:23,600 --> 00:11:28,840 Speaker 1: somehow learning or teaching itself things that should be beyond 172 00:11:28,880 --> 00:11:32,679 Speaker 1: its capabilities. We've heard some folks express concern that AI 173 00:11:32,840 --> 00:11:35,520 Speaker 1: is maybe smarter than we think it is and that 174 00:11:35,559 --> 00:11:38,240 Speaker 1: this is going to lead to catastrophe. But we've also 175 00:11:38,280 --> 00:11:41,760 Speaker 1: seen studies that say these concerns are based on faulty premises. 176 00:11:42,200 --> 00:11:45,200 Speaker 1: That early studies used a set of metrics that gave 177 00:11:45,280 --> 00:11:48,880 Speaker 1: us inaccurate pictures of what AI isn't able to do. 178 00:11:49,280 --> 00:11:52,120 Speaker 1: That because the metrics were designed a specific way, it 179 00:11:52,160 --> 00:11:55,920 Speaker 1: was almost like cherry picking your evidence. It was finding 180 00:11:55,960 --> 00:11:59,880 Speaker 1: things that seemed to support a particular hypothesis and ignoring 181 00:12:00,080 --> 00:12:03,840 Speaker 1: things that were refuting that hypothesis, and that when you 182 00:12:03,880 --> 00:12:08,120 Speaker 1: adjusted those metrics and you did the study again, those 183 00:12:08,240 --> 00:12:11,120 Speaker 1: emergent behaviors turned out to be nothing of the sort. 184 00:12:11,679 --> 00:12:14,800 Speaker 1: It sounds like, at least for the moment, we're not 185 00:12:14,960 --> 00:12:18,160 Speaker 1: having some sort of sky net situation. However, as we 186 00:12:18,240 --> 00:12:21,559 Speaker 1: close out twenty twenty three, right now, there are news 187 00:12:21,600 --> 00:12:26,360 Speaker 1: stories about self recursive AI models, that is, tools that 188 00:12:26,440 --> 00:12:31,319 Speaker 1: can make changes to and in theory, improvements to themselves 189 00:12:31,679 --> 00:12:35,840 Speaker 1: over time. The science fiction standard of an AI that 190 00:12:35,920 --> 00:12:39,719 Speaker 1: improves itself in cycles that increase in frequency and then 191 00:12:39,800 --> 00:12:42,120 Speaker 1: diminish in the amount of time it takes to do 192 00:12:42,200 --> 00:12:45,520 Speaker 1: them is one that comes to mind. Right if you've 193 00:12:45,520 --> 00:12:48,840 Speaker 1: got an AI that's able to improve itself and presumably 194 00:12:48,880 --> 00:12:53,160 Speaker 1: do so at a level that's at least comparable to humans, 195 00:12:53,520 --> 00:12:57,160 Speaker 1: if not better than what humans can do, then you 196 00:12:57,160 --> 00:13:00,559 Speaker 1: could get into this situation where it's making these changes 197 00:13:00,600 --> 00:13:03,520 Speaker 1: and improvements in cycles that are happening faster and faster, 198 00:13:04,040 --> 00:13:06,760 Speaker 1: and you have a runaway train on your hands. These 199 00:13:06,800 --> 00:13:10,120 Speaker 1: are the scenarios that come to mind when people cite 200 00:13:10,240 --> 00:13:13,720 Speaker 1: things like the tech singularity or perhaps even a potential 201 00:13:13,800 --> 00:13:19,520 Speaker 1: existential crisis for humanity. I think it's still largely science fiction. 