title
Andrew Ng: Opportunities in AI - 2023

description
This discussion took place on July 26, 2023, at Cemex Auditorium, Stanford University, and was hosted by the Stanford Graduate School of Business. This talk covers: Trends in AI Technologies and Tools Supervised Learning Generative AI The Adoption of AI Opportunities in AI Process for Building Startups AI Risks and Social Impact Explore the essentials of machine learning and develop practical skills to implement AI applications in this beginner-friendly program: https://www.coursera.org/specializations/machine-learning-introduction

detail
{'title': 'Andrew Ng: Opportunities in AI - 2023', 'heatmap': [{'end': 670.117, 'start': 641.931, 'weight': 1}, {'end': 1352.89, 'start': 1326.307, 'weight': 0.726}], 'summary': "Dr. andrew wu discusses ai opportunities, supervised learning applications, and the $100 billion financial value of large language models. the impact of ai on business, innovative cost-effective projects, $5 billion potential value in niche datasets, romantic relationship coaching, and ai's influence in shipping are also covered, including bering ai's success in saving 10% fuel per ship annually, with potential implications for humanity's survival and progress.", 'chapters': [{'end': 407.51, 'segs': [{'end': 94.922, 'src': 'embed', 'start': 43.324, 'weight': 0, 'content': [{'end': 51.647, 'text': 'About 8 million people, one in 1, 000 persons on the planet, have taken an AI class from him.', 'start': 43.324, 'duration': 8.323}, {'end': 58.23, 'text': 'And through both his education and his AI work, He has changed numerous lives.', 'start': 52.487, 'duration': 5.743}, {'end': 61.343, 'text': 'Please welcome Dr. Andrew Lee.', 'start': 59.093, 'duration': 2.25}, {'end': 68.052, 'text': 'Thank you, Lisa.', 'start': 67.432, 'duration': 0.62}, {'end': 69.093, 'text': "It's good to see everyone.", 'start': 68.192, 'duration': 0.901}, {'end': 74.615, 'text': 'So what I want to do today is chat to you about some opportunities in AI.', 'start': 69.733, 'duration': 4.882}, {'end': 78.236, 'text': "So I've been saying AI is the new electricity.", 'start': 75.475, 'duration': 2.761}, {'end': 86.339, 'text': "One of the difficult things to understand about AI is that it is a general purpose technology, meaning that it's not useful only for one thing,", 'start': 78.456, 'duration': 7.883}, {'end': 89.56, 'text': "but it's useful for lots of different applications, kind of like electricity.", 'start': 86.339, 'duration': 3.221}, {'end': 94.922, 'text': "If I was to ask you, what is electricity good for? You know, it's not any one thing, it's a lot of things.", 'start': 90.2, 'duration': 4.722}], 'summary': "Dr. andrew lee has taught ai to 8 million people, emphasizing ai's broad applicability like electricity.", 'duration': 51.598, 'max_score': 43.324, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE43324.jpg'}, {'end': 155.406, 'src': 'embed', 'start': 118.25, 'weight': 2, 'content': [{'end': 120.431, 'text': 'which is relatively new, exciting development.', 'start': 118.25, 'duration': 2.181}, {'end': 124.353, 'text': "If you're familiar with AI, you may have heard of other tools,", 'start': 121.131, 'duration': 3.222}, {'end': 130.755, 'text': "but I'm gonna talk less about these additional tools and I'll focus today on what I think are currently the two most important tools,", 'start': 124.353, 'duration': 6.402}, {'end': 133.036, 'text': 'which are supervised learning and generative AI.', 'start': 130.755, 'duration': 2.281}, {'end': 141.251, 'text': 'So supervised learning is very good at labeling things or very good at computing input to output or A to B mappings.', 'start': 134.204, 'duration': 7.047}, {'end': 144.014, 'text': 'Given input A, give me an output B.', 'start': 141.311, 'duration': 2.703}, {'end': 149.52, 'text': 'For example, given an email, we can use supervised learning to label it as spam or not spam.', 'start': 144.014, 'duration': 5.506}, {'end': 155.406, 'text': "The most lucrative application of this that I've ever worked on is probably online advertising, where, given an ad,", 'start': 150.301, 'duration': 5.105}], 'summary': 'Focusing on two important ai tools: supervised learning and generative ai.', 'duration': 37.156, 'max_score': 118.25, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE118250.jpg'}, {'end': 284.26, 'src': 'embed', 'start': 245.936, 'weight': 4, 'content': [{'end': 250.22, 'text': 'Then you have an AI engineering team train an AI model to learn from this data.', 'start': 245.936, 'duration': 4.284}, {'end': 255.605, 'text': 'And then finally, you would find maybe a cloud service to run the trained AI model.', 'start': 250.901, 'duration': 4.704}, {'end': 258.588, 'text': "And then you can feed it, you know, best policy I've ever had.", 'start': 256.106, 'duration': 2.482}, {'end': 260.13, 'text': "And that's positive sentiment.", 'start': 258.708, 'duration': 1.422}, {'end': 266.477, 'text': 'And so I think the last decade was maybe the decade of large-scale supervised learning.', 'start': 260.976, 'duration': 5.501}, {'end': 272.418, 'text': 'What we found, starting about 10, 15 years ago, was if you were to train a small AI model,', 'start': 267.117, 'duration': 5.301}, {'end': 279.819, 'text': 'so train a small neural network or small deep learning algorithm basically a small AI model, maybe not on a very powerful computer then,', 'start': 272.418, 'duration': 7.401}, {'end': 284.26, 'text': 'as you fed it more data, its performance would get better for a little bit, but then it would flatten out,', 'start': 279.819, 'duration': 4.441}], 'summary': 'Decade of large-scale supervised learning, small ai model performance flattens out with more data', 'duration': 38.324, 'max_score': 245.936, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE245936.jpg'}, {'end': 355.575, 'src': 'embed', 'start': 327.501, 'weight': 7, 'content': [{'end': 335.185, 'text': 'So if that was the last decade of AI, I think this decade, um is turning out to be also doing everything we had in supervised learning,', 'start': 327.501, 'duration': 7.684}, {'end': 339.147, 'text': 'but adding to it um, the exciting two of generative AI.', 'start': 335.185, 'duration': 3.962}, {'end': 344.45, 'text': 'So many of you, maybe all of you, were played with chat GPT and bot and so on.', 'start': 340.007, 'duration': 4.443}, {'end': 349.232, 'text': 'Uh, but just, you know, given