Artificial intelligence: immersion. Part 3. Development trends: from synthetic data to digital ethics
Progress in the field of artificial intelligence has accelerated significantly over the past year and does not intend to slow down. The reason is that AI technologies are not just a fashion trend, but a powerful tool for improving the operational efficiency of a business. It can be anticipated that some AI models within the next 2-5 years may catalyze transformations across specific industries.
The sectors of artificial intelligence evolution from which the most consequential outcomes for society and business should be expected include:
- Data-Centric AI
- Model-Centric AI
- Application-Focused AI
- Human-Centric AI
Some of these areas are complementary, and some compete with each other.
We tell you what promising trends are and which prominent technology startups are progressing within these spheres.
DATA-CENTRIC ARTIFICIAL INTELLIGENCE
Traditionally, the focus of AI development was to create models that worked out and optimized their algorithms on static data sets. However Data-Centric AI (DCAI) shifts the emphasis, no longer considering data as a constant. At DCAI, AI performance improvement is built on training, enriching the data used for training.
Innovations in data-driven artificial intelligence include synthetic data, knowledge graphs, data labeling, and annotations.
According to Gartner, significant growth awaits the scope of synthetic data. Synthetic data is data that is generated artificially, and not obtained from direct observations of the real world. Various methods for data generation exist, including constructing rigorous samples from authentic data, employing generative adversarial neural networks, and creating simulation scenarios that yield entirely novel sets of event data.
Today, synthetic data in DCAI models are already used in various industries, along with applications in computer vision and natural language applications.
Companies such as Snorkel AI, Modulos, and Shaped work in the DCAI sector.
MODEL-CENTRIC ARTIFICIAL INTELLIGENCE
Model-Centric AI systems focus on selecting the optimal combination of machine learning algorithms, programming languages, and AI platforms. The objective is to create highly efficient machine-learning models. The model-oriented approach has led to significant progress in the field of machine learning algorithms and has given rise to cloud services for building ML models. In this area, tech startups Dataiku and H2O.ai.
Innovations in this sector are in the development of AI using the laws of physics, composite, causal (causal) artificial intelligence, generative AI, basic models, and deep learning.
One the types of Model-centric AI is Composite AI. It is a fusion of various analytical methods, for example, deep learning, natural language processing, computer vision, and descriptive statistics. Since no artificial intelligence method is universal, composite AI has every chance to provide the basis for a significant increase in the range of solved business tasks and increase overall efficiency.
It is expected that in the next five years composite AI will become widespread and will lead to the transformation of the ways of doing business in various industries.
Technologies of Causal AI also tend to go beyond traditional machine learning models. Casual AI does not just determine the relationship of several facts, it identifies the root cause of events and realizes the influence of any variables that could lead to these events. This allows a much deeper understanding of cause-and-effect relationships.
APPLICATION-FOCUSED ARTIFICIAL INTELLIGENCE
Cloud services for AI, autonomous vehicles, operational AI systems, intelligent decision analysis, natural language processing, smart robots, and computer vision are all areas of Application-Focused AI development. In recent months, perhaps the most famous models for natural language processing have become chatbots OpenAI and Jasper.
HUMAN-CENTRIC ARTIFICIAL INTELLIGENCE
In Human-centric AI (HCAI), people are recognized as the only users of products and services, and models take into account human limitations on the use of results: business, moral and ethical. In this area, the greatest return is expected from the development of the category of Responsible AI and digital ethics.
Responsible AI is a term for solutions that integrate business and ethical constraints when implementing AI. The task of such an AI is to maintain a balance of risks and values for business and society. And also responsible AI must be transparent, fair, safe, and comply with legal requirements. According to Gartner analysts, responsible AI will take from 5 to 10 years to become mass, but eventually, it will become a catalyst for transformational processes for business.
Digital ethics is a more short—term trend. It is a system of values and moral principles in electronic interaction between individuals, organizations, and devices. Concerns pertaining to digital ethics, particularly those related to privacy, are of considerable importance to many, and resolutions to these issues are expected to be forthcoming in the near future.