202 00:13:19,880 --> 00:13:22,160 Speaker 1: I don't think it's something that we need to necessarily 203 00:13:22,920 --> 00:13:27,560 Speaker 1: concern ourselves about in real time. But it is one 204 00:13:27,559 --> 00:13:31,800 Speaker 1: of those things that reinforces this fear, uncertainty, and doubt 205 00:13:32,000 --> 00:13:37,280 Speaker 1: or fud about artificial intelligence. Now for a few specific 206 00:13:37,320 --> 00:13:41,120 Speaker 1: stories that happened throughout the year, Judge Beryl Howell ruled 207 00:13:41,200 --> 00:13:44,439 Speaker 1: over the summer that AI created works are not eligible 208 00:13:44,480 --> 00:13:48,800 Speaker 1: for copyright. Judge Howell determine only works from human authors 209 00:13:49,200 --> 00:13:53,200 Speaker 1: can be copyrighted, which is huge, right, because if you're 210 00:13:53,280 --> 00:13:56,559 Speaker 1: using AI to generate all your content but you can't 211 00:13:56,559 --> 00:14:00,440 Speaker 1: copyright that content, you might not be in as strong 212 00:14:00,480 --> 00:14:03,120 Speaker 1: a situation as you think you are. One the content 213 00:14:03,240 --> 00:14:05,480 Speaker 1: may not be very good, and two you have no 214 00:14:05,600 --> 00:14:07,560 Speaker 1: ownership over it. Right, you have no way to protect 215 00:14:07,600 --> 00:14:10,679 Speaker 1: yourself if someone just lifts your content and uses it 216 00:14:10,720 --> 00:14:13,800 Speaker 1: somewhere else because you cannot copyright it do to the 217 00:14:13,800 --> 00:14:16,439 Speaker 1: fact that it was not a work from human authorship. 218 00:14:16,880 --> 00:14:19,920 Speaker 1: Over in California, a group of artists brought a lawsuit 219 00:14:19,920 --> 00:14:24,000 Speaker 1: against the company's mid Journey, Stability AI, and devant Art. 220 00:14:24,720 --> 00:14:27,800 Speaker 1: They were making the case that these companies misused the 221 00:14:27,920 --> 00:14:32,040 Speaker 1: artist's own copyrighted works while they were training up their 222 00:14:32,160 --> 00:14:35,400 Speaker 1: own AI models, and the judge in that case dismiss 223 00:14:35,520 --> 00:14:38,640 Speaker 1: charges against both mid Journey and Deviant Art because they 224 00:14:38,640 --> 00:14:42,960 Speaker 1: were using tools made by the third defendant in the case, 225 00:14:43,040 --> 00:14:47,080 Speaker 1: Stability AI. The judge did indicate that the plaintiffs could 226 00:14:47,080 --> 00:14:50,480 Speaker 1: file an amended complaint and include the other two companies 227 00:14:50,560 --> 00:14:54,040 Speaker 1: if they amended the complaint so that it was relevant. 228 00:14:54,680 --> 00:14:57,080 Speaker 1: And out of the three artists who were part of 229 00:14:57,120 --> 00:15:01,120 Speaker 1: the lawsuit originally, only one had her claims really make 230 00:15:01,200 --> 00:15:04,040 Speaker 1: it out of all the dismissals, because it turns out 231 00:15:04,080 --> 00:15:08,360 Speaker 1: the other two had not copyrighted their works. The copyright 232 00:15:08,440 --> 00:15:11,040 Speaker 1: infringement only works against the person who did take the 233 00:15:11,120 --> 00:15:15,280 Speaker 1: time to copyright their works. It remains unclear how the 234 00:15:15,320 --> 00:15:18,280 Speaker 1: court will rule if training a generative AI model on 235 00:15:18,360 --> 00:15:21,880 Speaker 1: an artist's work without their permission amounts to copyright infringement, 236 00:15:21,920 --> 00:15:23,680 Speaker 1: but we'll have to keep our eyes on that. In 237 00:15:23,720 --> 00:15:27,800 Speaker 1: the following year, and Open Ai was in the news 238 00:15:27,840 --> 00:15:30,720 Speaker 1: an awful lot. This year, the company unveiled GPT four, 239 00:15:30,880 --> 00:15:33,560 Speaker 1: which is the latest version of their large language model. 