Generative AI is ushering in a new era of digital transformation, offering the potential to revolutionise everything from product development to customer engagement and operational efficiency.
In Professor Panos Constantinides’s latest blog, he delves into the key insights and strategies that senior leaders should harness to navigate this evolving landscape.
This article summarises the key takeaways from Panos’s insightful and thought-provoking blog.
The Benefits of Building Generative AI Models from Scratch
Studies show that there are significant benefits to building a proprietary generative AI model from scratch including improved data protection over sensitive data while using trusted and verified sources, better control over infrastructure response times, processing and scaling decisions, and improved anticipation of market trends to personalise customer experiences with renewed business models.
For example, Bloomberg used a 700 billion token corpus of data (363 billion token dataset consisting of English financial documents, plus a 345 billion token public dataset) to train a 50-billion parameter model. The Bloomberg model was found to perform better than existing open models, while still performing on par or better on general benchmarks.
The Costs of Implementing Generative AI
While there is a long list of strategic advantages of building and buying generative AI models – both open-source and proprietary – there are also substantial costs involved including computational power and data acquisition costs, human resource, and integration costs.
For example, training general-purpose AI models like OpenAI’s GPT-3 demands extensive computational power and access to vast datasets, which can be expensive for many organisations.
Most organisations will find it difficult to navigate this complex landscape of technological solutions and will often depend on third-party solutions provided by big tech companies. This only feeds into existing inequalities in the competitive dynamics for generative AI markets.
Strategic Recommendations
The analysis of the costs of generative AI points at the need for better collaboration and coordination between organisations, while at the same time protecting their unique assets, most notably data assets. Three long-term strategic decisions are proposed:
- Organisations should build their own proprietary generative AI model to fit their domain-specific business operations. Such models can be built on smaller parameter open-source models that are more cost efficient.
- License general-purpose generative AI models like Microsoft Copilot. General-purpose generative AI models can be most effective when used for routine tasks, benefiting from their large-scale optimisations and eliminating substantial upfront costs and technical complexities.
- Integrate the above with an orchestration framework. The key objective would be to generate data on the general-purpose model as outputs and then leverage that data as inputs for the proprietary model, while also ensuring interoperability between existing systems.
Implementation Considerations
The above strategic recommendations require several implementation actions:
- Build and protect proprietary datasets that are relevant and uniquely beneficial to strategic needs.
- Form strategic partnerships to mitigate and share the financial and technical burden of AI development.
- Develop digital governance frameworks to facilitate the integration of generative AI with existing systems and operations.
- Attract new and reskill existing employees to enhance workforce adaptability and maximise the productivity and innovation potential of AI technologies.
- Build ethical and legal frameworks to ensure compliance and applications of responsible AI.
By focusing on these strategies, senior leaders can unlock the transformative potential of generative AI to drive innovation, enhance efficiency, and maintain a competitive edge while navigating ethical, legal, and operational challenges.
Are you ready to lead your organisation into the digital future? Our intensive 4-day course, Leading Digital Transformation delivered by Professor Panos Constantinides, is designed for senior leaders who are eager to harness the power of digital technology to drive innovation and efficiency. Join us and become the visionary leader your company needs in the digital age.
