With the astonishing pace of generative Artificial Intelligence (AI) more broadly and OpenAI‘s ChatGPT becoming the world’s fastest-growing consumer application in history with 100 million users in 2 months, we have entered a next phase in the Smart Technology Era where AI is starting to show some more of its true power to comprehensively change the way we work, interact, create, play, and engage with technology, information, and people. It looks to be on course to have far-reaching implications for industries and society, and could potentially transform various job positions. From a rational optimistic perspective, if used in a smart, wise and responsible fashion, it can really amplify our humanity, but there are many challenges and risks that need to be navigated as we’ll also explore in this article.
The recent advancements in generative AI have also revolutionized several industries, such as art, music, fashion, and gaming. These models can perform a range of tasks, including text-to-image, text-to-audio, text-to-3D, text-to-code, and even text-to-science. They can optimize both creative and non-creative tasks, and have enormous implications for the industry and society as a whole. However, there are limitations and challenges that come with these models. One of the biggest challenges is the amount of time and computation power needed to run them, as well as the difficulty in finding adequate datasets. Additionally, these models can be biased due to the data they are trained on, and there is still a lack of understanding around the ethical implications of their use.
Despite these limitations, generative AI has enormous potential for growth and development in the future. As the technology advances, it will be important to address these challenges and limitations in order to fully realize the benefits of generative AI. Furthermore, individuals and businesses should stay informed on the latest developments in the field to take advantage of the opportunities presented by these models. Boston Consulting Group (BCG) has recently published a The CEO’s Guide to the Generative AI Revolution article emphasizing how business leaders should focus on how generative AI will impact their organizations and their industries and what strategic choices will enable them to exploit opportunities and manage challenges. They specifically mention that the choices are centered on three key pillars:
At the upcoming SwissCognitive, World-Leading AI Network virtual conference Redefining Business Performance with Generative AIon March 28, 2023, we’ll be discussing topics such as: what are the generative AI-propelled opportunities across businesses, industries, and domains; asking if generative AI is possibly overhyped or underhyped; and what are the challenges and solutions. These questions will also be examined later in this article.
To get a balanced perspective on Generative AI and its application landscape, Sequoia Capital‘s “Generative AI: A Creative New World” article, which was written by a generative AI large language model (GPT-3) and human collaborators, discusses the four waves in Generative AI that started out with small models (pre-2015), then the race to scale with the Transformer neural network architecture for natural language understanding (introduced by Google Research in the paper “Attention is All you Need“) and OpenAI‘s Generative Pre-trained Transformer (GPT) models (GPT, GPT-2, and GPT-3), followed by the third wave of better, faster and cheaper in 2022, and killer Generative AI applications emerging in 2023 as the fourth wave of which the generative AI large language model ChatGPT is a prime example.
Apart from a plethora of startups focusing on the Generative AI tech stack and applications (both model and application layers for text, code, image, speech, video, 3D, and other), we can expect AI-infused applications from all the tech giants, in particular Microsoft (with the support of OpenAI‘s impressive AI technologies) and Google who will have their own suite of impressive and useful AI-fuelled offerings as well and is already bringing AI into Google’s Workspace. Google’s PaLM large language model is similar to the GPT series created by OpenAI or Meta’s LLaMA family of models. Microsoft has for example announced a new AI-powered tool called Microsoft 365 Copilot, which is currently being tested with selected commercial customers. This tool combines the power of large language models with business data and Microsoft 365 apps to increase skills, creativity, and productivity. This new service will essentially integrate the functionality of a ChatGPT-like large language model with most-used Microsoft applications like Word, Excel, PowerPoint, Outlook and Teams.
OpenAI has also recently released their much improved next generation GPT-4 model to the public on their ChatGPT Plus offering which can solve difficult problems with greater accuracy due to its broader general knowledge and problem solving capabilities. It has more advanced reasoning capabilities, can handle a much longer context (more than 25000 words of text), and is more creative and collaborative with respect to generating, editing and iterating with users on technical as well as creative writing tasks such as writing screenplays, composing songs or learning a user’s writing style. GPT-4 can also accept images as inputs and generate captions analyses and classifications. It is also safer and more aligned by being more likely to producing factual responses and less likely to correspond to request for disallowed content.
Reid Hoffman co-wrote a book with GPT-4 called Impromptu: Amplifying Our Humanity Through AI that offers readers a travelog of the future through exploring how AI, and in particular generative AI Large Language Models like GPT-4, can elevate humanity across key areas like education, business, justice, journalism, social media and creativity. It also explores how we might address risk as we continue to develop AI technologies that can boost human progress at a time when the need for rapid solutions at scale has never been greater.
