{"version": "1.0.0", "segments": [{"startTime": 4.59, "endTime": 12.619, "body": "It's episode 58 of the Digitalization TikToks podcast, and your host Don Mack and myself, Jonas Narender are back behind the microphones.", "speaker": "Jonas Norinder"}, {"startTime": 13.229, "endTime": 22.829, "body": "Before we dive into today's conversation, I want to give you a thank you to all of our listeners who've been sharing the show and sending us feedback and suggesting guests.", "speaker": "Jonas Norinder"}, {"startTime": 22.95, "endTime": 26.569, "body": "We've been getting quite a few of those lately, so keep them coming.", "speaker": "Jonas Norinder"}, {"startTime": 26.95, "endTime": 34.409, "body": "Your support continues to fuel these conversations, and it also helps us bring on incredible guests like the one we have on today.", "speaker": "Jonas Norinder"}, {"startTime": 35.4, "endTime": 40.34, "body": "As you know, as we're all aware, AI is everywhere, but so are the myths surrounding it.", "speaker": "Jonas Norinder"}, {"startTime": 40.509, "endTime": 47.04, "body": "Like there are some belief that AI is always objective to fears that it's coming for everyone's jobs.", "speaker": "Jonas Norinder"}, {"startTime": 47.639, "endTime": 52.15, "body": "Today we are tackling seven of the most common misconceptions head on.", "speaker": "Jonas Norinder"}, {"startTime": 52.599, "endTime": 58.88, "body": "Ron brings a wealth of experience in AI strategy and implementation, and he's here to separate fact from fiction.", "speaker": "Jonas Norinder"}, {"startTime": 59.709, "endTime": 71.709, "body": "We'll explore why AI isn't as unbiased as people may think, what it can and can't do compared to us humans, and why bigger models don't always mean better results.", "speaker": "Jonas Norinder"}, {"startTime": 72.11, "endTime": 83.73, "body": "Plus we'll also talk about the myth that AI is only for big tech and how startups and smaller organizations are making big waves with the help of open source tools and community driven innovation.", "speaker": "Jonas Norinder"}, {"startTime": 83.98, "endTime": 86.41, "body": "So quite a few things to unpack here today.", "speaker": "Jonas Norinder"}, {"startTime": 87.319, "endTime": 96.22, "body": "The episode is really all about demystifying AI and giving you a clearer, more grounded view of where this technology is today and where it's headed.", "speaker": "Jonas Norinder"}, {"startTime": 96.76, "endTime": 99.62, "body": "So Dan, are you ready to bust some AI myths today?", "speaker": "Jonas Norinder"}, {"startTime": 100.54, "endTime": 101.29, "body": "Hi, Enas.", "speaker": "Speaker 2"}, {"startTime": 101.379, "endTime": 108.419, "body": "Well, we're definitely not the first ones out there to be busting myths, but I think it's worth giving it a shot and we've got a great topic to, to bust some myths.", "speaker": "Speaker 2"}, {"startTime": 108.5, "endTime": 109.36, "body": "So let's do it.", "speaker": "Speaker 2"}, {"startTime": 109.739, "endTime": 112.879, "body": "Uh, the name of Ron's company, Kung Fu AI is interesting.", "speaker": "Speaker 2"}, {"startTime": 113.18, "endTime": 118.419, "body": "So perhaps Ron can touch a little bit on the origin and maybe even tell us a little bit about who Kai is.", "speaker": "Speaker 2"}, {"startTime": 119.239, "endTime": 125.989, "body": "As my general lack of flexibility will reveal, I'm anything but a specialist in kung fu or any kind of martial arts.", "speaker": "Speaker 2"}, {"startTime": 126.36, "endTime": 139.919, "body": "But when I think about martial arts, it seems that there are some parallels to AI such as things like lots of training and refinement over time, a goal of eventually achieving mastery, and being able to adapt to new situations or new data.", "speaker": "Speaker 2"}, {"startTime": 140.539, "endTime": 146.74, "body": "So it seems like we're connecting with a good resource to discuss AI and clear up some of these misconceptions that you've mentioned.", "speaker": "Speaker 2"}, {"startTime": 147.02, "endTime": 149.66, "body": "I'm very much looking forward to our discussion today with Ron.", "speaker": "Speaker 2"}, {"startTime": 151.429, "endTime": 151.94, "body": "Excellent.", "speaker": "Jonas Norinder"}, {"startTime": 152.039, "endTime": 152.96, "body": "Thank you for that, Don.", "speaker": "Jonas Norinder"}, {"startTime": 153.039, "endTime": 155.119, "body": "Yes, I am also looking forward to it a lot.", "speaker": "Jonas Norinder"}, {"startTime": 155.199, "endTime": 156.46, "body": "So let's get to it.", "speaker": "Jonas Norinder"}, {"startTime": 156.52, "endTime": 158.6, "body": "I want to introduce