May 30, 2024
Jakarta – Generative artificial intelligence exploded into the public consciousness in 2022, and AI is beginning to take root in the business world in 2023. Now is a pivotal year for the future of AI, as researchers and companies seek to establish how this evolutionary leap in technology will have the greatest impact on the businesses that make up our daily lives.
Generative AI has already reached the “hobby” stage, with the number of openly licensed foundational models growing in the last year. Many open models, powered by fine-tuning techniques and datasets developed by the open source community, can now outperform all but the most powerful closed-source models in most benchmarks, despite having a much smaller number of parameters.
While the ever-expanding capabilities of cutting-edge models will grab media attention, the most impactful developments are likely to be those focused on finding the right business use cases, the ability to customize based on real business needs, and proper governance that makes generative AI trustworthy, sustainable, and accessible to both enterprises and end users.
As we approach mid-2024, here are some key AI game-changers to watch.
First, there are advances in small-scale language models.
Domain-specific models, especially large language models (LLMs), may be reaching a stage of diminishing returns from increasing parameter count. Large models have driven the ongoing golden age of AI, but they are not without their drawbacks. Only very large companies have the funding and server space to train and maintain energy-intensive models with hundreds of billions of parameters.
According to estimates by the University of Washington, training a single model the size of GPT-3 requires the annual electricity consumption of more than 1,000 homes, and the typical daily energy consumption of a ChatGPT query is comparable to the daily energy consumption of 33,000 homes in the U.S. Smaller models, on the other hand, consume much less resources.
These advances in smaller models have three key benefits: they help democratize AI by running at low cost on more readily available hardware; they can run locally on small devices, enabling more sophisticated AI in scenarios like edge computing and the Internet of Things (IoT); and they improve the explainability of AI. The larger the model, the harder it is to pinpoint exactly where and how it makes critical decisions. Explainable AI is essential to understand, improve, and trust the output of AI systems.
The second is Bring Your Own Model (BYOM).
Key to our vision of AI for business is the notion of empowerment: every organization should be able to deploy open AI models to achieve its own goals, and every company has its own regulations to comply with: laws, societal norms, industry standards, market demands, architectural requirements, and more.
One model cannot rule the world. Organizations want the right model for the right use case. They also need to be able to use their own models based on their needs. That's why custom foundational models are growing in popularity as companies look to avoid sending data to third parties and build models tailored to their specific needs. These customized language models can better recognize nuances in language that generic models may miss.
For example, a customer service chatbot may be more effective when customized with data from a company's customer service department. This data helps the model better identify customer needs and respond appropriately. Similarly, a model used to detect fraud may be more effective when customized with data from a company's fraud detection system.
The third is responsible AI.
Without responsible AI and AI governance, organizations cannot adopt AI at scale. While there is no doubt that businesses can gain a competitive advantage by adopting AI, organizations must also consider the return on investment and the risks that AI poses, including privacy, accuracy, explainability, and bias. Generative AI is ushering in a new era, but above all, it must be trustworthy.
Trust is our license to do business and our customers value every decision we make. Organizations cannot afford to break that trust with AI. As an example, a few years ago IBM published high-level principles around trust and transparency.
We believe that the purpose of AI is to augment, not replace, human expertise, judgment, and decision-making. Data and the insights generated from it are owned by their creators, not IT partners. And finally, powerful new technologies like AI must be transparent, explainable, and free of harmful or inappropriate bias if society is to trust them.
Organizations cannot shift the responsibility of AI ethics. CEOs must drive policies and processes that provide transparency and accountability across the board, making clear how and where technology is being used, as well as the source of datasets and underlying models. Business leaders must foster a culture focused on AI ethics, with the goal of optimizing the beneficial impact of AI while mitigating risks and negative impacts for all stakeholders. Responsible AI cannot be an afterthought.
Ultimately, despite all the hype, AI is a long-term game, not a short-term game. Companies need to think carefully about how AI can benefit their business in the long term, rather than adopting it for short-term gains.
By carefully considering the right models that deliver real value to the business, while maintaining proper governance from start to finish of AI adoption, enterprises can scale their business with more agility while ensuring they maintain the trust of all stakeholders.
