How Generative AI Is Changing Industries
Giovanni Sisinna
??Portfolio-Program-Project Management, Technological Innovation, Management Consulting, Generative AI, Artificial Intelligence??AI Advisor | Director Program Management @ISA | Partner @YOURgroup
Thank you for reading my latest article “How Generative AI Is Changing Industries”.
The purpose of this article is to provide an overview of the use of generative AI in various fields, including art, music, film and writing. It explores its potential applications and limitations, as well as ethical and regulatory considerations and potential impact on the environment
1. Introduction to Generative AI
Generative AI is a rapidly advancing area of artificial intelligence that is transforming how machines can create and learn. The term "generative" refers to the ability of an AI system to create new and original content by using complex algorithms to analyze patterns in data and generate new data that fits those patterns. The purpose of Generative AI is to enable machines to create content that is both novel and relevant to human users, allowing for the creation of entirely new forms of art, music, writing, and more.
At its core, Generative AI relies on neural networks, which are designed to mimic the way the human brain works. These networks are trained on vast amounts of data, allowing them to recognize patterns and generate new content that is similar to what they have learned. This process is what enables Generative AI systems to create content that is similar to what humans might produce, but with unique variations and twists.
Generative AI is rapidly transforming the world of content creation that wase previously thought to be the sole domain of humans. It has a wide range of applications, from creating new works of art to generating music, writing, and even entire films. Companies like Google, Facebook, Microsoft and OpenAI are investing heavily in this area, with new systems and applications being developed all the time. One of the most well-known examples of Generative AI is ChatGPT, which can generate text on any topic from a few prompts, making it a powerful tool for content creation.
Despite its potential, Generative AI also raises a number of ethical concerns. There are worries that these systems could be used to create fake news or other types of propaganda, as well as concerns around the impact on human creativity and job displacement. There is also a growing concern around the carbon emissions generated by these systems, with some estimates suggesting that they could have a significant impact on the environment.
2. Technical Aspects of Generative AI
Generative AI models are built to generate outputs that are novel and unique based on the input data that they are trained on. These models work by learning the underlying patterns and relationships between the input data and then using that information to generate new outputs that are similar to the input data but different enough to be considered novel.
Generative AI is a subset of machine learning, a type of artificial intelligence that enables computers to learn from patterns in data without human intervention. Generative AI is a recent breakthrough that goes beyond pattern recognition and can produce content on demand.
There are different types of generative models, such as Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs). GANs consist of two neural networks that compete against each other in a game-like manner: a generator that generates data and a discriminator that evaluates how realistic the generated data is. VAEs, on the other hand, work by compressing the input data into a lower-dimensional representation and then generating new outputs based on that representation.
Text-based machine learning models like ChatGPT, GPT-3, and BERT, are trained using self-supervised learning. In this method, a model is fed a massive amount of text to generate predictions based on patterns in the data. These models can accurately predict the end of a sentence or produce high-quality written content in seconds.
The technical aspects of Generative AI are complex and involve a number of key components, including neural networks, loss functions, and optimization algorithms. The neural networks are the core building blocks of generative models, and they are responsible for learning the underlying patterns and relationships in the input data. The loss functions are used to evaluate how well the generative model is performing, and optimization algorithms are used to adjust the model's parameters to improve its performance over time.
Developing a Generative AI model is an expensive and resource-intensive process that only well-resourced tech giants have attempted.
3. Is Generative AI Really Creative?
There is much debate about whether Generative AI can truly be considered creative. Some argue that the technology is simply reproducing patterns and structures that it has been trained on, rather than exhibiting genuine creativity. Others, however, point to recent examples of Generative AI creating novel and innovative content.
In a recent article (“There’s No Such Thing as ‘Generative AI’”, University of Cambridge) Harry Law suggests that the lack of consensus on what constitutes Artificial Intelligence and the origins of machine learning make it difficult to define Generative AI accurately.