240 00:15:34,000 --> 00:15:37,600 Speaker 1: They started taking on enterprise clients companies that want to 241 00:15:37,640 --> 00:15:40,440 Speaker 1: tap into the power of that language model to do 242 00:15:40,880 --> 00:15:45,280 Speaker 1: various things. Chat gpt got access to current events. That 243 00:15:45,440 --> 00:15:47,760 Speaker 1: was a big deal. When it first launched, chat gpt 244 00:15:48,000 --> 00:15:52,880 Speaker 1: could not access any information that came after September twenty 245 00:15:52,960 --> 00:15:56,920 Speaker 1: twenty one. That's as far up as it could access info. However, 246 00:15:57,080 --> 00:15:59,840 Speaker 1: now it has access to the Internet, so it can 247 00:16:00,000 --> 00:16:04,280 Speaker 1: whole from current events. Another big ongoing story was how 248 00:16:04,320 --> 00:16:08,320 Speaker 1: open AI's CEO, which was Sam Altman, for all but 249 00:16:08,440 --> 00:16:10,680 Speaker 1: a couple of days this year. More on that in 250 00:16:10,840 --> 00:16:14,440 Speaker 1: just a second. He met with various leaders and regulators 251 00:16:14,480 --> 00:16:16,800 Speaker 1: around the world, and the purpose of those meetings was 252 00:16:16,840 --> 00:16:21,440 Speaker 1: to discuss potential regulations for AI, because obviously a lot 253 00:16:21,480 --> 00:16:25,560 Speaker 1: of legislators have concerns about artificial intelligence, So how can 254 00:16:25,600 --> 00:16:29,200 Speaker 1: we allow for the continuation of development so that, say, 255 00:16:29,280 --> 00:16:33,360 Speaker 1: the United States doesn't fall behind other countries while also 256 00:16:33,480 --> 00:16:38,120 Speaker 1: preventing potential disaster. Now, clearly Altman has a vested interest 257 00:16:38,160 --> 00:16:40,840 Speaker 1: in the outcome of these discussions, and in fact, some 258 00:16:41,040 --> 00:16:46,120 Speaker 1: critics worried that Altman's suggestions were really calculated to just 259 00:16:46,160 --> 00:16:49,240 Speaker 1: make it harder for smaller AI startups to catch up 260 00:16:49,280 --> 00:16:52,440 Speaker 1: to open ai and thus give the leader in the 261 00:16:52,480 --> 00:16:57,320 Speaker 1: field even more advantages. And Altman, according to these critics, 262 00:16:57,400 --> 00:17:00,400 Speaker 1: wasn't trying to make ai safer, but he would trying 263 00:17:00,440 --> 00:17:03,920 Speaker 1: to slow down the competition. This brings us to the 264 00:17:04,000 --> 00:17:08,240 Speaker 1: massive story of Sam Altman being unceremoniously fired by the 265 00:17:08,240 --> 00:17:10,920 Speaker 1: board of directors, only to be welcomed back to the 266 00:17:10,960 --> 00:17:16,400 Speaker 1: company literally days later. It's been quite the roller coaster ride. 267 00:17:16,480 --> 00:17:20,640 Speaker 1: So Altman had appeared at open AI's very first developers conference. 268 00:17:20,800 --> 00:17:23,919 Speaker 1: He had made several high profile announcements about the direction 269 00:17:24,040 --> 00:17:27,560 Speaker 1: and future of open AI's products, like its large language 270 00:17:27,560 --> 00:17:32,000 Speaker 1: models and its chatbot. And then not long after he 271 00:17:32,160 --> 00:17:35,960 Speaker 1: finished up at this developers conference, he gets a message 272 00:17:36,040 --> 00:17:38,240 Speaker 1: and he has to attend a zoom call with the 273 00:17:38,240 --> 00:17:40,560 Speaker 1: Board of directors, and that's what he finds out. Bam, 274 00:17:40,920 --> 00:17:44,800 Speaker 1: He's been fired. Now, that last story has a lot 275 00:17:44,840 --> 00:17:47,439 Speaker 1: to do with the gap between the original vision of 276 00:17:47,440 --> 00:17:50,159 Speaker 1: what open ai was supposed to be and what it 277 00:17:50,320 --> 00:17:53,120 Speaker 1: actually has become now. As I said at the beginning 278 00:17:53,160 --> 00:17:57,280 Speaker 1: of this episode, open ai started off as a