Gaurav Tewari, founder and Managing Partner of Omega Venture Partners.
Generative artificial intelligence (GAI) is projected to add $2.6 trillion to $4.4 trillion in economic value, according to research by McKinsey. Solutions such as ChatGPT, DALL-E, GitHub Copilot and Bard have flooded into our business, education and creative workflows at an astonishing rate. Unsurprisingly, contemporary discourse relating to the power unlocked by AI tends to skew overwhelmingly toward generative AI.
However, generative models represent just a segment of the broader ecosystem of AI solutions powering compelling new capabilities. Here are four less-discussed AI advancements that hold the potential to fuel tremendous value across a wide range of business use cases.
1. Graph Neural Networks
Deep learning, a subset of artificial intelligence, uses artificial neural networks (ANNs) to process data. Specifically, graph neural networks (GNNs) are designed to understand data organized in graphs and shape advanced predictions.
GNNs employ a “message passing” framework, where graph nodes update their knowledge by combining information from neighboring nodes. This allows the neural network to identify useful patterns and make predictions.
GNNs provide advanced analysis across a variety of industries, including:
• Modeling molecular structures to augment drug discovery.
• Particle physics simulations for scientific research.
• Trader assistance for market forecasting.
However, to be most effective, GNNs must overcome issues in scalability because modeling complex relationships demands significant storage and computational power. They also require pre-training on labeled data, which can be expensive to obtain. Nevertheless, GNNs hold great promise to revolutionize deep-learning strategies for complex solutions and have the potential to catalyze a surge of real-world innovation.
2. Causal AI
While AI has the ability to perform certain impressive tasks, many leading experts believe that deep learning still has a significant journey ahead toward optimized productivity. Despite its power in correlation-related tasks, AI cannot interpret cause-and-effect, nor why associations and correlations exist.
Unlike traditional machine learning approaches that focus on correlation, causal AI seeks to uncover causal relationships, which means understanding how changes in one variable directly influence changes in another. Causal AI can both identify the true cause of outcomes and potentially change those outcomes before they occur. Feeding the model hypothetical scenarios yields insights into how the inputs should be changed to achieve desired results.
Innovations unlocked by causal AI span various sectors, including:
• E-commerce: Causal forecasting systems enable a world-modeling mechanism to optimize both supply chains and business decisions.
• Manufacturing: Businesses can deploy casual predictive maintenance to continuously monitor processes and dissect why manufactured parts are not meeting quality standards.
• Healthcare: Causal ML shows the potential to combine medical studies and complex data sets for expert clinical decision making that surpasses current ML solutions.
By identifying the underlying causes of a behavior or event, causal AI provides insights that statistical models are incapable of, thereby opening the door to novel solutions where traditional AI falls short.
3. Digital Twins
Creating an exact virtual replica of a physical object, place or being is no longer a science fiction fantasy. Data collected from a real-world entity can be relayed to a “digital twin,” which can then run simulations with the goal of generating valuable insights to improve the physical version.
The digital twin market is expected to grow from $8.6 billion in 2022 to nearly $138 billion by 2030 at a 42.6% CAGR. Digital twins are transcending previous capabilities across a wide variety of use cases, including:
• Power generating equipment: By simulating how power grids will react to extreme weather and ever-changing conditions, we can reduce outages, increase efficiency and optimize electricity allocation.
• Clinical research: About 80% of medical trials face delays in the enrollment phase, but digital twins of patients could create a wider and more diverse range of test subjects in a fraction of the time.
• Supply chain management: Digital twins can optimize fleet management, packing performance and distribution route efficiency to minimize supply chain costs.
Simulations of physical entities in a virtual world augment research, reduce costs and provide valuable insight applicable to the physical world.
4. Swarm Intelligence
Another promising subset of AI, swarm intelligence (SI), is a phenomenon inspired by the collective brilliance we observe in nature by animals and insects. Take honey bees, for example, whose collective grasp in high winds protects their hive and queen. Analogously, SI approaches problems in complex environments via a decentralized multi-agent system.
SI is particularly well-suited to the field of the Internet of Things (IoT). IoT-based systems are complex and consist of multiple objects, so powerful SI algorithms are optimal for analyzing and monitoring such intricate environments.
Some other common use cases include:
• Sustainability: SI solutions could augment the transition to environmentally sustainable practices such as more efficient waste management systems.
• Data science: Swarm intelligence is proving effective at managing and analyzing big data, as well as optimizing clustering and feature selection in data mining.
• Traffic control: The self-controlled, stochastic methods of SI outperform traditional traffic flow methods that aim to reduce congestion.
Simply put, SI provides valuable insights using a decentralized problem-solving approach, which has inspired many novel real-world applications.
While these AI advancements hold great promise, I would advise business leaders to take an incremental approach to harnessing AI’s potential within their business. Start by identifying a few high-impact areas where AI could drive real value—predictive maintenance to reduce equipment downtime, for example.
Run small-scale pilot projects focused on these pain points, measure the results and use hard ROI data to make the case for further AI adoption. Resist the urge to transform everything at once. Develop in-house expertise through strategic hires or upskilling, but focus their efforts on targeted solutions rather than on moonshots. Foster a culture of experimentation around AI, understanding that iterations and occasional failures are part of the process.
Looking beyond generative AI, it is crucial to embrace and harness the potential of these lesser-known advancements for continued growth and transformative change in our society.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?
Read the full article here
