*Per Thiago Viola
If 2023 is remembered as the year AI became dominant in the consumer market, it's a safe bet that 2024 will be the year companies follow in droves. But for executives who aspire to not just follow, but truly lead this coming revolution, there are certain principles they should keep in mind.
First, companies must be AI value creators.
There are three distinct ways of consuming generative AI: the first is purchasing software that has generative AI built in, the second is querying pre-existing third-party models via API calls, and the third is creating (and then querying) your own own foundational models that leverage public and private data.
Today, most companies focus on the first two adoption patterns because they represent the easiest path to experimenting and discovering valuable use cases. But while it's nice for consumers to ride the wave as AI users who don't have to pay attention to how it works, forward-thinking companies simply don't have that luxury. They have proprietary information, intellectual property and confidential business data to protect – and ethical, reputational and legal requirements to meet.
Any company that wants to make the most of AI must participate in all value creation opportunities from the underlying models, rather than delegating its capabilities, its strategy and, most importantly, its data to third parties. Base models are, in essence, a new representation of data. This new representation has the power to reveal valuable insights from data. Historically, the invention of each new data representation has proven to be fundamental to the advancement of science and technology, as well as the creation of new businesses. In this case it is no different.
Companies must carefully consider the competitive advantage they will lose by giving up their data to be encoded in a non-proprietary base model, as well as the value of the insights contained in their data. A company that is capable of creating its own AI models (and remember, they don't all have to be huge) is a company that controls its own destiny. By creating such models, you can train, tune and govern your own AI to consistently obtain the maximum return from these evolving technologies – as value creators, the company will have real ownership over protection, control, innovation and monetization of what will become one of your most important assets: business foundation models that encode your most valuable data.
The task of creating or using a basic template aimed at a company may seem daunting, but it is not. That's why we created Watsonx, to empower companies to become value creators, own the source of their competitive advantage, and control their destiny.
Second, business leaders must invest in the community.
It is already evident that wherever AI goes in the coming years, a closed model will not dominate them all. This revolution will be driven by the energy and ingenuity of the entire AI community – a decidedly open community. By integrating a combination of the best open source templates, private templates and, ultimately, their own created templates, companies can put themselves in a position to make the most of that community.
That's exactly what we're doing at IBM by partnering with Hugging Face, a pillar of the open source ecosystem with more than 250,000 AI models shared across its platforms to date. By integrating Hugging Face with our enterprise AI platform, watsonx, we are creating a future for AI that harnesses the creativity and diversity of a broad community to remain open, vibrant, and infinitely customizable.
Take for example climate and Earth science. These areas of business and science continue to be full of bottlenecks, especially with regard to access to the latest information and the ability to analyze this data quickly and efficiently due to their size. NASA estimates that 250,000 terabytes of new data will be online by 2024. How can we solve this?
IBM's response recently was to open source our geospatial AI base model watsonx.ai, built using NASA satellite data on Hugging Face. By putting this model – now the largest geospatial base model on Hugging Face and the first open source AI base model built with NASA – directly into the hands of the AI community, we can use the transformative power of collaboration to improve the way how we protect our planet and its resources.
Third, companies must ensure that their AI can work anywhere and efficiently.
By developing hybrid and open cloud technologies, companies can optimize costs, performance and latency. At IBM, we're making it easier for companies to manage their most valuable data and train, tune, and deploy AI models seamlessly across public and private clouds and on their own premises. The future of these technologies lies in agile, economical and efficient options for companies of all sizes – and the most successful will be those that are prepared to thrive in any environment.
Finally, while business leaders act urgently, they must also act responsibly.
There's no doubt we've reached an inflection point for AI — and executives' instinct to act boldly is a positive one. No one wants to be left behind in the race for these technologies or miss this unique moment full of opportunities.
But in the eyes of customers, investors, employees and each company, a license is needed to operate this exciting new technology: that license is trust. Unless each of us embeds responsible governance at the heart of our use of AI, its new risks will, over time, overshadow its benefits.
This moment, even more than most, demands trusted leadership from the private sector – and, in turn, will reward that trusted leadership. Good AI is governed AI, and for those who hope to lead the way, encouraging this principle in everything they do will go a long way toward consolidating their leadership position.
*Thiago Viola, AI director at IBM Latin America.
Notice: The opinion presented in this article is the responsibility of its author and not of ABES - Brazilian Association of Software Companies