a piece of text, which we call a prompt, like I love eating.', 'start': 344.71, 'duration': 4.522}, {'end': 355.575, 'text': "If you run this multiple times, maybe you get bagels, cream cheese or my mother's meatloaf or um altar, friends,", 'start': 349.592, 'duration': 5.983}], 'summary': 'This decade focuses on supervised and generative ai, including chat gpt and bot, with exciting advancements.', 'duration': 28.074, 'max_score': 327.501, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE327501.jpg'}], 'start': 5.657, 'title': 'Ai opportunities and supervised learning', 'summary': "Discusses the opportunities in ai, presenting dr. andrew wu's insights and discussing supervised learning, its applications in labeling and mapping tasks, impact of using large-scale data and ai models, and integration of generative ai.", 'chapters': [{'end': 133.036, 'start': 5.657, 'title': 'Opportunities in ai: a chat with dr. andrew wu', 'summary': "Discusses the opportunities in ai and presents dr. andrew wu's insights on ai as a general purpose technology, highlighting its impact and discussing the two most important tools, supervised learning and generative ai.", 'duration': 127.379, 'highlights': ['Dr. Andrew Wu is a notable figure in the AI field, having impacted 8 million individuals through his AI classes and contributing to various AI initiatives.', 'AI is considered a general purpose technology, similar to electricity, with diverse applications across different fields.', 'Supervised learning and generative AI are emphasized as the two most important tools in AI, with supervised learning excelling in recognizing and labeling things, and generative AI being a recent and exciting development.']}, {'end': 407.51, 'start': 134.204, 'title': 'Supervised learning in ai', 'summary': 'Discusses supervised learning in ai, highlighting its applications in labeling and mapping tasks, such as spam detection in emails, online advertising for user click prediction, ship route optimization for fuel efficiency, and restaurant review sentiment analysis. it also delves into the impact of using large-scale data and ai models and the integration of generative ai in the field.', 'duration': 273.306, 'highlights': ['Supervised learning applications include spam detection, online advertising, ship route optimization, and restaurant review sentiment analysis.', 'Large AI models trained on extensive data significantly improved performance compared to smaller models.', 'The last decade focused on large-scale supervised learning, while the current decade integrates generative AI into the field.']}], 'duration': 401.853, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE5657.jpg', 'highlights': ['Dr. Andrew Wu impacted 8M individuals through AI classes.', 'AI is a general purpose technology with diverse applications.', 'Supervised learning excels in recognizing and labeling things.', 'Generative AI is a recent and exciting development.', 'Large AI models trained on extensive data significantly improved performance.', 'Supervised learning applications include spam detection and online advertising.', 'Supervised learning is emphasized as an important tool in AI.', 'The current decade integrates generative AI into the field.']}, {'end': 910.604, 'segs': [{'end': 452.065, 'src': 'embed', 'start': 408.371, 'weight': 3, 'content': [{'end': 417.696, 'text': 'So by taking texts that you find on the Internet or other sources and by using this input-output supervised learning to try to repeatedly predict the next words.', 'start': 408.371, 'duration': 9.325}, {'end': 425.861, 'text': 'if you train a very large AI system on hundreds of billions of words or, in the case of the largest model, is now more than a trillion words,', 'start': 417.696, 'duration': 8.165}, {'end': 427.862, 'text': 'then you get a large language model like ChatGP.', 'start': 425.861, 'duration': 2.001}, {'end': 432.301, 'text': 'There are additional other important technical details.', 'start': 428.82, 'duration': 3.481}, {'end': 433.841, 'text': 'I talked about predicting the next word.', 'start': 432.421, 'duration': 1.42}, {'end': 438.522, 'text': 'Technically, these systems predict the next subword, a part of word called the token.', 'start': 434.261, 'duration': 4.261}, {'end': 446.524, 'text': 'Then there are other techniques like ROHF for further tuning the AI output to be more hopeful, honest, and harmless.', 'start': 439.022, 'duration': 7.502}, {'end': 452.065, 'text': 'But at the heart of it is this using supervised learning to repeatedly predict the next word.', 'start': 446.944, 'duration': 5.121}], 'summary': 'Using supervised learning to train a large language model like chatgp on trillions of words to predict the next subword.', 'duration': 43.694, 'max_score': 408.371, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE408371.jpg'}, {'end': 584.212, 'src': 'embed', 'start': 539.254, 'weight': 1, 'content': [{'end': 546.899, 'text': 'And some of them turned out to be really valuable, but this is a realistic timeline for building and deploying a commercial grade AI system.', 'start': 539.254, 'duration': 7.645}, {'end': 552.867, 'text': 'In contrast with prompt-based AI where you write a prompt, this is what the workflow looks like.', 'start': 548.362, 'duration': 4.505}, {'end': 562.337, 'text': 'You can specify a prompt that takes maybe minutes or hours, and then you can deploy it to the Cloud and that takes maybe hours or days.', 'start': 553.788, 'duration': 8.549}, {'end': 569.986, 'text': 'So there are now certain AI applications that used to take me, you know, literally six months, maybe a year,', 'start': 562.998, 'duration': 6.988}, {'end': 573.928, 'text': 'to build that many teams around the world can now build in maybe a week.', 'start': 569.986, 'duration': 3.942}, {'end': 578.45, 'text': 'And I think this is already starting, but the best is still yet to come.', 'start': 574.568, 'duration': 3.882}, {'end': 584.212, 'text': 'This is starting to open up a flood of a lot more AI applications that can be built by a lot of people.', 'start': 579.05, 'duration': 5.162}], 'summary': 'Commercial ai systems can now be built in a week, reducing development time from months to days, opening up a flood of new ai applications.', 'duration': 44.958, 'max_score': 539.254, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE539254.jpg'}, {'end': 683.526, 'src': 'heatmap', 'start': 641.931, 'weight': 