Ron.", "speaker": "Jonas Norinder"}, {"startTime": 158.679, "endTime": 166.339, "body": "We're very excited to have you on as our guest today, Ron, you are co-founder and CTO at Kung Fu AI.", "speaker": "Jonas Norinder"}, {"startTime": 167.039, "endTime": 173.309, "body": "Ron is a seasoned AI expert and serial and technology entrepreneur who's helped launch multiple success startups.", "speaker": "Jonas Norinder"}, {"startTime": 173.699, "endTime": 178.63, "body": "And at Kung Fu AI he's all about making AI practical, ethical and impactful.", "speaker": "Jonas Norinder"}, {"startTime": 178.96, "endTime": 180.32, "body": "So Ron, welcome to the", "speaker": "Jonas Norinder"}, {"startTime": 180.32, "endTime": 180.6, "body": "show.", "speaker": "Speaker 3"}, {"startTime": 181.47, "endTime": 182.869, "body": "Thank you so much guys for having me on.", "speaker": "Speaker 3"}, {"startTime": 182.919, "endTime": 184.82, "body": "I'm, I'm really excited about this conversation.", "speaker": "Speaker 3"}, {"startTime": 185.899, "endTime": 188.22, "body": "Well, thanks, Ron, and let's go ahead and jump right in.", "speaker": "Speaker 2"}, {"startTime": 188.259, "endTime": 191.44, "body": "We have 7 misconceptions to cover today, so let's get right to them.", "speaker": "Speaker 2"}, {"startTime": 191.699, "endTime": 197.0, "body": "The first one we'd like to talk about is AI and its objectivity and being unbiased.", "speaker": "Speaker 2"}, {"startTime": 197.179, "endTime": 197.88, "body": "Is that true?", "speaker": "Speaker 2"}, {"startTime": 198.99, "endTime": 199.949, "body": "It's really not.", "speaker": "Speaker 3"}, {"startTime": 200.029, "endTime": 213.419, "body": "I mean, AI systems are getting more and more powerful, but you know, I think, I think most people are probably aware of this now, that they're trained on an enormous amount of data, and it's very true with most computer systems, garbage in, garbage out.", "speaker": "Speaker 3"}, {"startTime": 213.49, "endTime": 217.529, "body": "That's never been more true than with artificial intelligence.", "speaker": "Speaker 3"}, {"startTime": 217.789, "endTime": 221.41, "body": "The main technique we use these days is called supervised learning.", "speaker": "Speaker 3"}, {"startTime": 222.02, "endTime": 233.74, "body": "It's a machine learning technique, and the challenge there is that if you have a data set that is biased or unfair in some way, the models will learn those patterns.", "speaker": "Speaker 3"}, {"startTime": 233.779, "endTime": 234.49, "body": "They'll pick it up.", "speaker": "Speaker 3"}, {"startTime": 234.699, "endTime": 235.619, "body": "They're like sponges.", "speaker": "Speaker 3"}, {"startTime": 235.82, "endTime": 237.94, "body": "They, they will soak it up and they will replicate that.", "speaker": "Speaker 3"}, {"startTime": 238.039, "endTime": 248.6, "body": "And so when we're working with clients, uh, this is one of the first things we talk about because It can be bad from a predictive perspective.", "speaker": "Speaker 3"}, {"startTime": 248.88, "endTime": 254.0, "body": "You can have less generalizable results, but it can also have really major ethical issues.", "speaker": "Speaker 3"}, {"startTime": 254.11, "endTime": 277.484, "body": "So for example, we've built systems that were designed to make healthcare risk decisions or within the financial sectors, and we Spent, I mean, I, I, I would say, you know, maybe half of the work is vetting the data and removing bias and making sure that the AI system is not being fed data that would skew it one way or the other.", "speaker": "Speaker 3"}, {"startTime": 277.595, "endTime": 292.204, "body": "And the, the challenge is, if you don't do this technique and you think you've built an AI system that's objective, uh, it's not, and you will just be perpetrating, you know, bad behavior, um, and without even realizing it.", "speaker": "Speaker 3"}, {"startTime": 293.39, "endTime": 303.57, "body": "So it sounds like you really have to be very thoughtful about what you put in on the front end to, to make sure that you don't create that bias of even unwillingly.", "speaker": "Jonas Norinder"}, {"startTime": 303.989, "endTime": 304.91, "body": "Yes, you really do.", "speaker": "Speaker 3"}, {"startTime": 304.97, "endTime": 324.609, "body": "And, and just to put a fine point on that, for example, if you're dealing with, um, We built a system to do real estate pricing and we were looking at decades of data, but there were clearly racial biases within that data and so we spent a lot of time removing that so the historical legacies weren't perpetrated by, you know, modern AI