The Development Pattern Of AI Is Evolving From “Data” To “Knowledge”

Original article was published by Md Ashikquer Rahman on Artificial Intelligence on Medium


Photo by Istocphoto

Since the artificial intelligence (AI) revolution was born more than half a century ago, it has brought a huge impact to the entire world. Especially in the past ten years, AI has been transformed from the research direction of the academic field into an indispensable part of our daily life. Today, the AI ​​business strategy we take for granted is mainly built around data, and proprietary data has even become the most strategic resource reserve for AI companies. But in the next few years, proprietary data will no longer be a unique asset, which means that the differentiated advantage based on proprietary data will continue to decline at the sustainable level. Therefore, the entire world is likely to transform from a data-based AI strategy to a knowledge-based AI strategy.

The development of big data benefits from the deployment of many sensors, the popularization of Internet connections, and the substantial improvements in computing power, communication capabilities, and digital storage. This also makes AI technology training transformed from small academic research projects to large-scale Enterprise production-level applications. In essence, big data requires complex AI models to analyze and extract knowledge and insights from them, and these AI models require massive amounts of big data for training and optimization. Therefore, AI companies often regard data as an important strategic reserve, and this trend has become more common in the field of venture capital. In fact, many start-ups have recently made data collection the core of their business strategy. More and more similar manufacturers are beginning to emphasize the unique data set they have and the long-term strategy to further obtain other proprietary data, and regard this as a sustainable barrier to employment. In addition, as AI tools and AI-as-a-service platforms have enabled AI model development to quickly enter the commercialization stage, coupled with the continuous emergence of public open data sets, people’s need to establish and defend their own data fence has become more and more significant.

In today’s technology ecosystem, whoever has more advanced AI programs and who has more control over proprietary data can get more returns from the market. This is also seen as a huge and sustainable competitive advantage. Manufacturers represented by Google and Netflix have developed and sorted out large-scale authoritative data sets in their operations for many years. Countless other companies are full of admiration to follow them and hope to replicate their success. However, in the face of Netflix’s complex and sophisticated data strategy, competing media service providers and drama production companies simply cannot match it.

However, with the expected increase in data exchange capabilities and willingness to exchange, we believe that in the next ten years, the barriers to entry established by proprietary data may not be sustainable. Although data will still become an important fuel for the AI ​​value engine, knowledge will become more and more important in AI business strategies.

Push the AI ​​value pyramid up to the knowledge level

If there is an AI value pyramid, then its base part is undoubtedly data, and the higher it goes, the greater the proportion of knowledge. Now that we are in an era where “information is at your fingertips, but knowledge is hard to find”, it has become a necessity to push the AI ​​value pyramid to the knowledge level.

In fact, we have seen numerous data exchange initiatives aimed at promoting and accelerating this trend. We hope to exchange commercialized data sharing in exchange for valuable knowledge and even business feasibility. In short, data will become more abundant, available, reliable, standardized and low-cost, and all this also means that data will become a typical commodity. On this basis, the concept of using data as a barrier to entry will also be self-defeating.

With the proliferation of Internet of Things (IoT) devices, the feasibility of data sharing will also reach new heights. In addition, emerging technologies, protocols and standards for data consolidation, sharing and exchange will also keep up in time. Looking to the future, as long as there is a clear motivation and willingness, the ability to share data will also become an important advantage. As the entry barrier of data collapses under the impact of AI technology, more organizations will tirelessly collect their own proprietary data and use it as an important commodity. Of course, the acquisition and use of this data is still quite difficult, and the return may not be obvious, so it may cause distortions at the strategic level. This is because although most organizations have regarded AI as part of their business system, AI is still not part of traditional skills or core expertise. In addition, the long-term absence of AI training engineers, developers, product owners, and managers will also exacerbate this strategic imbalance and ultimately make data sharing solutions aimed at knowledge exchange widely recognized by the market.

The EU’s recent initiative to generate knowledge through data exchange is a typical case of combining creativity and willingness to cooperate. They hope to establish a “single data market” to help individuals, companies and other organizations use non-personal data as materials to extract insights and make better decisions, so as to compete with current mainstream technology giants.

Another big factor that impacts the sustainability of proprietary data is the emergence of new data solutions. Such solutions can implement model training using relatively small data sets. Synthetic data solutions (such as generative adversarial networks) and other sample minimization techniques (such as data augmentation) are expected to enable companies to build disruptive AI products without a lot of data.

Establish a knowledge development strategy

The future of the AI ​​revolution will reshape the real market that companies rely on to survive, so we must establish a targeted business strategy. The transition from data to knowledge will also bring new frameworks, partnerships and business models, including all parties that provide data, information, AI models, storage and computing capacity for knowledge creation. Faced with this unprecedentedly vast market, companies should set out as soon as possible to formulate a development strategy that focuses more on knowledge elements:

• Establish knowledge reserves to replace data reserves, and regard this basic principle as the core of future business strategies. Enterprises and organizations should prepare for the knowledge-centric era-in this new era, whoever can ask the right questions, find the most relevant prediction results and design the most disruptive AI application program Can occupy the commanding heights of market competition.

• Use AI technology in a top-down manner to organize business systems around application and product layers. AI models should be developed and trained based on specific vertical industries and assumptions. For example, develop specific healthcare applications based on imaging, diagnosis, telemedicine, pharmacology and other clinical applications; or build traffic management systems for fleet management, public transportation, and other traffic participation factors. The development of such solutions requires us to combine rich knowledge based on specific fields with practical experience, while matching contextual information with well-tuned AI models.

• The data acquisition plan will only be a short-term tactical pursuit, while knowledge-based exchanges and partnerships are more long-term business strategies that are worth training. Last year, the Israel Innovation Authority launched a pilot program to achieve knowledge-based business cooperation between hospitals and technology start-ups. This cooperation established dozens of specific projects between startups and hospitals, promoted the active exchange of raw (and almost unusable) data between hospitals, and helped startups accumulate new and valuable knowledge. .

• Finally, the knowledge-oriented business transformation should also affect the human resource strategy within the organization. Companies should formulate appropriate and wise human resource management strategies for future AI development. Although some start-up companies still need to invest heavily in recruiting data engineers and scientists, the most ideal way should be to design the AI ​​team as a management team, responsible for establishing and promoting AI knowledge partnerships, inventing AI-based applications/products, and Make creative explorations on the bright prospects of the AI ​​revolution. All of this essentially represents a redesign of the architecture from data-centric to knowledge-centric. In addition, the AI ​​team should also help people understand the context in which they operate. The most important point is to ensure that each team member makes full use of their understanding of AI and specific functional areas through a holistic approach, instead of just playing the role of regular AI experts.

In summary, the future of AI depends on the shift from emphasis on proprietary data sets to sharing data and creating knowledge across entities. In order to successfully implement related AI strategies, companies must correctly combine data, information, AI models, storage, computing capacity and other elements to ensure that their business is deeply rooted in knowledge, the most important and core differentiated resource.