{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# GENERATIVE AI : A NEW REVOLUTION","metadata":{}},{"cell_type":"markdown","source":"## Table of Contents\n\n* [Introduction](#Introduction)\n* [What is generative AI ? ](#What-is-generative-AI-?)\n* [Generative ai applications](#Generative-ai-applications)\n* [How does it work ?](#How-does-it-work-?)\n* [Timeline](#Timeline)\n* [Big companies contributions](#Big-companies-contributions)\n* [Ethical considerations](#Ethical-considerations)\n* [Expectations and future of generative ai](#Expectations-and-future-of-generative-ai)\n* [Conclusion](#Conclusion)\n* [References](#References)","metadata":{}},{"cell_type":"markdown","source":"# Introduction","metadata":{}},{"cell_type":"markdown","source":"Artificial intelligence has been the subject of much discussion lately, especially with the emergence of generative AI, which is just a subcategory of AI. However, it is worth noting that it is one of the most fascinating results of research in this field. This is due to its multiple applications and its flexibility, which far surpass the methods and algorithms of traditional (discriminative) AI. With the rise of this technology, we can now perform complex tasks such as content creation (images, videos, code, etc.) and even engage in lengthy conversations with machines in real-time on any subject, surpassing even human capabilities sometimes . The purpose of this article is to shed light on the secrets of generative AI and demystify the different aspects of this technology, considering the scientific, technical, economic, social, and ethical contexts, in order to have a clear vision of its current state and even its future .","metadata":{}},{"cell_type":"code","source":"import matplotlib.pyplot as plt\nimport matplotlib.image as mpimg\n\n# Chemin de l'image\nchemin_image = '/kaggle/input/images/img2.png'\n\n# Charger l'image\nimg = mpimg.imread(chemin_image)\n\n# Afficher l'image\nplt.imshow(img)\nplt.axis('off')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2023-07-14T21:58:56.216719Z","iopub.execute_input":"2023-07-14T21:58:56.217145Z","iopub.status.idle":"2023-07-14T21:58:56.470577Z","shell.execute_reply.started":"2023-07-14T21:58:56.217113Z","shell.execute_reply":"2023-07-14T21:58:56.469507Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" chatgpt sprints to one million users1","metadata":{}},{"cell_type":"markdown","source":"# What is generative AI ?","metadata":{}},{"cell_type":"markdown","source":"we can define generative AI in several possible ways, and here is the official definition from Wikipedia : \" Generative artificial intelligence or generative AI is a type of artificial intelligence (AI) system capable of generating text, images, or other media in response to prompts. Generative AI models learn the patterns and structure of their input training data, and then generate new data that has similar characteristics\". The difference between generative AI and traditional AI, or more technically, discriminative AI, lies in the algorithms and methods used, but above all, in the possibilities offered by each of them. The table below helps illustrate this :\n","metadata":{}},{"cell_type":"markdown","source":"| discriminative ai || generative ai |\n|------------------------------------------------------------------||---------------------------------------------|\n| detecting an email as a spam or not spam||generating emails|\n| predicting if an image is a dog or cat image||generating cat/dog images|\n| predicting housing prices||manipulating images/texts|\n| recognizing voices||synthesizing artificial audios|","metadata":{}},{"cell_type":"code","source":"%%html\n\n","metadata":{"execution":{"iopub.status.busy":"2023-07-14T21:58:24.188484Z","iopub.execute_input":"2023-07-14T21:58:24.188982Z","iopub.status.idle":"2023-07-14T21:58:24.197689Z","shell.execute_reply.started":"2023-07-14T21:58:24.188948Z","shell.execute_reply":"2023-07-14T21:58:24.196344Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Generative ai applications ","metadata":{}},{"cell_type":"markdown","source":"Generative AI has revolutionized the world, and one of the reasons behind this is undoubtedly the fact that it can be leveraged in almost every domain. We're talking about applications in healthcare, education, banking, conversations, art, and much more. However, it's important to note that some of these applications are more prevalent than others, at least temporarily. Particularly, content creation applications are quite widespread, including image generation, video generation, text generation of various kinds, and blog article writing. This is made possible due to the availability and accessibility of tools like ChatGPT, Midjourney, DALL-E, and others.\n\nBeyond everyday and basic applications, we find more advanced and sophisticated uses of generative AI, such as its adoption in the medical field to detect diseases, improve medical imaging, or contribute to the development of new medications. Software developers also employ this technology in their work to interpret, generate, or enhance code, and sometimes even design complex programs. Marketing and advertising experts take advantage of generative AI to generate advertisements, conduct market research, and create personalized product descriptions, among other tasks.