systems.", "speaker": "Speaker 3"}, {"startTime": 325.97, "endTime": 333.429, "body": "Now, let's get to misconception number 2, and that is that AI can fully think and understand like a human.", "speaker": "Jonas Norinder"}, {"startTime": 333.5, "endTime": 335.92, "body": "It sounds great and beautiful if that's the case.", "speaker": "Jonas Norinder"}, {"startTime": 336.13, "endTime": 338.17, "body": "So, Ron, what's your thought on that?", "speaker": "Jonas Norinder"}, {"startTime": 339.079, "endTime": 341.48, "body": "Yeah, we are definitely not there yet.", "speaker": "Speaker 3"}, {"startTime": 342.16, "endTime": 352.94, "body": "There have been major strides, major improvements made even within the last 12 months, and, you know, quote, reasoning models or thinking models.", "speaker": "Speaker 3"}, {"startTime": 353.589, "endTime": 355.16, "body": "Truly are getting better.", "speaker": "Speaker 3"}, {"startTime": 355.279, "endTime": 375.48, "body": "They truly are more and more able to reflect upon their own output and use that to identify errors, use that to essentially identify dead ends when solving problems and backtrack, but they do not reason or think in a way that humans do.", "speaker": "Speaker 3"}, {"startTime": 375.519, "endTime": 376.079, "body": "At this point.", "speaker": "Speaker 3"}, {"startTime": 376.16, "endTime": 378.019, "body": "It is very, very much.", "speaker": "Speaker 3"}, {"startTime": 378.47, "endTime": 383.429, "body": "Um, still a very little understood world.", "speaker": "Speaker 3"}, {"startTime": 383.519, "endTime": 385.54, "body": "This is called model interpretability.", "speaker": "Speaker 3"}, {"startTime": 386.2, "endTime": 398.399, "body": "And we are at the very, very early stages of actually understanding how these systems think and reason, but it is very different than how, uh, uh, humans reason and think.", "speaker": "Speaker 3"}, {"startTime": 398.47, "endTime": 407.88, "body": "And so don't, you know, when you're interacting with these models and you, you, you get a result, it's, you know, they can be very convincing, but I would advise, you know, everybody.", "speaker": "Speaker 3"}, {"startTime": 408.5, "endTime": 420.519, "body": "To apply your own human reason and analysis to the output because we, we're dealing with systems that are getting powerful, more powerful every year, but, but they are very different than humans still.", "speaker": "Speaker 3"}, {"startTime": 421.329, "endTime": 429.709, "body": "So, so Ron, I, I think that begs the follow-up question, are we ever going to get to a point where AI can fully think and understand like a human?", "speaker": "Jonas Norinder"}, {"startTime": 430.429, "endTime": 431.109, "body": "I, I do.", "speaker": "Speaker 3"}, {"startTime": 431.23, "endTime": 433.309, "body": "I, I believe that very, very strongly.", "speaker": "Speaker 3"}, {"startTime": 433.459, "endTime": 439.67, "body": "Everything we're seeing, um, indicates that we are, um, making progress in that direction.", "speaker": "Speaker 3"}, {"startTime": 439.75, "endTime": 442.48, "body": "In fact, I would, I would actually go further.", "speaker": "Speaker 3"}, {"startTime": 442.75, "endTime": 469.63, "body": "Um, I think that we will not only achieve AGI, artificial general intelligence, by the year 2030, and that is, and AGI is basically This idea that these, these models can perform pretty much all human level reasoning tasks, but I think we're going to get to ASI, artificial superintelligence, within our lifetimes, and those models will be able to reason at a level that is beyond human capabilities.", "speaker": "Speaker 3"}, {"startTime": 470.779, "endTime": 472.01, "body": "Well, that's really interesting, Ron.", "speaker": "Speaker 2"}, {"startTime": 472.059, "endTime": 478.98, "body": "It's, it makes me think that we're gonna have to periodically touch base with you, uh, as we move to the future here since some of the misconceptions will be changing.", "speaker": "Speaker 2"}, {"startTime": 479.1, "endTime": 480.91, "body": "So interesting to see how things evolve.", "speaker": "Speaker 2"}, {"startTime": 481.7, "endTime": 481.809, "body": "Yes.", "speaker": "Speaker 3"}, {"startTime": 482.779, "endTime": 489.98, "body": "All right, well, let's go on to our next one, and, and this pertains to, to jobs, and I, I think really is one of the largest areas of concern for people.", "speaker": "Speaker 2"}, {"startTime": 490.459, "endTime": 492.73, "body": "AI will inevitably take my job.", "speaker": "Speaker 2"}, {"startTime": 493.1, "endTime": 494.14, "body": "What are your thoughts on that?", "speaker": "Speaker 2"}, {"startTime": 494.929, "endTime": 