This message is also aligned with some key thoughts in my book “Democratizing Artificial Intelligence to Benefit Everyone: Shaping a Better Future in the Smart Technology Era” which takes us on a holistic sense-making journey and lays a foundation to synthesize a more balanced view and better understanding of AI, its applications, its benefits, its risks, its limitations, its progress, and its likely future paths. Specific solutions are also shared to address AI’s potential negative impacts, designing AI for social good and beneficial outcomes, building human-compatible AI that is ethical and trustworthy, addressing bias and discrimination, and the skills and competencies needed for a human-centric AI-driven workplace. The book aims to help with the drive towards democratizing AI and its applications to maximize the beneficial outcomes for humanity and specifically arguing for a more decentralized beneficial human-centric future where AI and its benefits can be democratized to as many people as possible. It also examines what it means to be human and living meaningful in the 21st century and share some ideas for reshaping our civilization for beneficial outcomes as well as various potential outcomes for the future of civilization.
So let ‘s dive a little bit deeper. With a little bit of assistance from ChatGPT, let’s briefly explore Generative AI in more detail, the taxonomy of popular generative AI models, its applications, its business opportunities, its challenges and limitations, if it is currently over- and under-hyped, and its future.
What is Generative AI?
Generative AI is a subset of artificial intelligence that allows machines to create and generate new content. Unlike other types of AI, such as machine learning and deep learning, generative AI is focused on the creation of new data rather than analyzing existing data to make predictions.
At the heart of generative AI are generative models, which are mathematical algorithms that generate new data based on patterns learned from existing data. These models are trained on large datasets of images, text, or other types of data, and then use that knowledge to create new content that is similar in style, tone, or structure to the original data.
Generative models can be used for a wide range of applications, from generating images and videos to creating music and poetry. They can even be used to generate realistic human-like conversations or to generate synthetic data for use in scientific research.
There are several different types of generative models used in generative AI, including Variational Auto-encoders (VAEs), Generative Adversarial Networks (GANs), and autoregressive models (of which Recurrent Neural Networks (RNNs) and Transformer Neural Networks are examples). Each of these models works in a slightly different way, but they all share the goal of creating new data that is similar to the original data used to train the model.
In general, generative AI is an exciting field that is rapidly evolving and has the potential to revolutionize the way we think about creativity and innovation. By using generative models to create new content, machines can help humans unlock new ideas, insights, and possibilities that would otherwise be impossible to achieve.
How Does Generative AI Work?
Generative AI is built on a variety of technologies and techniques that work together to create new content. One of the most important technologies used in generative AI is neural networks, which are mathematical models that are designed to simulate the way the human brain works. These networks are used to analyze and learn from large datasets of existing content, and then generate new content based on what they have learned.
Another important technology used in generative AI is reinforcement learning, which involves training a model to learn from its environment by rewarding or punishing it based on its actions. Reinforcement learning is often used in gaming and robotics, where the goal is to teach a model how to make decisions based on a set of rules and objectives.
Generative adversarial networks (GANs) are another key technology used in generative AI. GANs work by pitting two neural networks against each other: one network generates new content, while the other network evaluates that content to determine if it is real or fake. This back-and-forth process helps to refine the generative model over time, resulting in more realistic and high-quality content.
Transformers and large language models have also played a major role in the recent advances of generative AI. These models are based on natural language processing (NLP) and can be trained to generate realistic, human-like language. This has led to the creation of advanced chatbots, language translation tools, and even algorithms that can write news articles or fiction stories.
Training data is also a crucial component of generative AI. To create accurate and high-quality content, generative models must be trained on large datasets of existing content. These datasets can come from a variety of sources, such as publicly available data, user-generated content, or proprietary data. The quality and diversity of the training data can have a significant impact on the accuracy and quality of the generated content.
As can be seen, generative AI is a complex and dynamic field that relies on a range of technologies and techniques to create new content. By combining these technologies in innovative ways, researchers and developers are pushing the boundaries of what is possible with artificial intelligence, and creating new opportunities for creativity and innovation.