1, 'content': [{'end': 642.751, 'text': "So that's my prompt.", 'start': 641.931, 'duration': 0.82}, {'end': 643.872, 'text': "And now I'm just gonna run it.", 'start': 643.112, 'duration': 0.76}, {'end': 647.554, 'text': "And I've never run it before, so I really hope- oh, thank goodness we got the right answer.", 'start': 644.192, 'duration': 3.362}, {'end': 656.881, 'text': 'And this is literally all the code that takes to build a sentiment classifier.', 'start': 652.816, 'duration': 4.065}, {'end': 663.469, 'text': 'Um, and so today, you know, developers around the world can take literally maybe like 10 minutes to build a system like this.', 'start': 657.121, 'duration': 6.348}, {'end': 670.117, 'text': "Um, and that's a very exciting development.", 'start': 664.17, 'duration': 5.947}, {'end': 683.526, 'text': "So One of the things I've been working on was trying to teach online classes about how to use prompting, not just as a consumer tool,", 'start': 671.538, 'duration': 11.988}], 'summary': 'Building sentiment classifier takes 10 minutes, exciting development for developers worldwide.', 'duration': 41.595, 'max_score': 641.931, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE641931.jpg'}, {'end': 736.629, 'src': 'embed', 'start': 702.565, 'weight': 0, 'content': [{'end': 710.727, 'text': 'But the vast majority of financial value from AI today is, I think, supervised learning, where, for a single company like Google,', 'start': 702.565, 'duration': 8.162}, {'end': 713.028, 'text': 'can be worth more than 100 billion US dollars a year.', 'start': 710.727, 'duration': 2.301}, {'end': 718.583, 'text': 'Um, and also, there are millions of developers building supervised learning applications.', 'start': 713.048, 'duration': 5.535}, {'end': 725.505, 'text': "So it's already massively valuable and also with tremendous momentum behind it, just because of the sheer effort in, you know,", 'start': 718.663, 'duration': 6.842}, {'end': 727.166, 'text': 'finding applications and building applications.', 'start': 725.505, 'duration': 1.661}, {'end': 732.988, 'text': 'And in generative AI is the really exciting new entrance, uh, which is much smaller right now.', 'start': 728.086, 'duration': 4.902}, {'end': 736.629, 'text': "And then there are the other tools I'm including for completeness.", 'start': 734.348, 'duration': 2.281}], 'summary': 'Supervised learning in ai is worth over 100 billion usd annually for google, with millions of developers building applications, while generative ai is a smaller but exciting new entrance.', 'duration': 34.064, 'max_score': 702.565, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE702565.jpg'}, {'end': 803.535, 'src': 'embed', 'start': 776.874, 'weight': 7, 'content': [{'end': 781.979, 'text': 'But this light shaded region in green or orange, that light shaded region,', 'start': 776.874, 'duration': 5.105}, {'end': 789.467, 'text': 'is where the opportunity is for either new startups or for large companies incumbents to create and to enjoy value capture.', 'start': 781.979, 'duration': 7.488}, {'end': 795.373, 'text': 'But one thing I hope you take away from this slide is that all of these technologies are general purpose technologies.', 'start': 790.188, 'duration': 5.185}, {'end': 803.535, 'text': 'So, in the case of supervised learning, A lot of the work that had to be done over the last decade, but is continuing for the next decade,', 'start': 796.074, 'duration': 7.461}], 'summary': 'Opportunity for startups and large companies in green or orange region, all technologies are general purpose', 'duration': 26.661, 'max_score': 776.874, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE776874.jpg'}], 'start': 408.371, 'title': 'Ai language modeling and its impact', 'summary': "Covers training a large ai system to create a language model like chatgp using supervised learning and discusses the impact of large language models on reducing development time to build commercial-grade ai systems, leading to a flood of ai applications, with supervised learning's financial value estimated at over 100 billion us dollars a year for a single company like google.", 'chapters': [{'end': 452.065, 'start': 408.371, 'title': 'Ai language modeling', 'summary': 'Describes training a large ai system on hundreds of billions of words to create a language model like chatgp, which uses supervised learning to predict the next subword and incorporates techniques like rohf for tuning the ai output.', 'duration': 43.694, 'highlights': ['The AI system is trained on hundreds of billions of words, resulting in a large language model like ChatGP.', 'Supervised learning is used to predict the next subword, known as a token, in the text data.', 'Additional techniques like ROHF are employed to further tune the AI output to be more hopeful, honest, and harmless.']}, {'end': 910.604, 'start': 452.125, 'title': 'Impact of large language models', 'summary': 'Discusses the impact of large language models as a developer tool, reducing the time to build commercial-grade ai systems from months to days, leading to a flood of ai applications, and the financial value of supervised learning estimated at over 100 billion us dollars a year for a single company like google.', 'duration': 458.479, 'highlights': ['Supervised learning financial value estimated at over 100 billion US dollars a year for a single company like Google.', 'Reduction in time to build commercial-grade AI systems from months to days, allowing many teams to build AI applications in a week.', 'Generative AI emerging as an exciting new entrance, expected to more than double in the next three years due to developer interest, venture capital investments, and large corporates exploring applications.', 'Discussion on short-term fads and the concept of defensible long-term value in software products.', 'Comparison between the time taken to build systems using supervised learning and prompt-based AI, highlighting the significantly reduced time with prompt-based AI.']