508.989, "body": "Yeah, I, this is a really, I, I think, complicated and nuanced discussion, because I, um, it really depends upon, I think, understanding the time frames and how things will change.", "speaker": "Speaker 3"}, {"startTime": 509.529, "endTime": 517.789, "body": "Candidly, you know, I like to joke, ultimately, AI is going to take all of our jobs, but that might be in 100 years, OK.", "speaker": "Speaker 3"}, {"startTime": 518.219, "endTime": 543.19, "body": "And by that time, I'm very, very optimistic that as a part of that process, we will be unlocking our exploration of science and improving engineering and healthcare in so many dramatic ways through these AI powered systems that it will, it will be a liberating event and not this, this idea that, that, you know, humanity will be unemployed and desolate.", "speaker": "Speaker 3"}, {"startTime": 543.729, "endTime": 546.609, "body": "So, in the short term, it's going to be very sporadic.", "speaker": "Speaker 3"}, {"startTime": 546.76, "endTime": 555.729, "body": "In like any, like any disruptive technology, as these AI systems become more powerful, they're really fractal on the, on the leading edge.", "speaker": "Speaker 3"}, {"startTime": 555.77, "endTime": 563.809, "body": "And what that means is they have a capability that sometimes can be amazingly powerful and sometimes can be really silly and stupid.", "speaker": "Speaker 3"}, {"startTime": 563.849, "endTime": 569.59, "body": "If, if you've interacted with language models, you'll see every now and then they'll make just the, the silliest mistakes.", "speaker": "Speaker 3"}, {"startTime": 569.859, "endTime": 573.19, "body": "In some ways it's analogous to the way humans work.", "speaker": "Speaker 3"}, {"startTime": 573.299, "endTime": 578.65, "body": "We can be so brilliant, and then sometimes we can misspeak or make the silliest errors.", "speaker": "Speaker 3"}, {"startTime": 578.979, "endTime": 591.82, "body": "So over time, as these models become more powerful, as they are able to reason and think like humans, like we just discussed, They are going to be leveraged to augment human jobs.", "speaker": "Speaker 3"}, {"startTime": 591.95, "endTime": 599.07, "body": "So if you are a scientist, you're gonna have AI enabled assistants that are gonna, um, help you do your job.", "speaker": "Speaker 3"}, {"startTime": 599.14, "endTime": 602.07, "body": "And I think we're going to see that across all the fields.", "speaker": "Speaker 3"}, {"startTime": 602.15, "endTime": 610.13, "body": "Interestingly, Um, if you're in software, you know, my field, it's probably being the most disruptive of any field so far.", "speaker": "Speaker 3"}, {"startTime": 610.429, "endTime": 612.349, "body": "Uh, AI coding assistants are everywhere.", "speaker": "Speaker 3"}, {"startTime": 612.429, "endTime": 613.59, "body": "We use them every day.", "speaker": "Speaker 3"}, {"startTime": 613.71, "endTime": 618.07, "body": "They're, I, I think they're twice as powerful now as they were 6 months ago.", "speaker": "Speaker 3"}, {"startTime": 618.45, "endTime": 632.599, "body": "So this is going to happen pretty much in every domain, but I would encourage people to view it not as job removal, but as job assistance and job empowerment through AI.", "speaker": "Speaker 3"}, {"startTime": 632.679, "endTime": 637.309, "body": "And so, for example, We're going to have AI powered legal assistance.", "speaker": "Speaker 3"}, {"startTime": 637.46, "endTime": 640.59, "body": "Think, think about how overwhelmed our legal system is, right?", "speaker": "Speaker 3"}, {"startTime": 640.71, "endTime": 658.429, "body": "And if, if you could have an, an AI legal assistant that could instantaneously go review all relevant case law into months' worth of work in just seconds, that's going to help people get a more fair trial, to me, things like that only make the world a better place.", "speaker": "Speaker 3"}, {"startTime": 659.789, "endTime": 668.59, "body": "And, and I do agree with you on that, Ron, on the, on the side or your, your comment about AI being more of an assistant right now.", "speaker": "Jonas Norinder"}, {"startTime": 668.669, "endTime": 672.659, "body": "I mean, we really all should be looking at ways that AI can help us.", "speaker": "Jonas Norinder"}, {"startTime": 673.15, "endTime": 675.39, "body": "Because if we don't, then somebody else is going to do it.", "speaker": "Jonas Norinder"}, {"startTime": 675.44, "endTime": 677.229, "body": "So, I think that's a really, really good point.", "speaker": "Jonas Norinder"}, {"startTime": 677.549, "endTime": 683.84, "body": "I think it might make our listeners