A Taxonomy of Popular Generative Models
The January 2023 article ChatGPT is not all you need. A State of the Art Review of large Generative AI models provides, amongst others, a taxonomy of the main generative models published recently. The diagram below shows this taxonomy classified according to their input and generated formats. Over the past two years, a multitude of large generative models, such as Stable Diffusion and ChatGPT, have been introduced, showcasing impressive capabilities in tasks like automatic generation of artistic images and general question answering. Generative AI can effectively and creatively transform text to other texts (e.g. ChatGPT), text to images (e.g. DALLE-2 model), text to 3D images (e.g. Dreamfusion model), images to text (e.g. Flamingo model), text to video (e.g. Phenaki model), text to audio (e.g. AudioLM model), text to code (e.g. Codex model), and even generate scientific texts and algorithms (e.g. Galactica and AlphaTensor models, respectively).
Various Generative AI models have been developed through collaborations between different entities. Notably, Microsoft invested $1 billion in OpenAI to help with the development of their models, while Google acquired DeepMind in 2014. Academic institutions such as KAUST (King Abdullah University of Science and Technology), Carnegie Mellon University, and Nanyang Technological University Singapore collaborated to develop VisualGPT, while Tel Aviv University in Israel developed the Human Motion Diffusion Model. Additionally, some models were developed through collaborations between companies and universities. For instance, Stable Diffusion was developed by Runway, Stability AI, and Ludwig-Maximilians-Universität München, Soundify was developed by Runway and Carnegie Mellon University, and DreamFusion was developed by Google and University of California, Berkeley. The following diagram highlights some of the main generative AI models by notable developers.
Applications of Generative AI
Generative AI has numerous applications across various industries, ranging from art and fashion to gaming and healthcare. Some of the most exciting applications of generative AI include:
AIMultiple has recently updated their top 16 generative AI applications in 2023 article to the Top 70+ Generative AI Applications in 2023 and divides the applications into two main categories: General applications (visual, audio, text-based, code-based, and other) and Industry-specific applications such as healthcare, education, fashion, banking, customer services, marketing, search engine optimization, and human resources.
It is clear that generative AI has the potential to transform many different industries and sectors, and to create new opportunities for innovation and creativity. As the field continues to evolve and advance, we can expect to see even more exciting applications of generative AI in the future.
What are the current applications and potentials of Generative AI in business?
Generative AI has a wide range of potential applications in business. Here are a few examples:
Generative AI has the potential to transform many aspects of business operations, enabling businesses to create new types of content and products, improve efficiency and productivity, and enhance customer experiences. As Generative AI technology continues to advance and businesses explore more of its possibilities, we can expect to see even more transformative applications and innovations in the future.
What are the generative AI-propelled opportunities across businesses, industries, and domains?
The World Economic Forum‘s Strategic Intelligence page highlights the following areas of impact and opportunities for Generative AI (see also diagram below):
Generative AI has the potential to create many opportunities across businesses, industries, and domains. Here are a few examples:
The opportunities created by Generative AI are wide-ranging and diverse, and they will continue to expand as the technology develops and matures. By leveraging the power of Generative AI, businesses can unlock new possibilities and create innovative solutions to complex problems.
How is business performance redefined by Generative AI?
Generative AI has the potential to redefine business performance in several ways. Here are a few examples:
From this we can see that Generative AI has the potential to revolutionize the way businesses operate, improving efficiency, driving innovation, and creating new opportunities for growth and expansion.
Is generative AI Overhyped or Underhyped?
The hype surrounding Generative AI is a topic of debate in the AI community. Some argue that it is overhyped, while others argue that it is underhyped.
On the one hand, some experts believe that Generative AI is overhyped. They argue that the current state of the technology is not as advanced as some of the media coverage suggests. They also point out that many of the applications touted as being powered by Generative AI are actually using other AI techniques, such as supervised learning.
On the other hand, some experts argue that Generative AI is underhyped. They believe that the technology has the potential to revolutionize many industries and domains, but that the full potential of Generative AI is not yet widely understood or appreciated.
In reality, the truth is likely somewhere in between. While it is true that the technology is not yet fully mature and that there are limitations and challenges that need to be overcome, Generative AI has already shown significant promise in a wide range of applications, from content creation and product design to healthcare and scientific research.
When people have limited knowledge about the underlying Generative AI technology and its applications, we also see a lot of hype. The diagram below which views ChatGPT through the lens of the Dunning-Kruger effect (which occurs when a person’s lack of knowledge and skills in a certain area cause them to overestimate their own competence), provides some perspective on how ChatGPT as an example can be on this spectrum of believing the hype and understanding reality. Apart from that, we know that ChatGPT also sometimes goes to the extent of providing an answer that seems grounded and factual, but sometime it is fabricated or hallucinated or just plain misleading.