nonprofit 279 00:17:57,440 --> 00:18:03,120 Speaker 1: organization dedicated to developing useful, ethical, and safe artificial intelligence, 280 00:18:03,480 --> 00:18:07,919 Speaker 1: but AI is really expensive, and in an effort to 281 00:18:08,119 --> 00:18:12,280 Speaker 1: fund the operation and not just constantly be begging for 282 00:18:12,359 --> 00:18:17,040 Speaker 1: funding from various parties, Sam Altman created a for profit 283 00:18:17,240 --> 00:18:20,800 Speaker 1: arm of open Ai, and since then the company has 284 00:18:20,800 --> 00:18:24,560 Speaker 1: made some very aggressive moves in the artificial intelligence space, 285 00:18:24,760 --> 00:18:28,639 Speaker 1: sometimes with Altman issuing statements that made it seem like 286 00:18:28,680 --> 00:18:32,280 Speaker 1: even he thought it might be a bit much and 287 00:18:32,359 --> 00:18:36,120 Speaker 1: a bit too aggressive. And yet the aggressive moves kept 288 00:18:36,119 --> 00:18:38,320 Speaker 1: on coming, and it reached a point where the board 289 00:18:38,320 --> 00:18:41,359 Speaker 1: of directors, who were people who were originally part of 290 00:18:41,359 --> 00:18:45,439 Speaker 1: the nonprofit open Ai version, were really concerned enough to 291 00:18:45,720 --> 00:18:49,520 Speaker 1: relieve Altman of his job. But the backlash following that 292 00:18:49,560 --> 00:18:53,040 Speaker 1: move prompted a near total shakedown of the board and 293 00:18:53,119 --> 00:18:56,800 Speaker 1: Altman is back in the driver's seat because whether the 294 00:18:56,840 --> 00:19:00,560 Speaker 1: concerns are relevant or not, you had parties like Microsoft, 295 00:19:00,640 --> 00:19:04,880 Speaker 1: which has dedicated ten billion dollars in investments to open 296 00:19:04,920 --> 00:19:10,359 Speaker 1: ai over the near future, and without consulting Microsoft first 297 00:19:10,359 --> 00:19:14,680 Speaker 1: before firing the CEO, it really shook out the Apple carts. 298 00:19:14,680 --> 00:19:20,560 Speaker 1: So we see that commerce can overpower concern right. I 299 00:19:20,560 --> 00:19:23,240 Speaker 1: think it's safe to say that every year from here 300 00:19:23,320 --> 00:19:25,920 Speaker 1: on out is going to be AI's year for good 301 00:19:25,960 --> 00:19:30,360 Speaker 1: and for bad, and twenty twenty three certainly qualified. AI 302 00:19:30,359 --> 00:19:32,560 Speaker 1: was part of stories that were even outside the world 303 00:19:32,560 --> 00:19:36,280 Speaker 1: of technology. Played a part in Hollywood negotiations, as both 304 00:19:36,320 --> 00:19:40,560 Speaker 1: the WGA and sag AFTRA when on strike. Both unions 305 00:19:40,600 --> 00:19:44,480 Speaker 1: express concerns about AI's role and entertainment moving forward. So 306 00:19:44,560 --> 00:19:47,359 Speaker 1: I expect we'll see lots more stories in that vein 307 00:19:47,760 --> 00:19:51,800 Speaker 1: as we move forward. But that's an overview of AI 308 00:19:51,880 --> 00:19:55,080 Speaker 1: and generative AI in twenty twenty three. We'll be back 309 00:19:55,119 --> 00:19:59,480 Speaker 1: with more short episodes about big tech news stories throughout 310 00:19:59,480 --> 00:20:02,040 Speaker 1: the year over the next few days. I hope you're 311 00:20:02,119 --> 00:20:05,880 Speaker 1: all well, and I'll talk to you again really soon. 312 00:20:12,440 --> 00:20:17,080 Speaker 1: Tech Stuff is an iHeartRadio production. For more podcasts from iHeartRadio, 313 00:20:17,400 --> 00:20:21,080 Speaker 1: visit the iHeartRadio app, Apple Podcasts, or wherever you listen 314 00:20:21,160 --> 00:20:22,240 Speaker 1: to your favorite shows.