Generative AI is not “truly artificial intelligence”, according to Law. It is a machine learning system that relies on humans to produce outputs. Generative AI systems are trained on vast corpuses of data created by humans and will require human input for widespread adoption. This is because the machine learning systems underpinning Generative AI have no understanding of the outputs they produce. They can only predict the next output in a sequence based on the historical data they have been trained on. This lack of understanding is not just a problem for today's machine learning systems but also for future systems that will be trained on their outputs.
Law also notes that Generative AI is not new. Computer systems capable of generating text and attempting the Turing test have existed since the mid-20th century. The recent popularity of Generative AI can be attributed to enthusiastic reporting and the AI industry's ballooning budgets.
Finally, Law highlights the ethical issues surrounding Generative AI. The creation of Generative AI models requires vast amounts of data, much of which is labeled by low-paid workers in the Global South. Generative AI models also require human oversight to ensure their reliable operation, which will lead to a dramatic expansion of the number of workers required to oversee these systems. The reliance on human labor in the creation and operation of Generative AI models is a cause for concern, given the potential for exploitation.
4. Generative AI Race
In the world of technology, there is always a new innovation that captures people's imagination. Currently, the talk of the town is ChatGPT. It has drawn over one million users within five days of its launch, making it one of the fastest consumer-product launches in history.
Microsoft has invested $10 billion in OpenAI, with the intention of integrating ChatGPT-like capabilities, such as generating text, images, and videos that appear to have been created by humans, into its software. Meanwhile, Google has released a similar model that creates music from a text description of a song. Other companies, such as Baidu, are also working on developing their own chatbots.
In a recent article (“The race of the AI labs heats up. ChatGPT is not the only game in town” Jan 30th, 2023) The Economist affirms that the race for AI supremacy is heating up, and the results of this race will determine the pace of the AI age and who will dominate it.
Corporate research and development organizations have traditionally been the source of scientific advances in America. However, over the years, firms have moved away from basic science towards developing existing ideas due to the rising cost of research and the increasing difficulty of capturing its fruits. Nonetheless, AI is once again shaking things up by transforming how the tech industry thinks about innovation and its engines, which are the corporate research labs. These labs, whether part of big tech firms or independent startups, are in an epic race for AI supremacy.
Recent breakthroughs in big AI globally have come mostly from giant companies because they have the computing power and because this is a rare area where results of basic research can be quickly incorporated into products. Amazon, Meta Alphabet (with its controlled Google Research and DeepMind), OpenAI, Microsoft and Baidu are only some of the companies that produce considerable AI research
The Chinese labs appear to have a big lead in the subdiscipline of computer vision, which involves analyzing images. The Beijing Academy of Artificial Intelligence (BAAI) has built what it says is the world's biggest natural-language model, Wu Dao 2.0. While expert opinion varies on who is ahead on the merits, no model enjoys an unassailable advantage. AI knowledge diffuses quickly, and researchers from competing labs often move between organizations, bringing expertise and experience with them. Additionally, since the best AI brains are scientists at heart, they often made their defection to the private sector conditional on a continued ability to publish their research and present results at conferences.
In terms of the sort of AI that ChatGPT is popularizing, the big battle is between Microsoft and Alphabet. The Economist conducted tests on both companies' AI systems and found that neither AI emerged as clearly superior for the reasons outlined above.
One variable that may determine the ultimate outcome of the race for AI supremacy is how labs are organized. OpenAI, a small firm with few revenue streams to protect, may find itself with more latitude than rivals to release products to the public. Insiders note that OpenAI's rapid progress in recent years has allowed it to poach experts from rivals including DeepMind. To keep up, Alphabet, Amazon, and Meta may need to rediscover their ability to move fast and break things.
Size has been everything so far in Generative AI. However, there are limits to how big the models can get, and ways to fine-tune a model to a specific task that dramatically reduce the need to scale up are being developed all the time. Novel methods to do more with less are being developed, and capital is flowing into Generative AI startups. Venture capitalists are betting that not all the value will be captured by big tech. The AI race is just getting started.
