1 00:00:02,280 --> 00:00:02,840 Hi guys. 2 00:00:02,840 --> 00:00:03,880 Welcome to this session. 3 00:00:03,880 --> 00:00:10,160 So in this session we want to talk about the key terminologies we use in the world of AI and the players 4 00:00:10,160 --> 00:00:11,800 in generative AI currently. 5 00:00:12,280 --> 00:00:18,960 So if you look at what is generative AI, it is going to be the AI systems that can create new content 6 00:00:19,000 --> 00:00:25,520 around text, images, music, coding by learning patterns from large data sets. 7 00:00:25,720 --> 00:00:32,840 Now, the key difference between traditional AI and generative AI is that generative AI can generate 8 00:00:33,160 --> 00:00:37,200 original outputs instead of merely analyzing data. 9 00:00:37,600 --> 00:00:42,320 Now, the popular tools which we are using currently in this particular field are ChatGPT, which is 10 00:00:42,320 --> 00:00:49,000 for text generation, Midjourney for image generation, you have pseudo AI, which is for music creation 11 00:00:49,000 --> 00:00:49,640 as well. 12 00:00:49,680 --> 00:00:56,640 And the key applications which we have at this moment are text generation, image generation, coding 13 00:00:56,680 --> 00:00:59,600 assistance, and enterprise solutions. 14 00:00:59,600 --> 00:01:02,510 So for enterprise solutions, we are using Adobe Sensei. 15 00:01:02,870 --> 00:01:05,830 Coding assistance is being done by GitHub Copilot. 16 00:01:05,830 --> 00:01:13,510 So there are different key applications now in in the market which which provides these kind of services. 17 00:01:13,910 --> 00:01:17,430 Now if you look at the core generative AI terminologies. 18 00:01:17,430 --> 00:01:18,350 So LMS. 19 00:01:18,390 --> 00:01:21,070 LMS are large language models. 20 00:01:21,070 --> 00:01:28,390 So these are AI trained on enormous amount of data sets to understand and generate human like text. 21 00:01:28,830 --> 00:01:34,150 Now these are powerful tools like these are built on like ChatGPT for Claude. 22 00:01:34,310 --> 00:01:37,750 These are all LMS which we can use now. 23 00:01:38,310 --> 00:01:44,230 At the background of it there is neural networks being used, which is a brain inspired systems for 24 00:01:44,230 --> 00:01:47,590 learning and decision making, which we will get now. 25 00:01:47,590 --> 00:01:54,510 There is also Gans, which is Generative Adversarial Networks, which is basically two neuro neural 26 00:01:54,510 --> 00:01:58,550 networks that compete to create and refine the outputs. 27 00:01:58,710 --> 00:02:03,630 It is used in artistic tools like stable diffusion, which we'll see in this course. 28 00:02:04,390 --> 00:02:08,070 Also there is NLP, which is natural language processing. 29 00:02:08,230 --> 00:02:15,430 It is a subfield of AI focused on enabling machines to understand, interpret and respond to human language. 30 00:02:15,750 --> 00:02:22,670 Now it also enables AI to understand respond in human language, which can be done and it gives a better 31 00:02:22,670 --> 00:02:23,870 user experience. 32 00:02:25,350 --> 00:02:30,910 So there are in ChatGPT specifically, if you look at it, there are certain specific terminologies 33 00:02:31,350 --> 00:02:36,390 which you'll get to see which is like prompting prompts, which is like inputs or instructions, which 34 00:02:36,390 --> 00:02:40,790 we give to the AI guiding the AI's response. 35 00:02:40,950 --> 00:02:48,310 Now there is temperature adjusting creativity, which is you will get in OpenAI playground or workbenches 36 00:02:48,310 --> 00:02:55,390 or API usage, wherein there is a degree which is low and high, low being precise, which is at 0.2 37 00:02:55,590 --> 00:02:58,830 and high can be imaginative, which is 0.8. 38 00:02:58,830 --> 00:02:58,940 eight. 39 00:02:59,180 --> 00:03:06,980 There is also a concept of tokens, units of data, basically in in the form of text, images and symbols. 40 00:03:07,300 --> 00:03:13,180 Uh, different models have different token limits, like GPT four have thousands of tokens for input 41 00:03:13,180 --> 00:03:13,820 and output. 42 00:03:14,260 --> 00:03:17,060 Similarly, there is something we call as context window. 43 00:03:17,060 --> 00:03:20,580 It is a memory unit for conversation history. 44 00:03:20,940 --> 00:03:28,180 Older details are dropped off as the window fills and there is hallucinations, which is AI generated 45 00:03:28,180 --> 00:03:32,260 responses that are factually incorrect or nonsensical. 46 00:03:33,340 --> 00:03:39,060 Now, if you look at the tech giants who are basically driving the generative AI, currently there is 47 00:03:39,060 --> 00:03:45,940 Microsoft, which is partnered with OpenAI, and they have tools like Azure AI, GitHub, copilot, 48 00:03:46,220 --> 00:03:47,820 Co-Pilot, which is for coding. 49 00:03:48,140 --> 00:03:54,420 Then there is Google or DeepMind, and their tools are Gemini Palm models. 50 00:03:54,580 --> 00:03:57,580 They're integrated into Google Workspace, Google products. 51 00:03:57,940 --> 00:03:59,180 Then there is Nvidia. 52 00:03:59,420 --> 00:04:03,700 Nvidia focuses on manufacturing GPUs and AI infrastructure. 53 00:04:04,100 --> 00:04:07,340 It powers the hardware behind AI innovations. 54 00:04:07,820 --> 00:04:09,300 Other than this, there's Adobe. 55 00:04:09,420 --> 00:04:15,020 Adobe has tools like Adobe Sensei, which enhances creativity for professionals. 56 00:04:16,220 --> 00:04:22,260 Lastly, if you look at the specialized AI companies driving innovations, there is OpenAI specifically 57 00:04:22,260 --> 00:04:24,380 focusing on GPT four Dall-E. 58 00:04:24,700 --> 00:04:28,580 It's a versatile text and image generation platform. 59 00:04:28,740 --> 00:04:31,700 Then there is anthropic, which has who has created Claude. 60 00:04:31,900 --> 00:04:35,140 It prioritizes AI safety and ethical development. 61 00:04:35,340 --> 00:04:41,020 Then there is stable AI, which has created stable diffusion, which is an open source image creation 62 00:04:41,020 --> 00:04:41,700 platform. 63 00:04:42,060 --> 00:04:47,220 And then there is hugging face, which is basically a collaborative platform for AI models and data 64 00:04:47,260 --> 00:04:47,740 sets. 65 00:04:48,100 --> 00:04:54,540 So these are all the developments, companies, platforms, terms, terminologies which you will get 66 00:04:54,540 --> 00:04:56,940 to see a lot in the AI space.