pretty happy to hear too that, uh, you know, we won't be replaced maybe 100 years from now.", "speaker": "Jonas Norinder"}, {"startTime": 683.95, "endTime": 688.469, "body": "Um, so maybe we don't have to be that worried about that in the short term.", "speaker": "Jonas Norinder"}, {"startTime": 689.659, "endTime": 692.9, "body": "The 4th misconception is an interesting one.", "speaker": "Jonas Norinder"}, {"startTime": 692.94, "endTime": 697.2, "body": "I think AI systems are always accurate and reliable.", "speaker": "Jonas Norinder"}, {"startTime": 697.859, "endTime": 718.78, "body": "Yeah, this is, this is something that, uh, still surprises me because with, with everybody using, uh, you know, chatbots on a daily basis, you know, you would think that people would see the mistakes that AI models make, but every now and then I, I encounter people that, that just take it face value and, and, and believe anything, these, these models.", "speaker": "Speaker 3"}, {"startTime": 719.239, "endTime": 722.69, "body": "Uh, say, or produce, and it's, and it's, you need to be careful there.", "speaker": "Speaker 3"}, {"startTime": 722.77, "endTime": 725.2, "body": "These models can fail, they can produce errors.", "speaker": "Speaker 3"}, {"startTime": 725.289, "endTime": 727.03, "body": "They are not omniscient.", "speaker": "Speaker 3"}, {"startTime": 727.77, "endTime": 735.159, "body": "And, and they are especially prone to error, um, when faced with data that's out of their training distribution.", "speaker": "Speaker 3"}, {"startTime": 735.25, "endTime": 745.15, "body": "And what that means is if you put them into scenarios where they have not seen enough training data to have generalized in a strong way.", "speaker": "Speaker 3"}, {"startTime": 745.369, "endTime": 754.239, "body": "And so, Uh, I, I don't, I don't think this is too hard to get your head around once you just understand that like humans, these AI systems, uh, will make mistakes.", "speaker": "Speaker 3"}, {"startTime": 754.28, "endTime": 761.559, "body": "And in fact, I actually believe that, that we will never have infallible, uh, systems, whether they're AI or otherwise.", "speaker": "Speaker 3"}, {"startTime": 761.64, "endTime": 764.179, "body": "I just don't think we live in that type of universe.", "speaker": "Speaker 3"}, {"startTime": 764.979, "endTime": 777.159, "body": "Yeah, maybe this goes with the whole, uh, thought about don't believe everything you see on the internet, and, uh, and I think with AI it, it comes across with a very confident answer a lot of times you just think it's right, but, uh,", "speaker": "Speaker 2"}, {"startTime": 777.21, "endTime": 778.28, "body": "that seems like so confident.", "speaker": "Speaker 3"}, {"startTime": 778.359, "endTime": 781.969, "body": "Yeah, sometimes it, it, it's just so compelling you, you wanna believe it,", "speaker": "Speaker 3"}, {"startTime": 782.44, "endTime": 783.299, "body": "right, right.", "speaker": "Speaker 2"}, {"startTime": 783.84, "endTime": 785.869, "body": "All right, well, let's go on to our next misconception.", "speaker": "Speaker 2"}, {"startTime": 785.96, "endTime": 790.09, "body": "I think we're up to number 5 now, and this guy has to do with the, the thought bigger is better.", "speaker": "Speaker 2"}, {"startTime": 790.4, "endTime": 794.599, "body": "If you think in terms of AI, is it true that the bigger the model, the better.", "speaker": "Speaker 2"}, {"startTime": 795.619, "endTime": 796.469, "body": "No, it's really not.", "speaker": "Speaker 3"}, {"startTime": 796.549, "endTime": 802.169, "body": "And this is, there's a little bit of a subtle point here that I, I need to make to be really clear.", "speaker": "Speaker 3"}, {"startTime": 802.789, "endTime": 822.989, "body": "Most models, frontier models, and we're talking about, you know, uh, anthropic cloud or chat GPT from OpenAI, these frontier models are enormous, and they have been trained on, you know, quote, all of the internet, and they have very, very, very diverse capability, and, and it's, they're incredible.", "speaker": "Speaker 3"}, {"startTime": 823.07, "endTime": 825.169, "body": "They're amazingly powerful.", "speaker": "Speaker 3"}, {"startTime": 825.95, "endTime": 832.799, "body": "But they are not capable of performing every task better than any model out there.", "speaker": "Speaker 3"}, {"startTime": 832.919, "endTime": 845.669, "body": "I would say, in fact, the vast majority of the way we help our clients is by building proprietary custom models that leverage their proprietary data.", "speaker": "Speaker 3"}, {"startTime": 845.809, "endTime": 853.9, "body": "And these models can