\n\nEducation, in turn, is a field that has been significantly impacted by the advancements of generative AI. The content generation capabilities have enabled the creation of personalized lessons that cater to specific needs and the writing of scientific articles covering a wide range of subjects. In summary, these are some of the most popular applications of generative AI. The graph below summarizes the aforementioned points in terms of numbers :\n\n\n\n\n\n\n\n\n","metadata":{}},{"cell_type":"code","source":"import matplotlib.pyplot as plt\nimport matplotlib.image as mpimg\n\n# Chemin de l'image\nchemin_image = '/kaggle/input/images/img.png'\n\n# Charger l'image\nimg = mpimg.imread(chemin_image)\n\n# Afficher l'image\nplt.imshow(img)\nplt.axis('off')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2023-07-14T21:59:31.300329Z","iopub.execute_input":"2023-07-14T21:59:31.301386Z","iopub.status.idle":"2023-07-14T21:59:31.526647Z","shell.execute_reply.started":"2023-07-14T21:59:31.301346Z","shell.execute_reply":"2023-07-14T21:59:31.525731Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n \n Rate of generative AI adoption in the workplace (us) 2023,by industry2","metadata":{}},{"cell_type":"markdown","source":"# How does it work ?","metadata":{}},{"cell_type":"markdown","source":"Now that we have a general idea about generative AI, let's explore some of the most popular generative methods and algorithms that underlie the functioning of this technology.","metadata":{}},{"cell_type":"markdown","source":"\n\n>Generative Adversarial Networks (GANs): GANs are one of the most popular and widely used generative AI methods. They consist of a generator network that learns to generate synthetic data, and a discriminator network that learns to distinguish between real and fake data. GANs have been successfully applied to various domains, including image generation, text generation, and music generation.\n\n>Variational Autoencoders (VAEs): VAEs are another popular method for generative modeling. They combine elements of autoencoders and probabilistic modeling to learn a latent representation of the data. VAEs have been used for tasks such as image synthesis, text generation, and anomaly detection.\n\n>Transformer-based Models: Transformer models, originally introduced by the \"Attention Is All You Need\" paper, have had a significant impact on generative AI. Models like OpenAI's GPT (Generative Pre-trained Transformer) have achieved impressive results in natural language generation tasks, such as text completion, machine translation, and story generation.\n\n>PixelRNN and PixelCNN: PixelRNN and PixelCNN are generative models specifically designed for generating images pixel by pixel. They model the conditional distribution of each pixel given the previous pixels in the image. These models have been used to generate high-resolution and realistic images.\n\n>Flow-based Models: Flow-based generative models are gaining popularity due to their ability to learn invertible transformations from the data. Models like Glow and RealNVP learn bijective mappings between the data space and a simple latent space. Flow-based models have shown promising results in image generation tasks.\n\n>Reinforcement Learning for Sequence Generation: Reinforcement Learning (RL) techniques have been used in sequence generation tasks such as language generation and music generation. By framing the generation process as a sequential decision-making problem, RL algorithms can learn to generate high-quality and coherent sequences.\n\n>Diffusion models:Diffusion models are another important class of generative models that have gained significant attention in recent years. Diffusion models, also known as probabilistic diffusion models (PDMs), are designed to model the process of diffusion, where a simple distribution is iteratively transformed into a more complex distribution to generate samples.\nThe key idea behind diffusion models is to perform a series of transformations on a noise sample to gradually approximate the target distribution. The transformations are typically designed to be computationally tractable and reversible. By iteratively applying these transformations, diffusion models generate samples that progressively resemble the desired data distribution 3\n\n\n\n\n\n","metadata":{}},{"cell_type":"markdown","source":"# Timeline ","metadata":{}},{"cell_type":"markdown","source":"\n\n\n2020: \n* OpenAI's DALL·E - OpenAI introduced DALL·E, a generative model that generates images from textual descriptions. It showcased the ability to generate highly detailed and imaginative images based on textual prompts.\n\n* OpenAI's CLIP - OpenAI released CLIP, a model that connects vision and language by jointly training on large-scale image and text datasets. CLIP demonstrated impressive capabilities in understanding images and generating textual descriptions.\n\n* VQ-VAE-2 - Razavi et al. proposed VQ-VAE-2, a generative model based on Variational Autoencoders (VAEs) and Vector Quantization (VQ). VQ-VAE-2 achieved remarkable results in image generation and representation learning.