The developments in AI are transforming not just the tech industry, but also business, culture, and society. These innovations have the potential to significantly impact the future, and companies that are at the forefront of the AI race will have a significant advantage. While it is too early to determine the full extent of AI's impact, one thing is clear: companies that invest in AI and develop the most advanced models will be better positioned to succeed in the future.
5. Environment Impact
In a recent article ("The Generative AI Race Has a Dirty Secret" Wired, February 10, 2023) Chris Stokel-Walker argues that Generative AI requires significant amounts of computing power, which has led to concerns over its environmental impact.
Tech giants Google and Microsoft recently announced major overhauls to their search engines, incorporating generative artificial intelligence tools that use Large Language Models (LLMs). This move is expected to offer users a more accurate and richer search experience. However this comes with a dirty secret - a dramatic rise in computing power and energy requirements, leading to huge carbon emissions.
Training LLMs such as those underpinning Microsoft's souped-up Bing search engine and Google's equivalent, Bard, requires parsing and computing linkages within massive volumes of data, requiring companies with sizable resources to develop them. According to Carlos Gómez-Rodríguez, a computer scientist at the University of Coru?a in Spain, only Big Tech companies currently possess the necessary computational power to train LLMs. Third-party analysis estimates that training GPT-3, which ChatGPT is based on, consumed 1,287 MWh and led to emissions of more than 550 tons of carbon dioxide equivalent.
领英推荐
Integrating ChatGPT into Bing, which handles half a billion searches every day, would require at least four or five times more computing power per search at a minimum, according to Martin Bouchard, cofounder of Canadian data center company QScale. Furthermore, the current data centers and infrastructure in place would not be able to cope with this, requiring significant investments in hardware. This move could increase data centers' greenhouse gas emissions, which already account for around one percent of the world's greenhouse gas emissions, according to the International Energy Agency.
To address this issue, companies running search engines could reduce their net contribution to global heating by moving data centers onto cleaner energy sources and redesigning neural networks to become more efficient. This would reduce the inference time required for large models, minimizing the amount of computing power needed for an algorithm to work on new data.
Google spokesperson Jane Park suggests that combining efficient models, processors, and data centers with clean energy sources can reduce the carbon footprint of a machine learning system by as much as 1,000 times. Microsoft has committed to becoming carbon negative by 2050, while Google aims to achieve net-zero emissions across its operations and value chain by 2030.
Despite the potential environmental impacts, Nafise Sadat Moosavi, a lecturer in natural language processing at the University of Sheffield who works on sustainability in natural language processing, suggests that the move to incorporate AI-powered search engines is beneficial for end-users. Moosavi emphasizes the importance of focusing on the efficiency aspect and reducing the amount of energy and carbon generated by LLMs.
6. Text Generation
Generative AI is increasingly being used to generate text-based content such as articles, books, and other written materials. The technology is capable of generating coherent and grammatically correct text that can be used for a wide range of applications, from writing articles to chatbots and virtual assistants. One of the most well-known examples of Generative AI for text generation are GPT models, such as GPT-3, which can generate highly convincing human-like text based on a given prompt.
The release of ChatGPT has led to concerns about the end of the study of the humanities, and human-generated journalism. Furthermore, all the major tech companies are developing comparable word-spinning technologies to integrate into the tools we use every day, such as search engines and email and word processors.
In a recent article (“How AI Will Change Writing Forever”, Jan 20, 2023) Ann Kjellberg affirms that in 2023, significant changes are happening in the way writing is produced and disseminated.
The advantages of using ChatGPT and others Generative AI for text generation are numerous and will transform writing across society, from universities to newsrooms.
However, there are also limitations and concerns to using Generative AI for text generation.
Despite the potential of AI-generated text to replace human writers, some working writers are tentatively enthusiastic about the possibilities. Others argue that a computer could never replace a writer in the creation of actual literary art.
7. Music Generation
Generative AI is transforming the music industry, with many artists and producers using AI tools to create music that would have been impossible to produce without it. AI is being used to generate music in a variety of genres, from pop and hip-hop to classical and jazz.