often be orders of magnitude smaller than frontier models, which are now in the, you know, trillions of parameters, right?", "speaker": "Speaker 3"}, {"startTime": 854.179, "endTime": 863.609, "body": "We'll build models that sometimes can be as small as hundreds of millions of parameters, which, you know, just 10 years ago would have been enormous.", "speaker": "Speaker 3"}, {"startTime": 863.88, "endTime": 869.21, "body": "And these smaller models can have very, very sophisticated specialized capabilities.", "speaker": "Speaker 3"}, {"startTime": 869.289, "endTime": 870.229, "body": "So you can build.", "speaker": "Speaker 3"}, {"startTime": 870.89, "endTime": 890.159, "body": "You can take your own company data and use that to make failure predictions, or you can spot manufacturing defects, or you could automate financial decisions, or you could, we've built systems that can do cancer risk prediction 5 years in the future when in fact it just was approved by the FDA.", "speaker": "Speaker 3"}, {"startTime": 890.46, "endTime": 893.479, "body": "This is superhuman capability looking at.", "speaker": "Speaker 3"}, {"startTime": 893.789, "endTime": 898.239, "body": "Uh, mammograms and picking up on biological signs of cancer that humans can't see.", "speaker": "Speaker 3"}, {"startTime": 898.359, "endTime": 901.789, "body": "And these are much, much smaller models than these frontier models.", "speaker": "Speaker 3"}, {"startTime": 902.039, "endTime": 913.989, "body": "So you don't always have to go spend, you know, $10 billion to go embrace AI and have it actually give you proprietary, um, differentiated advantage within your own company.", "speaker": "Speaker 3"}, {"startTime": 915.78, "endTime": 920.14, "body": "Glad to hear that there still are cases where where small sometimes is better.", "speaker": "Jonas Norinder"}, {"startTime": 920.179, "endTime": 926.159, "body": "It doesn't always have to be big and elaborate to create value, and it sounds like with AI there was definitely something to that.", "speaker": "Jonas Norinder"}, {"startTime": 926.659, "endTime": 933.369, "body": "Now we're quickly getting closer to the end with misconception number 6, and that could potentially be an ugly one.", "speaker": "Jonas Norinder"}, {"startTime": 933.969, "endTime": 940.679, "body": "AI can easily explain its decision, and it's the emphasis on easily that makes me say that it could get ugly.", "speaker": "Jonas Norinder"}, {"startTime": 940.82, "endTime": 943.82, "body": "Now I don't know, Ron, if that's the case or not, so, so let's hear it.", "speaker": "Jonas Norinder"}, {"startTime": 943.9, "endTime": 945.479, "body": "What's your verdict from your perspective?", "speaker": "Jonas Norinder"}, {"startTime": 946.119, "endTime": 967.95, "body": "Yes, it is, it is absolutely true that these powerful AI systems can be hard to understand, meaning they can give you the right answer, and We can still not understand how it came to its answer, and this is especially true of larger models.", "speaker": "Speaker 3"}, {"startTime": 968.28, "endTime": 976.02, "body": "Um, and it's, it's almost always true of large, deep learning systems or artificial neural networks that are um.", "speaker": "Speaker 3"}, {"startTime": 977.0, "endTime": 982.679, "body": "Through supervised learning example or process I mentioned earlier where you train on a bunch of data.", "speaker": "Speaker 3"}, {"startTime": 982.799, "endTime": 989.739, "body": "These models are incredibly powerful, but their decision making is spread out across all of these parameters.", "speaker": "Speaker 3"}, {"startTime": 989.77, "endTime": 997.78, "body": "And even though the math is relatively simple, The outputs of these models are, they're, they're very much like black boxes.", "speaker": "Speaker 3"}, {"startTime": 998.09, "endTime": 1004.78, "body": "So, it can be a challenge, and there will be environments where, in fact, it makes them inappropriate choices.", "speaker": "Speaker 3"}, {"startTime": 1004.9, "endTime": 1022.429, "body": "So, for example, if you're building a system and you need to, in healthcare, let's say, and you are doing diagnoses or you're making clinical decisions, it is probably unacceptable for a model to tell you, you need surgery, but It being unable to tell you why.", "speaker": "Speaker 3"}, {"startTime": 1022.52, "endTime": 1034.16, "body": "And, and so in those environments, very often explainability and interpretability override the need for just raw predictive power, and there are other techniques that you typically have to lean on.", "speaker": "Speaker 3"}, {"startTime": 1034.359, "endTime": 1041.359, "body": "So