\n\n* GPT-3 - OpenAI unveiled GPT-3, a language model with 175 billion parameters, making it one of the largest AI models at the time. GPT-3 showcased exceptional performance in various natural language processing tasks, including text generation, translation, and question-answering.\n\n2021: \n* DALL·E 2 - OpenAI introduced DALL·E 2, an improved version of the original DALL·E model. DALL·E 2 showcased the ability to generate even more detailed and diverse images based on textual prompts.\n\n* GAN Image Synthesis Advancements - Researchers made significant progress in generative adversarial networks (GANs) for image synthesis. Techniques like Progressive Growing of GANs (PGGAN), CycleGAN, and StyleGAN2 improved the quality, resolution, and control of generated images.\n\n* GPT-3 Applications and Fine-Tuning - Developers and researchers explored various applications of GPT-3, including content generation, chatbots, code completion, creative writing, and more. GPT-3's large-scale language generation capabilities were fine-tuned for specific tasks, resulting in impressive results.\n\n2022: \n* VQ-VAE-3 and Improved Image Generation - Following VQ-VAE-2, researchers continued to improve the architecture with VQ-VAE-3, achieving better image generation and representation learning capabilities. These advancements contributed to the generation of high-quality and diverse images.\n\n* Text-to-Image Synthesis - Researchers made progress in text-to-image synthesis by leveraging large-scale language models and GANs. These models demonstrated the ability to generate images from textual descriptions with better accuracy and realism.\n\n2023:\n* Continued Research and Applications - Generative AI research continues to advance rapidly, with ongoing exploration of large-scale language models, image synthesis techniques, and multimodal generative models. Applications span various domains, including creative arts, design, entertainment, and content generation\n4\n\n\n\n\n\n","metadata":{}},{"cell_type":"markdown","source":"# Big companies contributions","metadata":{}},{"cell_type":"markdown","source":"In the field of generative artificial intelligence (AI), several prominent companies have emerged as key contributors, driving innovation and pushing the boundaries of what is possible.These companies have become significant contributors to the field of generative AI due to several key capacities:\n\nResearch and Development (R&D) : These companies invest heavily in research and development, maintaining dedicated teams of researchers and engineers focused on advancing generative AI technologies. They have the resources and infrastructure to conduct cutting-edge research, explore novel approaches, and develop state-of-the-art models.\n\nTalent Acquisition : Big companies have the ability to attract top talent in the field of AI. They hire experienced researchers, scientists, and engineers who specialize in generative AI and have a track record of producing impactful work. By assembling teams of experts, these companies can drive innovation and push the boundaries of what is possible in generative AI.\n\nComputational Resources : Generative AI often requires substantial computational power for training large-scale models. Big companies have access to powerful computing infrastructure, including GPUs and distributed computing systems, allowing them to train and deploy complex generative models at scale. This computational advantage enables them to tackle more challenging problems and experiment with larger datasets.\n\nData Resources : Large companies often possess extensive data resources, including diverse and high-quality datasets that can be used to train generative models. These datasets provide a foundation for developing robust and versatile generative AI systems across various domains. The availability of rich data resources gives these companies an edge in training models that can generate realistic and diverse outputs.\n\nIndustry Applications : Big companies have a strong presence across various industries, including technology, healthcare, finance, and entertainment. Their ability to apply generative AI techniques to real-world problems and develop practical applications gives them a significant impact on the field. They have the opportunity to leverage their existing platforms and user bases to integrate generative AI technologies into their products and services.\n\nOverall, the combination of strong R&D capabilities, access to talent and computational resources, rich data resources, and practical industry applications positions these companies as major contributors to the field of generative AI\n\nHere are some notable examples:\n\nOpenAI: OpenAI has been at the forefront of generative AI research and development. They have released several iterations of their GPT (Generative Pre-trained Transformer) models, which have demonstrated impressive capabilities in natural language processing tasks. OpenAI has also contributed to the development of reinforcement learning algorithms and frameworks.\n\nGoogle: Google has been actively involved in advancing generative AI. Their research teams have published influential papers on topics like generative adversarial networks (GANs), image synthesis, and text generation. Google's DeepMind division has made notable contributions to reinforcement learning and has developed state-of-the-art models in this area.