Generative AI

When ChatGPT launched in late 2022, it awakened the world to the transformative potential of artificial intelligence (AI). The technology underpinning the powerful new chatbot is one of the biggest step changes in the history of AI—rather than simply analyzing or classifying existing data, generative AI is able to create something entirely new, including text, images, audio, synthetic data and more. Across business, science and society itself, it will enable groundbreaking human creativity and productivity.

From ChatGPT to DALL-E, the latest class of generative AI applications has emerged from foundation models, which are complex machine learning systems trained on vast quantities of data (text, images, audio or a mix of data types) on a massive scale. With recent advances, companies can now build specialized image- and language-generating models on top of these foundation models. Most of today’s foundation models are large language models (LLMs) trained on natural language.

The power of these systems lies not only in their size, but also in the fact that they can be adapted quickly for a wide range of downstream tasks without needing task-specific training. In zero-shot learning, the model uses a general understanding of the relationship between different concepts to make predictions and does not use any specific examples. In-context learning builds on this capability, whereby a model can be prompted to generate novel responses on topics that it has not seen during training using examples within the prompt itself. In-context learning techniques include one-shot learning, which is a technique where the model is primed to make predictions with a single example. In few-shot learning, the model is primed with a small number of examples and is then able to generate responses in the unseen domain.

The introduction of pre-trained foundation models with unprecedented adaptability to new tasks will have far-reaching consequences. According to Accenture’s 2023 Technology Vision report, 97% of global executives agree that foundation models will enable connections across data types, revolutionizing where and how AI is used. To operate in tomorrow’s market, businesses will need to lean on the full capabilities that generative AI provides.

To realize quick returns, organizations can easily consume foundation models “off the shelf” through APIs. But to address their unique needs, companies will need to customize and fine-tune these models using their own data. Then the models can support specific tasks, such as powering customer service bots or generating product designs—thus maximizing efficiency and driving competitive advantage.

As foundation models broaden and extend what we can do with AI, the opportunities will only multiply. Companies will use them to transform human-AI collaboration, ushering in a new generation of AI applications and services. AI models will become our ever-present copilots, optimizing tasks and augmenting human capabilities. Generative AI will bring unprecedented speed and creativity to areas like design research and copy generation. It will take business process automation to a transformative new level, catalyzing a new era of efficiency in both the back and front offices. It will significantly boost productivity among software coders by automating code writing and rapidly converting one programming language to another. And in time, it will support enterprise governance and information security, protecting against fraud and improving regulatory compliance.

With the global AI market size expected to reach nearly $2 trillion by 2030 (source: Statista.com, March 2023), companies will find new ways to leverage these tools to solve complex problems and drive innovation.

Like any nascent technology, generative AI faces its share of challenges, risks and limitations. Importantly, generative AI providers cannot guarantee the accuracy of what their algorithms produce, nor can they guarantee safeguards against biased or inappropriate content. That means human-in-the-loop safeguards are required to guide, monitor and validate generated content. Inaccuracies are known as hallucinations, in which a model generates an output that is not accurate or relevant to the original input. This can happen due to incomplete or ambiguous input, incorrect training data or inadequate model architecture.

Generative AI also raises questions around legal ownership of both machine-generated content and the data used to train these algorithms. To navigate this, it’s important to consult with legal experts and to carefully consider the potential risks and benefits of using generative AI for creative purposes.

Moreover, foundation models possess certain characteristics that render them unsuitable for specific scenarios, at least for the time being. This introduces a whole new level of complexity to security, which is vital to ensure the smooth implementation of transformative technologies. It’s imperative for leaders to incorporate security measures throughout the entire process of designing, developing and deploying generative AI solutions, thereby safeguarding data, upholding privacy and averting misuse. Leaders must brace themselves for the unexpected, as even minor security breaches can result in significant repercussions.

Ultimately, it’s critical that generative AI technologies are responsible and compliant by design, and that models and applications do not create unacceptable business risks. When AI is designed and put into practice within an ethical framework, it creates a foundation for trust with consumers, the workforce and society as a whole.