it is definitely not the case that AI can easily explain its decisions or that they're easily interpretable.", "speaker": "Speaker 3"}, {"startTime": 1042.75, "endTime": 1043.319, "body": "Well, thanks, Ron.", "speaker": "Speaker 2"}, {"startTime": 1043.359, "endTime": 1056.099, "body": "It seems like we're on a, a, a theme of in the last few questions here or misconceptions about bigger isn't better, and I actually want to stick with that with our last, our 7th misconception here, and, and this pertains to who can accomplish AI development.", "speaker": "Speaker 2"}, {"startTime": 1056.67, "endTime": 1061.199, "body": "Uh, another possible misconception is that AI development is really for big tech companies only.", "speaker": "Speaker 2"}, {"startTime": 1061.52, "endTime": 1062.719, "body": "What's your opinion about this?", "speaker": "Speaker 2"}, {"startTime": 1063.579, "endTime": 1067.17, "body": "Yeah, that is, that is a huge misconception.", "speaker": "Speaker 3"}, {"startTime": 1067.9, "endTime": 1085.26, "body": "Every, pretty much every company, and I would almost go so far as to say almost all employees out there today are leveraging AI in one form or another, whether it's via, you know, chatbots or other, you know, summarization techniques and things like that.", "speaker": "Speaker 3"}, {"startTime": 1085.5, "endTime": 1087.17, "body": "But I'm really talking beyond that.", "speaker": "Speaker 3"}, {"startTime": 1087.229, "endTime": 1090.0, "body": "I'm talking about the ability to leverage AI.", "speaker": "Speaker 3"}, {"startTime": 1090.42, "endTime": 1102.13, "body": "In custom ways for your own business, for your own workflows, and that can range from as simple as integrating AI capabilities with your, within your own workflows.", "speaker": "Speaker 3"}, {"startTime": 1102.38, "endTime": 1110.099, "body": "There are incredibly powerful open source tools that you can now, uh, use to leverage perceptive capabilities.", "speaker": "Speaker 3"}, {"startTime": 1110.224, "endTime": 1119.785, "body": "That means you can understand unstructured data, you can understand images and, and you can read text and handwriting, you can understand images.", "speaker": "Speaker 3"}, {"startTime": 1119.905, "endTime": 1123.864, "body": "All of these capabilities are more and more available on open-source tools.", "speaker": "Speaker 3"}, {"startTime": 1124.104, "endTime": 1134.765, "body": "And if you need to push beyond that, there are really, really powerful APIs from all the hyperscalers out there that will allow you to, you know, for a small fee, go to the next level.", "speaker": "Speaker 3"}, {"startTime": 1135.229, "endTime": 1152.369, "body": "But even beyond that is every company out there is sitting on a gold mine of proprietary data, and it is absolutely feasible in this day and age to take that data and use it to train custom models that can be perceptive or.", "speaker": "Speaker 3"}, {"startTime": 1152.505, "endTime": 1168.604, "body": "Predictive, and then integrate those capabilities within your own business to give you either, um, some, some new feature capability for your end users, or to give you some differentiated capability that makes you more competitive.", "speaker": "Speaker 3"}, {"startTime": 1168.665, "endTime": 1172.185, "body": "And we've seen this over and over and over again.", "speaker": "Speaker 3"}, {"startTime": 1172.265, "endTime": 1178.564, "body": "Um, and really the only limitation to taking advantage of your own proprietary data to build.", "speaker": "Speaker 3"}, {"startTime": 1179.52, "endTime": 1188.01, "body": "Uh, custom AI systems is every now and then we'll, we'll encounter companies that have failed to persist the data.", "speaker": "Speaker 3"}, {"startTime": 1188.17, "endTime": 1190.68, "body": "So they are an expert in some domain.", "speaker": "Speaker 3"}, {"startTime": 1191.01, "endTime": 1193.89, "body": "They've been doing it for years or decades.", "speaker": "Speaker 3"}, {"startTime": 1194.089, "endTime": 1201.93, "body": "They want to automate it, and then we will engage with them, and it turns out they have not been keeping the data, the necessary data for training on it.", "speaker": "Speaker 3"}, {"startTime": 1201.959, "endTime": 1203.709, "body": "And so I always use the old adage.", "speaker": "Speaker 3"}, {"startTime": 1204.16, "endTime": 1206.839, "body": "That, you know, the best time to plant a tree was 20 years ago.", "speaker": "Speaker 3"}, {"startTime": 1207.04, "endTime": 1210.079, "body": "The second best time is today, and it's the same for your data.", "speaker": "Speaker 3"}, {"startTime": 1210.189, "endTime": 