}], 'duration': 502.233, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE408371.jpg', 'highlights': ['Supervised learning financial value estimated at over 100 billion US dollars a year for a single company like Google.', 'Reduction in time to build commercial-grade AI systems from months to days, allowing many teams to build AI applications in a week.', 'Generative AI emerging as an exciting new entrance, expected to more than double in the next three years due to developer interest, venture capital investments, and large corporates exploring applications.', 'The AI system is trained on hundreds of billions of words, resulting in a large language model like ChatGP.', 'Comparison between the time taken to build systems using supervised learning and prompt-based AI, highlighting the significantly reduced time with prompt-based AI.', 'Supervised learning is used to predict the next subword, known as a token, in the text data.', 'Additional techniques like ROHF are employed to further tune the AI output to be more hopeful, honest, and harmless.', 'Discussion on short-term fads and the concept of defensible long-term value in software products.']}, {'end': 1106.123, 'segs': [{'end': 976.924, 'src': 'embed', 'start': 911.735, 'weight': 0, 'content': [{'end': 920.739, 'text': 'But with the rise of iOS, with the rise of iPhone, someone also figured out how to build things like Uber and Airbnb and Tinder, the very long-term,', 'start': 911.735, 'duration': 9.004}, {'end': 924.461, 'text': 'very defensible businesses that created sustaining value.', 'start': 920.739, 'duration': 3.722}, {'end': 935.447, 'text': 'And I think, with the rise of GenDev AI or the rise of new AI tools, I think really what excites me is the opportunity to create those really deep,', 'start': 925.122, 'duration': 10.325}, {'end': 940.409, 'text': 'really hard applications that hopefully can create very long-term value.', 'start': 935.447, 'duration': 4.962}, {'end': 945.568, 'text': 'So the first trend I want to share is AI as a general purpose technology,', 'start': 942.406, 'duration': 3.162}, {'end': 950.432, 'text': 'and a lot of the work that lies ahead of us is to find the very diverse use cases and to build them.', 'start': 945.568, 'duration': 4.864}, {'end': 956.817, 'text': "There's a second trend I want to share with you, which relates to why AI isn't more widely adopted yet.", 'start': 951.633, 'duration': 5.184}, {'end': 961.2, 'text': 'it feels like a bunch of us have been talking about AI for like 15 years or something.', 'start': 957.699, 'duration': 3.501}, {'end': 968.562, 'text': 'But if you look at where the value of AI is today, a lot of it is still very concentrated in consumer software internet.', 'start': 961.7, 'duration': 6.862}, {'end': 972.123, 'text': 'right?. Once you go outside, you know tech or consumer software internet.', 'start': 968.562, 'duration': 3.561}, {'end': 975.604, 'text': "there's some AI adoption, but it all feels very early.", 'start': 972.123, 'duration': 3.481}, {'end': 976.924, 'text': 'So why is that??', 'start': 976.324, 'duration': 0.6}], 'summary': 'Rise of ios led to creation of long-term, defensible businesses like uber, airbnb, and tinder. ai presents opportunity for deep, long-term value creation and diverse use cases, but adoption outside consumer software internet still feels early.', 'duration': 65.189, 'max_score': 911.735, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE911735.jpg'}, {'end': 1106.123, 'src': 'embed', 'start': 1019.857, 'weight': 2, 'content': [{'end': 1026.361, 'text': 'hardly anyone has a hundred million or a billion users that you can write and apply one piece of software to.', 'start': 1019.857, 'duration': 6.504}, {'end': 1034.364, 'text': 'So once you go to other industries, as we go from the head of this curve on the left over to the long tail,', 'start': 1027.569, 'duration': 6.795}, {'end': 1036.888, 'text': "these are some of the projects I see and I'm excited about.", 'start': 1034.364, 'duration': 2.524}, {'end': 1043.935, 'text': 'I was working with a pizza maker that was taking pictures of the pizza they were making,', 'start': 1037.671, 'duration': 6.264}, {'end': 1047.358, 'text': 'because they needed to do things like make sure that the cheese is spread evenly.', 'start': 1043.935, 'duration': 3.423}, {'end': 1050.3, 'text': 'So this is about a $5 million project.', 'start': 1048.018, 'duration': 2.282}, {'end': 1059.046, 'text': "But that recipe of hiring 100 engineers or dozens of engineers to work on a $5 million project, that doesn't make sense.", 'start': 1051.36, 'duration': 7.686}, {'end': 1064.974, 'text': 'Um or, as another example, working with an agriculture company,', 'start': 1059.91, 'duration': 5.064}, {'end': 1069.977, 'text': 'that um with them we figured out that if we use cameras to find out how tall is the wheat,', 'start': 1064.974, 'duration': 5.003}, {'end': 1072.539, 'text': 'and wheat is often bent over because of wind or rain or something.', 'start': 1069.977, 'duration': 2.562}, {'end': 1079.651, 'text': 'and we can chop off the wheat at the right height, then that results in more food for the farmer to sell and is also better for the environment.', 'start': 1073.109, 'duration': 6.542}, {'end': 1088.573, 'text': 'But this is another $5 million project that that old recipe of hiring a large group of highly skilled engineers to work on this one project.', 'start': 1080.151, 'duration': 8.422}, {'end': 1089.454, 'text': "that doesn't make sense.", 'start': 1088.573, 'duration': 0.881}, {'end': 1094.515, 'text': 'And similarly, materials grading, cloth grading, sheet metal grading, many projects like this.', 'start': 1090.434, 'duration': 4.081}, {'end': 1100.32, 'text': "So, whereas to the left in the head of this curve there's a small number of, let's say,", 'start': 1095.638, 'duration': 4.682}, {'end': 1104.562, 'text': 'multi-billion dollar projects and we know how to execute those delivering value.', 'start': 1100.32, 'duration': 4.242}, {'end': 1106.123, 'text': 'In other industries.', 'start': 1105.202, 'duration': 0.921}], 'summary': 'In various industries, $5 million projects are being executed without hiring large groups of skilled engineers, resulting in increased efficiency and value.', 'duration': 86.266, 'max_score': 1019.857, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1019857.jpg'}], 'start': 911.735, 'title': 'Ai impact on business and cost-effective projects', 'summary': 'Delves into the rise of ai as a general purpose technology, its impact on business, challenges in wider ai adoption, concentration of ai value in multi-billion dollar projects, scaling software projects outside consumer internet, and innovative cost-effective projects such as optimizing wheat cutting for farmers resulting in increased food production and environmental benefits.', 'chapters': [{'end': 996.805, 'start': 911.735, 'title': 'The rise of ai and its impact on business', 'summary': 'Discusses the rise of ai as a general purpose technology with the potential to create long-term, deep applications, the challenges in wider ai adoption outside consumer software internet, and the concentration of ai value in multi-billion dollar projects.', 'duration': 85.07, 'highlights': ['AI as a general purpose technology with the potential to create very long-term, very defensible businesses like Uber and Airbnb and Tinder.', 'The challenges in wider AI adoption outside consumer software internet, with most of the AI value concentrated in multi-billion dollar projects like advertising, web search, and product recommendations.', 'The need to find diverse use cases and build deep applications with AI to create long-term value.']