One of the key advantages of Generative AI in music is its ability to allows artists to experiment with different sounds and styles quickly and easily, which can lead to new and innovative music. AI-generated music can also be used in film and TV scores, video games, and other media projects, where it can be a cost-effective and efficient way to create original music.
While some experts predict that AI will revolutionize the music industry, others are concerned about the potential impact on musicians and their creative output. It also raises concerns about the role of AI in music creation and whether it will displace human musicians.
Google's new Generative AI system, MusicLM, has raised questions about the potential impact of AI-generated music on the musical industry.
In a recent article (“Is Google Displacing Musicians with Its New Generative AI System: Music LM?”, Forbes, Jan 29, 2023) Cindy Gordon affirms that the system, which has not yet been released due to copyright issues, is built on a neural network and trained on a large music data set of over 280,000 hours of music. It can automatically produce innovative music tracks of diverse instruments, genres, and concepts based on text descriptions such as "a calming violin melody backed by a distorted guitar riff."
While AI-generated music has been around for some time, MusicLM promises higher quality fidelity audio that remains consistent over several minutes. Researchers from Google have also published an AI training dataset of 5,500 pieces of music to support other researchers working on automated song generation. However, the use of AI in music raises some difficult questions about ownership and creativity.
One of the main concerns is the risk of AI algorithms creating their own compositions and work, and who owns this work, the AI or the human. Furthermore, who owns the music when it's a blend of everything on the world wide web and creates a new song from the brilliance of our musicians? When you purchase music, are you also purchasing the right to use its audio as an AI training data? These are complex questions that require careful consideration and effective policy and legal legislation.
Musicians globally are trying to understand the impacts of AI on their industry. YouTube sensation and American Idol contestant, Taryn Southern, has already started composing music with AI. As AI continues to advance, it's imperative that the musical industry improves its legislation in regard to ownership of musicians' music and how AI algorithms should be treated and managed in the industry.
8. Image and Video Generation
Generative AI is also being used to create images and videos, with the potential to revolutionize the media and entertainment industries.
AI image generation is the process of creating images through AI algorithms. These technologies work by analyzing large datasets of visual information, and then generating new content based on that analysis. This is achieved through a process called diffusion, which interprets noise to create recognizable shapes using a collection of data.
Despite the potential benefits of Generative AI for image and video generation, there are also ethical concerns to consider.
There are many AI tools that can be used for Image Generation.
DALL-E 2, Midjourney, and Stable Diffusion are three popular AI image-generation tools that generate images through text-to-image prompts or image-to-image prompts. DALL-E 2 is a paid service, accessed through the OpenAI website, and generates photo-realistic images. Midjourney works entirely through Discord and is best for generating environmental concept art. Stable Diffusion is an open-source model that can be locally hosted or accessed through services like Dream Studios and is best suited for designers, artworkers, and producers.
While DALL-E 2 and Midjourney are best used for exploring different art styles and visual concepts in a short amount of time, Stable Diffusion's biggest benefit is making new iterations and artwork options to existing artwork through processes known as inpainting and outpainting. Inpainting replaces sections of existing artwork with new references, allowing for quick and efficient changes.
Creatives should introduce AI image generation tools into the pre-production stages of storyboarding, artwork, and concept development. These tools can improve workflow, free up internal resources, and ultimately reduce costs. However, it is worth noting that AI tools can be intimidating for first-time users, and results can vary significantly depending on a variety of factors, such as the unique differences between AI image generation tools and the quality of the prompts fed into the AI.
9. The Future
In conclusion, Generative AI offers many benefits, and its applications span multiple sectors but, on the other hand, it also generates significant challenges that need to be addressed to ensure its responsible development and use.
With the progress of technologies, Generative AI will in fact allow the creation of even more advanced systems. However, it is important to proceed with caution, bearing in mind the ethical, social, and environmental implications of these technologies. In this way, we can ensure that Generative AI is developed and used in ways that benefit society as a whole.