1217.599, "body": "If you, if you're going to build AI systems, start thinking about it as soon as possible because you may not be collecting the right data, but that's OK.", "speaker": "Speaker 3"}, {"startTime": 1217.76, "endTime": 1226.479, "body": "The sooner you recognize it, the sooner you can start collecting it, and that will enable you to do things in the future that will be very disruptive for your business.", "speaker": "Speaker 3"}, {"startTime": 1228.119, "endTime": 1228.92, "body": "OK, well, thanks, Ron.", "speaker": "Speaker 2"}, {"startTime": 1228.93, "endTime": 1234.67, "body": "I think that's a good one to wrap up with and hopefully we cleared up many of these misconceptions during our conversation.", "speaker": "Speaker 2"}, {"startTime": 1235.089, "endTime": 1245.829, "body": "One thing we like to do on our episodes is wrap up with some key takeaways, and if you could, it might be difficult, but could you talk about what the top 2 or 3 messages are that you'd like to leave with our audience?", "speaker": "Speaker 2"}, {"startTime": 1247.25, "endTime": 1249.489, "body": "Yes, I would say it's very, very simple.", "speaker": "Speaker 3"}, {"startTime": 1249.609, "endTime": 1252.65, "body": "AI is an incredibly powerful technology.", "speaker": "Speaker 3"}, {"startTime": 1252.719, "endTime": 1261.839, "body": "It's, it's, it's truly going to be the most disruptive technology of our lives, but that doesn't mean it's omniscient.", "speaker": "Speaker 3"}, {"startTime": 1261.93, "endTime": 1263.589, "body": "It doesn't mean it's all powerful.", "speaker": "Speaker 3"}, {"startTime": 1263.65, "endTime": 1266.43, "body": "It doesn't mean it's going to take everybody's jobs away.", "speaker": "Speaker 3"}, {"startTime": 1266.89, "endTime": 1269.29, "body": "And it doesn't mean that it's easily understood.", "speaker": "Speaker 3"}, {"startTime": 1269.41, "endTime": 1276.53, "body": "And so I would encourage everybody to approach AI with an open mind, but, but be aware of the downsides, be aware of the challenges.", "speaker": "Speaker 3"}, {"startTime": 1276.829, "endTime": 1284.369, "body": "AI, like any technology, is very, you know, has power, but with that power comes a lot of responsibility.", "speaker": "Speaker 3"}, {"startTime": 1285.229, "endTime": 1289.67, "body": "That was a very succinct way of wrapping up that episode, Ron, thank you very much.", "speaker": "Jonas Norinder"}, {"startTime": 1289.79, "endTime": 1306.17, "body": "I'm sure our listeners will find this entire conversation very useful for anybody researching AI enabled tools and solutions, which I'm pretty much just guessing, I'm, I'm guessing is everyone or somebody that's just interested in learning more about AI and what it can do.", "speaker": "Jonas Norinder"}, {"startTime": 1306.27, "endTime": 1308.189, "body": "So thank you so much for being on the podcast.", "speaker": "Jonas Norinder"}, {"startTime": 1308.229, "endTime": 1310.88, "body": "It has really been a pleasure talking to you today, Ron.", "speaker": "Jonas Norinder"}, {"startTime": 1311.43, "endTime": 1312.27, "body": "Thank you so much guys.", "speaker": "Speaker 3"}, {"startTime": 1312.31, "endTime": 1313.369, "body": "I really enjoyed this.", "speaker": "Speaker 3"}, {"startTime": 1314.329, "endTime": 1316.569, "body": "So with that, we are at the end of the episode.", "speaker": "Jonas Norinder"}, {"startTime": 1316.689, "endTime": 1323.369, "body": "If you have any questions about the topic around AI that we discussed today, please feel free to reach out to Ron directly.", "speaker": "Jonas Norinder"}, {"startTime": 1323.449, "endTime": 1326.709, "body": "We have included a link to his email in the show notes.", "speaker": "Jonas Norinder"}, {"startTime": 1327.25, "endTime": 1340.25, "body": "And as usual, Don and I are always open to your comments, your thoughts, any feedback you may have on the show, or any of the topics, or if you have any suggestions for guests, you can certainly reach out to us as well.", "speaker": "Jonas Norinder"}, {"startTime": 1340.329, "endTime": 1343.729, "body": "You can find our contact information in the show notes as well.", "speaker": "Jonas Norinder"}, {"startTime": 1344.31, "endTime": 1356.089, "body": "So if you liked what you heard today, please help spread the word by rating the show and if you're not already a subscriber and would like to be notified when new episodes are released, please subscribe to the series.", "speaker": "Jonas Norinder"}, {"startTime": 1356.39, "endTime": 1359.65, "body": "Thank you for listening to this episode of Digitalization TikToks.", "speaker": "Jonas Norinder"}], "language": "en-US"}