}, {'end': 1043.935, 'start': 997.665, 'title': 'Scaling software projects outside consumer internet', 'summary': 'Discusses the challenges of scaling software projects outside consumer internet, exemplifying the difficulty in finding a large user base to apply software to and the shift in focus from consumer internet to other industries.', 'duration': 46.27, 'highlights': ['The challenge of scaling software projects outside consumer internet due to the lack of a large user base to apply software to', 'The shift in focus from consumer internet to other industries in search of new projects']}, {'end': 1106.123, 'start': 1043.935, 'title': 'Innovative cost-effective projects', 'summary': 'Discusses the inefficiency of using a large group of highly skilled engineers for a $5 million project and presents examples of cost-effective solutions, like using cameras to optimize wheat cutting for farmers, resulting in increased food production and environmental benefits.', 'duration': 62.188, 'highlights': ['Using cameras to optimize wheat cutting for farmers, resulting in increased food production and environmental benefits.', 'Discussing the inefficiency of using a large group of highly skilled engineers for a $5 million project.', 'Highlighting the inefficiency of the traditional approach in executing smaller projects, such as materials grading, cloth grading, and sheet metal grading.', 'Emphasizing the effectiveness of small, cost-effective projects in delivering value compared to multi-billion dollar projects.']}], 'duration': 194.388, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE911735.jpg', 'highlights': ['AI as a general purpose technology with potential for long-term, defensible businesses like Uber and Airbnb (Paragraph 1)', 'Challenges in wider AI adoption outside consumer software internet, with most value concentrated in multi-billion dollar projects (Paragraph 1)', 'Using cameras to optimize wheat cutting for farmers, resulting in increased food production and environmental benefits (Paragraph 3)', 'Shift in focus from consumer internet to other industries in search of new projects (Paragraph 2)', 'Need to find diverse use cases and build deep applications with AI to create long-term value (Paragraph 1)', 'Inefficiency of using a large group of highly skilled engineers for a $5 million project (Paragraph 3)', 'Inefficiency of traditional approach in executing smaller projects, such as materials grading, cloth grading, and sheet metal grading (Paragraph 3)', 'Effectiveness of small, cost-effective projects in delivering value compared to multi-billion dollar projects (Paragraph 3)', 'Challenge of scaling software projects outside consumer internet due to lack of large user base to apply software to (Paragraph 2)']}, {'end': 1408.73, 'segs': [{'end': 1145.237, 'src': 'embed', 'start': 1106.123, 'weight': 0, 'content': [{'end': 1117.388, 'text': "I'm seeing a very long tail of tens of thousands of let's call them $5 million projects that until now have been very difficult to execute on because of the high cost of customization.", 'start': 1106.123, 'duration': 11.265}, {'end': 1129.683, 'text': 'The trend that I think is exciting is that the AI community has been building better tools that lets us aggregate these use cases and make it easy for the end user to do the customization.', 'start': 1118.953, 'duration': 10.73}, {'end': 1138.232, 'text': "So specifically, I'm seeing a lot of exciting low code and no code tools that enable the user to customize the AI system.", 'start': 1129.723, 'duration': 8.509}, {'end': 1143.436, 'text': 'What this means is instead of me needing to worry that much about pictures of pizza.', 'start': 1138.812, 'duration': 4.624}, {'end': 1145.237, 'text': 'We have tools.', 'start': 1144.497, 'duration': 0.74}], 'summary': 'Ai community developing low code tools for $5 million projects, making customization easier.', 'duration': 39.114, 'max_score': 1106.123, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1106123.jpg'}, {'end': 1207.859, 'src': 'embed', 'start': 1186.37, 'weight': 1, 'content': [{'end': 1197.38, 'text': 'or a technology called data-centric AI, whereby, instead of asking the pizza factory to write a lot of code, which is challenging,', 'start': 1186.37, 'duration': 11.01}, {'end': 1200.323, 'text': 'we can ask them to provide data which turns out to be more feasible.', 'start': 1197.38, 'duration': 2.943}, {'end': 1207.859, 'text': 'And I think the second trend is important, because I think this is a key part of the recipe for taking the value of AI,', 'start': 1201.997, 'duration': 5.862}], 'summary': 'Data-centric ai reduces code writing, making ai more feasible.', 'duration': 21.489, 'max_score': 1186.37, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1186370.jpg'}, {'end': 1325.206, 'src': 'embed', 'start': 1287.192, 'weight': 3, 'content': [{'end': 1294.054, 'text': "So that's why I ended up starting AI Fun, which is a venture studio that builds startups to pursue a diverse set of AI opportunities.", 'start': 1287.192, 'duration': 6.862}, {'end': 1303.188, 'text': 'And of course, in addition to lots of startups, incumbent companies also have a lot of opportunities to integrate AI into existing businesses.', 'start': 1295.064, 'duration': 8.124}, {'end': 1304.188, 'text': 'In fact,', 'start': 1303.448, 'duration': 0.74}, {'end': 1315.094, 'text': "one pattern I'm seeing for incumbent businesses is distribution is often one of the cyclical advantages of incumbent companies is they play the cards right,", 'start': 1304.188, 'duration': 10.906}, {'end': 1318.975, 