\n\nFacebook: Facebook's AI research division, Facebook AI Research (FAIR), has made significant contributions to generative AI. They have published research papers on various topics, including image generation, video synthesis, and language modeling. Facebook has also explored the application of generative models in virtual reality and augmented reality.\n\nNVIDIA: NVIDIA has played a crucial role in advancing generative AI by developing high-performance GPUs and software frameworks that enable faster training and inference for deep learning models. Their GPUs have been instrumental in training large-scale generative models and accelerating the development of new algorithms.\n\nMicrosoft: Microsoft has been actively involved in generative AI research and has made contributions in areas such as image synthesis, text generation, and speech synthesis. They have also developed tools and frameworks, such as Microsoft Cognitive Toolkit (CNTK) and Azure Machine Learning, that support generative AI research and development.\n\nThese are just a few examples, and many other companies, both large and small, have made valuable contributions to the field of generative AI. The collective efforts of these companies have propelled the advancements and applications of generative AI across various industries 5","metadata":{}},{"cell_type":"markdown","source":"# Ethical considerations ","metadata":{}},{"cell_type":"markdown","source":"Generative AI presents a range of ethical considerations that need to be carefully addressed as the technology advances. Here are some key ethical considerations related to generative AI:\n\nBias and Fairness: Generative AI models learn from data, and if the training data contains biases or reflects societal inequalities, the generated outputs can perpetuate or amplify those biases. It is crucial to ensure fairness and mitigate bias by carefully curating and diversifying training data and implementing robust evaluation methods.\n\nMisuse of AI-generated Content: The ease of generating realistic and convincing content raises concerns about potential misuse. AI-generated content can be used for spreading disinformation, creating deepfakes, or generating offensive and harmful material. Safeguards and regulations are necessary to prevent malicious uses and protect individuals' privacy and reputation.\n\nIntellectual Property and Copyright: Generative AI models have the capability to produce content that may infringe on intellectual property rights. Clarifying the ownership and legal implications of AI-generated creations is a complex issue that requires thoughtful consideration to balance the rights of creators and the potential for innovation.\n\nTransparency and Explainability: Generative AI models are often considered black boxes, making it challenging to understand how they arrive at their outputs. Ensuring transparency and explainability is essential, especially when AI-generated content influences decision-making or has significant consequences. Developing methods to interpret and explain the reasoning behind generative AI outputs is crucial for building trust and accountability.\n\nConsent and Privacy: AI systems that generate content based on user data or personal information raise concerns about consent and privacy. Collecting and using user data for generative AI should follow ethical guidelines, including obtaining informed consent and protecting individuals' privacy rights.\n\nEnvironmental Impact: Training large-scale generative AI models requires substantial computational resources, which have an environmental impact. Companies should strive to develop energy-efficient models and explore sustainable computing approaches to minimize the carbon footprint associated with generative AI research and deployment.\n\nAddressing these ethical considerations requires a multidisciplinary approach involving AI researchers, policymakers, ethicists, and society at large. Striking the right balance between innovation and responsible use of generative AI is crucial for harnessing its potential while minimizing the potential risks and ensuring its benefits are shared inclusively. 6","metadata":{}},{"cell_type":"markdown","source":"# Expectations and future of generative ai","metadata":{}},{"cell_type":"markdown","source":"Generative AI has already made significant strides and continues to show great promise for the future. Here are some expectations and potential developments we can anticipate:\n\nEnhanced Creativity: Generative AI will continue to advance in generating creative content across various domains, such as art, music, and writing. We can expect AI-generated works to become increasingly indistinguishable from human creations, blurring the lines between human and machine creativity.\n\nPersonalization and Assistance: Generative AI will enable personalized experiences by generating content tailored to individual preferences and needs. Whether it's personalized news summaries, custom product recommendations, or virtual assistants that adapt to your communication style, generative AI will play a crucial role in enhancing user experiences.