While it is important to be realistic about the current state of Generative AI, it is also important to recognize its potential and to continue investing in its development and application. With continued innovation and investment, Generative AI has the potential to transform many aspects of our lives and our society.
The Future of Generative AI
The future of generative AI is promising, with new trends and technologies emerging that are poised to drive further innovation and advancement in the field.
Sequoia Capital‘s “Generative AI: A Creative New World” article includes the chart below that illustrates a timeline for how one might expect to see Generative AI models progress and the associated applications that become possible for text, code, images, video, 3D, and gaming applications.
The potential impact of Generative AI on business and society is significant and multifaceted. It will depend on how the technology is developed and deployed. With careful consideration of ethical concerns and risks, Generative AI has the potential to transform many aspects of business and society for the better. Here are a few key ways in which Generative AI could shape our world:
1. Increased efficiency and productivity: Generative AI can help businesses automate repetitive tasks and generate new types of content and products more quickly and efficiently, leading to cost savings and increased productivity.
2. Enhanced customer experiences: Generative AI-powered chatbots and personalized content can help businesses deliver more personalized and engaging experiences to their customers, improving customer satisfaction and loyalty.
3. Creation of new types of content and products: Generative AI has the potential to create new types of content and products that were previously impossible, leading to new business opportunities and innovations.
4. Ethical concerns and risks: As with any new technology, Generative AI raises ethical concerns and risks, such as the potential for bias and misuse. It will be important for businesses and society to consider these risks and develop appropriate safeguards.
5. Job displacement and re-skilling: Generative AI could lead to job displacement in some industries, while creating new job opportunities in others. It will be important for businesses and society to invest in re-skilling programs to help workers transition to new types of work.
One major trend that we are seeing is the increasing use of generative AI in areas such as healthcare and education, where it can be used to create personalized content and improve learning outcomes.
Another trend that is emerging is the use of generative AI in robotics and automation. As robots become more advanced and sophisticated, they will need to be able to generate their own ideas and solutions to complex problems. Generative AI can help to enable this by allowing robots to generate new ideas and strategies based on their environment and objectives.
In addition to these emerging trends, there are also new technologies being developed in the field of generative AI. For example, researchers are working on developing more advanced generative models that can generate highly realistic and accurate content, such as images and videos.
However, the increasing use of generative AI also raises concerns about its potential impact on the job market and society as a whole. As generative AI becomes more advanced and capable, it may lead to the automation of many jobs, particularly those that involve repetitive tasks or data analysis. This could have significant implications for the workforce, and may require new policies and regulations to address.
Moreover, generative AI also raises ethical concerns, such as the potential for the creation of deepfakes or other forms of misinformation. As such, it is important for researchers and developers to prioritize ethical considerations and ensure that generative AI is used responsibly and ethically.
The future of generative AI is likely to be shaped by a combination of emerging trends, new technologies, and ethical considerations. As the field continues to evolve and develop, we can expect to see further advances in generative AI that have the potential to transform many different industries and sectors.
Challenges and Limitations of Generative AI
Despite the many potential benefits of generative AI, there are also several challenges and limitations that must be addressed. Some of these challenges include:
Herewith some potential solutions to some of the challenges:
The challenges and limitations of generative AI highlight the need for responsible and ethical use of the technology. Researchers and developers must prioritize issues such as bias and ethics, and work to address the challenges associated with creating truly original and high-quality content. At the same time, policymakers and society as a whole must also be aware of the potential risks associated with generative AI, and work to ensure that it is used in a responsible and ethical manner.
Generative AI has the potential to revolutionize a wide range of industries, from art and music to healthcare and finance. By using advanced algorithms and neural networks, generative models can create new and innovative content that was previously impossible to generate. However, as with any technology, there are also challenges and limitations that must be addressed, such as bias and ethics, the difficulty of creating original content, and the risks associated with its use.
Despite these challenges, the potential of generative AI is vast, and we can expect to see continued growth and innovation in the field in the years to come. Individuals and businesses can stay up-to-date on the latest developments by following industry news and attending conferences and seminars. By staying informed and taking advantage of the opportunities presented by generative AI, we can work together to build a better and more innovative future.
Democratizing Artificial Intelligence to Benefit Everyone: Shaping a Better Future in the Smart Technology Era
See also the Democratizing AI Newsletter: https://www.linkedin.com/newsletters/democratizing-ai-6906521507938258944/