'text': 'can allow them to integrate AI into their products quite efficiently.', 'start': 1315.094, 'duration': 3.881}, {'end': 1325.206, 'text': 'But just to be concrete, where are the opportunities? So I think of this as a, this is what I think of as the AI stack.', 'start': 1319.916, 'duration': 5.29}], 'summary': 'Ai fun is a venture studio building startups for diverse ai opportunities, with incumbent companies also having efficient ai integration opportunities through distribution.', 'duration': 38.014, 'max_score': 1287.192, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1287192.jpg'}, {'end': 1358.153, 'src': 'heatmap', 'start': 1326.307, 'weight': 0.726, 'content': [{'end': 1329.449, 'text': 'At the bottom level is the hardware semiconductor layer.', 'start': 1326.307, 'duration': 3.142}, {'end': 1334.552, 'text': 'Fantastic opportunities there, but very capital intensive, very concentrated.', 'start': 1329.889, 'duration': 4.663}, {'end': 1336.713, 'text': "So there's a lot of resources, relatively few winners.", 'start': 1334.672, 'duration': 2.041}, {'end': 1338.714, 'text': 'So some people can and should play there.', 'start': 1337.154, 'duration': 1.56}, {'end': 1340.475, 'text': "I personally don't like to play there myself.", 'start': 1338.794, 'duration': 1.681}, {'end': 1342.837, 'text': "There's also the infrastructure layer.", 'start': 1341.296, 'duration': 1.541}, {'end': 1347.099, 'text': 'Also fantastic opportunities, but very capital intensive, very concentrated.', 'start': 1343.157, 'duration': 3.942}, {'end': 1349.661, 'text': 'So I tend not to play that myself either.', 'start': 1347.66, 'duration': 2.001}, {'end': 1352.89, 'text': "And then there's a developer tool layer.", 'start': 1351.069, 'duration': 1.821}, {'end': 1358.153, 'text': "What I showed you just now was, I was actually using OpenAI's API as a developer tool.", 'start': 1353.49, 'duration': 4.663}], 'summary': 'Hardware and infrastructure layers are capital intensive and concentrated, while developer tools offer opportunities.', 'duration': 31.846, 'max_score': 1326.307, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1326307.jpg'}], 'start': 1106.123, 'title': 'Ai customization and opportunities', 'summary': 'Highlights the emergence of low code and no code ai tools, unlocking a potential $5 billion value in niche datasets, and discusses diverse opportunities in ai, emphasizing the potential for startups and incumbent companies to integrate ai into existing businesses.', 'chapters': [{'end': 1225.91, 'start': 1106.123, 'title': 'Ai customization for low code and no code tools', 'summary': 'Highlights the emergence of low code and no code ai tools, enabling industries to train and deploy their own custom ai systems, thereby unlocking a potential $5 billion value in niche datasets and extending the reach of ai beyond the tech world.', 'duration': 119.787, 'highlights': ['The AI community has developed low code and no code tools that allow industries to customize AI systems, potentially unlocking a $5 billion value in niche datasets.', 'Emerging technology, such as data-centric AI and prompting tools, is enabling industries to train and deploy their own custom AI systems, extending the reach of AI beyond the tech world.', 'The trend of low code and no code tools is making it easier for industries, such as the pizza making factory, to utilize their unique datasets, which are not accessible through traditional search engines.']}, {'end': 1408.73, 'start': 1225.91, 'title': 'Ai opportunities and startups', 'summary': 'Discusses the diverse opportunities in ai, the establishment of ai fun as a venture studio, and the different layers of the ai stack, emphasizing the potential for startups and incumbent companies to integrate ai into existing businesses.', 'duration': 182.82, 'highlights': ['The chapter discusses the diverse opportunities in AI and the establishment of AI Fun as a venture studio', 'Emphasizing the potential for startups and incumbent companies to integrate AI into existing businesses', 'Explanation of the different layers of the AI stack and the opportunities they present']}], 'duration': 302.607, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1106123.jpg', 'highlights': ['Low code and no code AI tools unlock $5 billion value in niche datasets', 'Data-centric AI and prompting tools extend AI reach beyond tech world', 'Industries utilize unique datasets with low code and no code tools', 'Diverse opportunities in AI for startups and incumbent companies', 'AI Fun established as a venture studio to integrate AI into businesses', 'Explanation of different layers of the AI stack and opportunities']}, {'end': 1638.264, 'segs': [{'end': 1434.86, 'src': 'embed', 'start': 1408.73, 'weight': 1, 'content': [{'end': 1412.811, 'text': 'they can generate enough revenue to pay the infrastructure and the tooling layer.', 'start': 1408.73, 'duration': 4.081}, {'end': 1415.632, 'text': 'So actually, let me mention one example.', 'start': 1413.591, 'duration': 2.041}, {'end': 1419.634, 'text': 'Amorai I was actually just texting the CEO yesterday.', 'start': 1415.792, 'duration': 3.842}, {'end': 1427.277, 'text': 'But Amorai is a company that we built that uses AI for romantic relationship coaching.', 'start': 1420.054, 'duration': 7.223}, {'end': 1434.86, 'text': "And just to point out, I'm an AI guy, and I feel like I know nothing really about romance.", 'start': 1428.518, 'duration': 6.342}], 'summary': 'Ai company amorai generates revenue for romantic relationship coaching.', 'duration': 26.13, 'max_score': 1408.73, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1408730.jpg'}, {'end': 1484.306, 'src': 'embed', 'start': 1459.258, 'weight': 0, 'content': [{'end': 1465.061, 'text': "We're able to build something pretty unique using AI for kind of romantic relationship mentoring.", 'start': 1459.258, 'duration': 5.803}, {'end': 1470.956, 'text': 'Um. and the interesting thing about applications like these is, when we look around um,', 'start': 1465.692, 'duration': 5.264}, {'end': 1476.36, 'text': 'you know how many teams in the world are simultaneously expert in AI and in relationships?', 'start': 1470.956, 'duration': 5.404}, {'end': 1484.306, 'text': "And so, at the application layer, I'm seeing a lot of exciting opportunities that seem to have a very large market,", 'start': 1476.961, 'duration': 7.345}], 'summary': 'Ai used for romantic relationship