\n\nImproved Natural Language Processing: Language models like GPT-3 have made significant strides in natural language processing. Future advancements will likely include better contextual understanding, improved coherence in generating longer texts, and enhanced handling of ambiguous queries. This will have broad implications for customer support, content creation, and language translation.\n\nVirtual Avatars and Characters: Generative AI will facilitate the creation of highly realistic and interactive virtual avatars and characters. These AI-powered virtual entities could be employed in gaming, virtual reality, education, and even customer service, providing dynamic and engaging experiences.\n\nAccelerated Scientific Research: Generative AI can aid scientific research by generating hypotheses, simulating experiments, and analyzing large datasets. This can significantly speed up the discovery process and enable scientists to explore a broader range of possibilities in fields such as drug discovery, material science, and climate modeling.\n\nHuman-AI Collaboration: Generative AI will not replace humans but rather augment their capabilities. We can expect increased collaboration between humans and AI, where AI systems act as creative partners, aiding professionals in various fields like design, music composition, and writing.\n\nImproved Generative Models: The development of new generative models will continue, aiming to overcome limitations such as biases, robustness, and interpretability. Researchers will focus on developing models that are more controllable, adaptable, and capable of generating diverse and high-quality outputs.\n\nDemocratization of Creativity: Generative AI tools will become more accessible to non-experts, enabling people without extensive technical knowledge to create impressive content. This will democratize the creative process, empowering individuals from diverse backgrounds to express their ideas and talents.\n\nSocietal Impact: Generative AI will impact various industries, ranging from entertainment and advertising to healthcare and education. While it presents immense opportunities, there will also be challenges, such as job displacement and concerns regarding privacy and security. Society will need to adapt and establish frameworks to ensure the responsible and beneficial use of generative AI technologies.\n\nOverall, the future of generative AI is promising, with advancements poised to revolutionize multiple aspects of our lives, unleashing new opportunities and transforming the way we create, interact, and solve complex problems. 7","metadata":{}},{"cell_type":"markdown","source":"# Conclusion","metadata":{}},{"cell_type":"markdown","source":"To sum up, generative artificial intelligence (AI) represents a groundbreaking technological advancement that has transformed various fields and opened up new possibilities. Throughout this report, we have explored the capabilities, applications, and evolution of generative AI.\n\nGenerative AI algorithms, such as generative adversarial networks (GANs) and recurrent neural networks (RNNs), have demonstrated the ability to generate realistic and coherent content, including images, music, text, and even human-like conversations. These algorithms learn from vast amounts of data, allowing them to produce novel and creative outputs that often surpass what was previously achievable through traditional programming methods.\n\nThe applications of generative AI are vast and have far-reaching implications across industries. In the field of art and creativity, generative AI has empowered artists and designers to explore new frontiers, inspiring innovation and pushing the boundaries of imagination. In healthcare, generative models have the potential to assist in medical imaging analysis, drug discovery, and personalized treatment plans. Moreover, generative AI has found applications in entertainment, gaming, advertising, and many other sectors, enhancing user experiences and enabling personalized content creation.\n\nWhile the potential benefits of generative AI are significant, it also raises ethical and societal considerations. Issues such as intellectual property rights, authenticity, and the potential for misuse and manipulation require careful attention. As generative AI becomes more powerful, it is crucial to establish clear guidelines and ethical frameworks to ensure its responsible and beneficial use.\n\nLooking ahead, the future of generative AI holds immense promise. Continued advancements in machine learning and deep learning techniques will likely lead to even more sophisticated and capable generative models. As these models improve, they will drive further innovation and impact various aspects of our lives, from entertainment and art to healthcare and education.\n\nGenerative AI has already made remarkable contributions and continues to revolutionize industries and human creativity. It is imperative that we approach its development and implementation with responsibility, addressing the ethical challenges while harnessing its immense potential for the betterment of society. By doing so, we can navigate the transformative power of generative AI and pave the way for a future that embraces the synergy between human ingenuity and machine intelligence.8","metadata":{}},{"cell_type":"markdown","source":"# References","metadata":{}},{"cell_type":"markdown","source":"1,2 statista.com\n\n3,4,5,6,7,8 chatgpt","metadata":{}}]}