Katie Madding is Chief Product Officer at Adjust, a measurement and analytics company that helps marketers grow their apps across platforms.
getty
As competition among Big Tech continues to rise, the true value of a company increasingly hinges on its data. As generative AI (GenAI) becomes central to innovation, many companies that had not previously considered entering the GenAI race are quickly realizing that public data is not just an asset; it’s the engine that powers the future of their business.
For Fortune 500 companies and tech giants, the race is no longer just about participating in the GenAI race (that’s inevitable) but also strategically deciding how to harness the power of large language models (LLMs). These models are trained on vast datasets and are the core of GenAI capabilities, enabling companies to generate, predict and innovate in ways that were previously unimaginable.
Apple’s recent decision to leverage OpenAI’s LLM instead of developing its own represents a pivotal moment for the tech industry: the build, buy or borrow dilemma. Do they build proprietary models, buy them from established players or borrow from open-source platforms?
This decision is critical because it determines how they capitalize on their data and intellectual property (IP) while balancing the risks of security, cost and control.
Build: The In-House Approach
By opting to develop LLMs internally, companies have the advantage of customizing and training them to their specific use cases and seamlessly integrating them within existing tech stacks. This tailored development process ensures that the AI solution aligns perfectly with the company’s goals and operations.
Another key benefit is that a company has full control over its data. Data ownership is crucial as it ensures that data remains confidential and allows a company to own its intellectual property, turning the AI system into a highly valuable asset.
But building an LLM in-house is not without its challenges. This approach requires significant investment—not just financially but also in terms of data, talent and infrastructure. Companies that opt to build are often those with a lot of money to spend, an abundance of data and a core business model that depends on creating new content or responses. These companies prioritize owning their IP and ensuring that their data remains confidential.
However, the costs involved are staggering; OpenAI’s GPT-3 reportedly cost over $4.6 million to train. Not to mention the environmental impact, as training AI models can consume more energy than 100 typical U.S. homes in a year.
Furthermore, companies that develop their own LLMs bear the full weight of legal and ethical responsibilities. Legislators around the world are rushing to put laws in place to manage the operation of AI technologies—and the burden of ensuring compliance with new and evolving regulations would be solely on the company.
Customizing smaller LLMs for specific use cases is emerging as a viable middle ground. Tailored models allow companies to address particular business needs without the overhead of building a massive LLM from scratch. However, even these require careful consideration of governance, risk management and the underlying data infrastructure.
Buy: Purchasing AI Solutions
Purchasing an LLM is a faster route to market. Companies can quickly implement ready-made LLMs without the significant time and financial investment required for in-house development and training. This approach allows companies to incorporate tried and trusted technologies from established providers, reducing the risks associated with developing and deploying their own LLM.
On the other hand, companies that choose to buy or borrow an LLM must navigate the risks associated with third-party services. These include the potential release of sensitive information, reliance on external platforms and the lack of visibility into how these models are trained and updated. There’s also the risk of "hallucinations" from these models—errors that could be attributed to the company buying or borrowing, potentially damaging its reputation.
Borrow: Partnering With AI Providers
Partnering with AI providers, like Apple’s collaboration with OpenAI, offers a middle ground between building and buying. It allows companies to access innovative technology and expertise without taking on the full responsibility of development. Partnerships can offer greater flexibility than outright purchases, resulting in solutions and collaboration that can be beneficial for both parties.
Legal and ethical responsibilities can also be shared, meaning the sole burden of navigating complex regulations isn’t solely on the company borrowing.
However, AI partnerships come with their own set of challenges. They often involve complex contractual agreements and careful negotiation of terms relating to data sharing, IP rights and responsibilities. Ensuring that the partners’ goals are aligned can also be challenging and conflicts can be detrimental to the partnership.
Another limitation of the "borrow" approach is that the company’s end users—whether they are consumers or other businesses—may have policies in place to prevent their data from being shared.
Additional Challenges And Opportunities
The legal landscape for GenAI is still evolving, with conflicting regulations between different regions and countries. Companies must navigate this environment carefully, balancing innovation with compliance and ethical responsibility. Apple’s decision to leverage OpenAI’s LLM can be seen as a strategic move to mitigate legal risks. However, it also requires ensuring that OpenAI’s practices align with Apple’s standards.
Data privacy and security considerations are critical in GenAI development. Regulations such as the General Data Protection Regulation (GDPR) in Europe and the California Consumer Privacy Act (CCPA) in the U.S. impose strict requirements that companies must follow.
Outsourcing AI development can reduce the direct handling of sensitive data, potentially lowering the risk of noncompliance. Nevertheless, companies are still accountable for the data shared, so they must ensure that their partners adhere to all applicable privacy regulations.
Build, Buy Or Borrow: A Strategic Move
With estimates that GenAI could add between $2.6 trillion and $4.4 trillion in annual value to the global economy, the opportunities in this evolving space are extensive. For tech giants and Fortune 500 companies, the ability to effectively leverage LLMs will determine who reaps the benefits.
The decision to build, buy or borrow is not just a technical choice; it’s a strategic move that could define the future of these companies in the GenAI era.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
{Categories} _Category: Takes{/Categories}
{URL}https://www.forbes.com/councils/forbestechcouncil/2024/09/20/the-genai-tech-debate-build-buy-or-borrow/{/URL}
{Author}Katie Madding, Forbes Councils Member{/Author}
{Image}https://imageio.forbes.com/specials-images/imageserve/66ec40c7c01413827c2ee815/0x0.jpg?format=jpg&height=600&width=1200&fit=bounds{/Image}
{Keywords}Innovation,/innovation,Innovation,/innovation,technology,standard{/Keywords}
{Source}POV{/Source}
{Thumb}{/Thumb}