mentoring with large market potential.', 'duration': 25.048, 'max_score': 1459.258, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1459258.jpg'}, {'end': 1565.389, 'src': 'embed', 'start': 1539.984, 'weight': 2, 'content': [{'end': 1545.667, 'text': 'They came to me and they said, hey, Andrew, you should build a business to use AI to make ships more fuel efficient.', 'start': 1539.984, 'duration': 5.683}, {'end': 1550.37, 'text': 'The specific idea was, think of it as a Google Maps for ships.', 'start': 1546.608, 'duration': 3.762}, {'end': 1556.353, 'text': 'We can suggest a ship or tell a ship how to steer so that you still get to your destination on time,', 'start': 1550.47, 'duration': 5.883}, {'end': 1558.875, 'text': 'but using it turns out about 10 percent less fuel.', 'start': 1556.353, 'duration': 2.522}, {'end': 1565.389, 'text': 'Um, and so what we now do is we spend about a month validating the idea.', 'start': 1559.725, 'duration': 5.664}], 'summary': 'Proposed using ai to make ships 10% more fuel efficient, with a google maps-like system.', 'duration': 25.405, 'max_score': 1539.984, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1539984.jpg'}], 'start': 1408.73, 'title': 'Ai for romantic relationship coaching', 'summary': 'Discusses the use of ai for romantic relationship coaching and making ships more fuel efficient, emphasizing market potential and startup development process.', 'chapters': [{'end': 1638.264, 'start': 1408.73, 'title': 'Ai for romantic relationship coaching', 'summary': 'Discusses using ai for romantic relationship coaching and ai for making ships more fuel efficient, highlighting the potential in the market and the process of building startups.', 'duration': 229.534, 'highlights': ['The potential of using AI for romantic relationship coaching and making ships more fuel efficient is discussed, highlighting the unique market opportunities and the process of building startups.', 'The company Amorai is mentioned, which uses AI for romantic relationship coaching, and the process of building startups is transparently shared.', 'The idea of using AI to make ships more fuel efficient and the process of validating the idea and recruiting a CEO for the project is explained.']}], 'duration': 229.534, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1408730.jpg', 'highlights': ['The potential of using AI for romantic relationship coaching and making ships more fuel efficient is discussed, highlighting the unique market opportunities and the process of building startups.', 'The company Amorai uses AI for romantic relationship coaching, and the process of building startups is transparently shared.', 'The idea of using AI to make ships more fuel efficient and the process of validating the idea and recruiting a CEO for the project is explained.']}, {'end': 2205.776, 'segs': [{'end': 1687.615, 'src': 'embed', 'start': 1662.158, 'weight': 0, 'content': [{'end': 1669.603, 'text': 'And 10% fuel savings translates the rough water amount to maybe $450, 000 in savings in fuel per ship per year.', 'start': 1662.158, 'duration': 7.445}, {'end': 1673.285, 'text': "And of course, it's also frankly quite a bit better for the environment.", 'start': 1670.083, 'duration': 3.202}, {'end': 1680.549, 'text': "And I think this is not a I think would not have existed if not for Dylan's fantastic work.", 'start': 1674.326, 'duration': 6.223}, {'end': 1683.857, 'text': 'um, and then also, you know, Mitsui, bringing this idea to me.', 'start': 1680.549, 'duration': 3.308}, {'end': 1687.615, 'text': 'And, I like this example because this is another one.', 'start': 1684.659, 'duration': 2.956}], 'summary': '10% fuel savings equals $450,000 per ship per year, also beneficial for environment.', 'duration': 25.457, 'max_score': 1662.158, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1662158.jpg'}, {'end': 1770.992, 'src': 'embed', 'start': 1725.978, 'weight': 1, 'content': [{'end': 1735.309, 'text': "And so I've learned that if I can just help get accurate technical validation and then use AI resources to make sure the AI tech is built quickly and well.", 'start': 1725.978, 'duration': 9.331}, {'end': 1738.932, 'text': "And I think we've always managed to help the companies build a strong technical team.", 'start': 1735.71, 'duration': 3.222}, {'end': 1743.776, 'text': 'Then partnering with subject matter experts often results in exciting new opportunities.', 'start': 1739.312, 'duration': 4.464}, {'end': 1756.243, 'text': "I want to share with you one other weird lesson I've learned about building startups, which is I like to engage only when there's a concrete idea.", 'start': 1745.876, 'duration': 10.367}, {'end': 1765.169, 'text': "This runs counter to a lot of the advice you hear from the design thinking methodology, which often says, don't rush to solutioning.", 'start': 1756.263, 'duration': 8.906}, {'end': 1767.55, 'text': 'Explore a lot of alternatives before you develop a solution.', 'start': 1765.389, 'duration': 2.161}, {'end': 1770.992, 'text': 'Honestly, we tried that, it was very slow.', 'start': 1768.511, 'duration': 2.481}], 'summary': 'Ai validation leads to quick and strong tech; concrete ideas for faster startup growth.', 'duration': 45.014, 'max_score': 1725.978, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1725978.jpg'}, {'end': 1927.862, 'src': 'embed', 'start': 1886.235, 'weight': 3, 'content': [{'end': 1888.237, 'text': 'So AI is a very powerful technology.', 'start': 1886.235, 'duration': 2.002}, {'end': 1894.021, 'text': 'To state something you probably guessed, my teams and I, we only work on projects that move humanity forward.', 'start': 1888.897, 'duration': 5.124}, {'end': 1901.508, 'text': 'And we have multiple times killed projects that we assess to be financially sound based on ethical grounds.', 'start': 1894.302, 'duration': 7.206}, {'end': 1902.569, 'text': 'It turns out.', 'start': 1901.588, 'duration': 0.981}, {'end': 1909.434, 'text': "I've been surprised and sometimes dismayed at the creativity of people to come up with good ideas.", 'start': 1902.569, 'duration': 6.865}, {'end': 1914.578, 'text': 'sorry, to come up with really bad ideas that seem profitable but really should not be built.', 'start': 1909.434, 'duration': 5.144}, {'end': 1917.68, 'text': "We've killed a few projects on those, on those grounds.", 'start': 1915.539, 'duration': 2.141}, {'end': 1924.081, 'text': 'And then I think it has to be acknowledged that AI today does have problems with bias, fairness, accuracy.', 'start': 1918.66, 'duration': 5.421}, {'end': 1927.862, 'text': 'Um, but also, you know, the technology is improving quickly.', 'start': 1924.101, 'duration': 3.761}], 'summary': "Ai technology can have ethical issues, and it's improving quickly.", 'duration': 41.627, 'max_score': 1886.235, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1886235.jpg'}, {'end': 1972.466, 'src': 'embed', 'start': 1945.788, 'weight': 5, 'content': [{'end': 1952.312, 'text': 'When I think of the biggest risk of AI, I think that one of the biggest risks is the disruption to jobs.', 'start': 1945.788, 'duration': 6.524}, {'end': 1963.458, 'text': 'This is a diagram from a paper by our friends at the University of Pennsylvania and some folks at OpenAI analyzing the exposure of different jobs to AI automation.', 'start': 1953.392, 'duration': 10.066}, {'end': 1972.466, 'text': 'And it turns out that whereas the previous wave of automation mainly the most exposed jobs were often the lower wage jobs,', 'start': 1964.318, 'duration': 8.148}], 'summary': 'Ai poses a risk to jobs, as shown by university of pennsylvania and openai research.', 'duration': 26.678, 'max_score': 1945.788, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1945788.jpg'}, {'end': 2173.359, 'src': 'embed', 'start': 2145.209, 'weight': 6, 'content': [{'end': 2149.932, 'text': 'leading to massive depopulation of some parts of the planets, or much lower odds,', 'start': 2145.209, 'duration': 4.723}, {'end': 2154.335, 'text': 'but maybe someday an asteroid doing to us what it had done to the dinosaurs.', 'start': 2149.932, 'duration': 4.403}, {'end': 2163.197, 'text': 'I think, if you look at the actual real extinction risk to humanity, AI having more intelligence, even artificial intelligence in the world,', 'start': 2154.995, 'duration': 8.202}, {'end': 2164.737, 'text': 'will be a key part of the solution.', 'start': 2163.197, 'duration': 1.54}, {'end': 2173.359, 'text': 'So I feel like, if you want humanity to survive and thrive for the next thousand years, rather than slowing AI down, which some people propose,', 'start': 2164.757, 'duration': 8.602}], 'summary': "Ai's increasing intelligence is key to humanity's survival and thriving for the next thousand years.", 'duration': 28.15, 'max_score': 2145.209, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE2145209.jpg'}], 'start': 1638.984, 'title': "Ai's influence in shipping and future implications", 'summary': "Highlights bering ai's success in enabling ships to save 10% fuel, translating to approximately $450,000 per ship annually, and discusses the ethical considerations, risks, and benefits of ai, along with its potential impact on humanity's survival and progress.", 'chapters': [{'end': 1824.84, 'start': 1638.984, 'title': 'Ai startup success story', 'summary': 'Highlights the success story of bering ai, where the company has enabled hundreds of ships to save 10% fuel, amounting to approximately $450,000 per ship per year, thus proving the effectiveness of ai in maritime shipping.', 'duration': 185.856, 'highlights': ['Bering AI enabled hundreds of ships to save 10% fuel, resulting in approximately $450,000 per ship per year.', 'The importance of subject matter expertise in combination with AI in developing innovative solutions.', 'The efficiency of engaging with concrete startup ideas for quick validation and execution.']}, {'end': 2205.776, 'start': 1825.46, 'title': "Ai's impact and future", 'summary': "Discusses the efficient utilization of subject matter experts' ideas, the ethical considerations in ai project selection, the risks and benefits of ai, including job disruption and the potential for ai to contribute to humanity's survival and advancement.", 'duration': 380.316, 'highlights': ["AI teams quickly validating and building on subject matter experts' ideas leads to efficient innovation.", 'Emphasis on ethical grounds in AI project selection, with the rejection of financially sound projects for ethical reasons.', 'Risk assessment of AI projects based on bias, fairness, and accuracy, leading to the rejection of potentially profitable yet unethical ideas.', "Concern over AI's impact on job disruption and the responsibility to ensure the well-being of individuals affected by automation.", "AI's potential to contribute to humanity's survival and advancement, with a focus on managing real extinction risks."]}], 'duration': 566.792, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/5p248yoa3oE/pics/5p248yoa3oE1638984.jpg', 'highlights': ['Bering AI enabled ships to save 10% fuel, translating to approximately $450,000 per ship annually', 'The importance of subject matter expertise in combination with AI for innovative solutions', "AI teams quickly validating and building on subject matter experts' ideas for efficient innovation", 'Emphasis on ethical grounds in AI project selection, rejecting financially sound projects for ethical reasons', 'Risk assessment of AI projects based on bias, fairness, and accuracy, leading to rejection of potentially profitable yet unethical ideas', "Concern over AI's impact on job disruption and responsibility to ensure well-being of affected individuals", "AI's potential to contribute to humanity's survival and advancement, focusing on managing real extinction risks", 'Efficiency of engaging with concrete startup ideas for quick validation and execution']}], 'highlights': ['Supervised learning financial value estimated at over 100 billion US dollars a year for a single company like Google', 'Dr. Andrew Wu impacted 8M individuals through AI classes', 'Bering AI enabled ships to save 10% fuel, translating to approximately $450,000 per ship annually', 'Low code and no code AI tools unlock $5 billion value in niche datasets', 'Large AI models trained on extensive data significantly improved performance', 'Generative AI emerging as an exciting new entrance, expected to more than double in the next three years', 'AI as a general purpose technology with potential for long-term, defensible businesses like Uber and Airbnb', 'The potential of using AI for romantic relationship coaching and making ships more fuel efficient is discussed', 'Using cameras to optimize wheat cutting for farmers, resulting in increased food production and environmental benefits', 'Supervised learning excels in recognizing and labeling things']}