Machine learning and AI research, which is now a critical technology in almost every sector and corporation, is just too extensive for anybody to understand. This column seeks to compile some of the most important recent findings and articles — primarily in artificial intelligence, but not exclusively — and explain why they matter.
Scientists did an intriguing experiment this week in AI to forecast how “market-driven” systems like food delivery and ride-hailing services influence the whole economy when they are optimised for different goals, such as generating income. A team from ETH Zurich built a system that can interpret tree heights from satellite pictures, while another set of researchers tested a system that can forecast a startup’s success using public online data, illustrating the adaptability of AI.
Salesforce’s AI Economist, an open-source research environment for learning how AI may enhance economic policy, is the foundation for the market-driven platform work. Several of the AI Economist’s researchers were involved in the latest research, which was documented in a report released in March.
The purpose, according to the coauthors, was to look at two-sided marketplaces like Amazon, DoorDash, Uber, and TaskRabbit, which have more market strength due to rising demand and supply. The researchers trained a system to analyse the impact of interactions between platforms (e.g., Lyft) and customers using reinforcement learning, a sort of AI system that learns to solve a multi-level issue by trial and error (e.g., riders).
Using their AI system, the researchers discovered that during economic downturns, a platform built to maximize revenue tends to raise fees and make more money from buyers and sellers, putting societal welfare at risk. When platform fees are set (for example, owing to legislation), they discover that a platform’s revenue-maximizing incentive often matches the larger economy’s welfare concerns.
The conclusions are hardly groundbreaking, but the co authors believe the system — which they aim to open source — may help a business or policymaker examine the platform economy under various situations, designs, and regulatory concerns. “We use reinforcement learning as a framework to characterise strategic operations of platform organisations that optimise pricing and matching in reaction to changes in the environment, such as an economic shock or a regulatory change,” they continued.
Turning our attention away from platform businesses and toward the venture capital that fuels them, Skopas researchers claim to be able to predict a startup’s ability to attract investments using publicly available data. Skopas is a startup that uses AI to characterise companies based on criteria like technology, market, and finances. The co authors claim that using data from startup websites, social media, and business registers, they can get prediction results “similar to those using structured data accessible in private databases.”
It is not a novel concept to use artificial intelligence in due diligence. Algorithms are being used by companies like Correlation Ventures, EQT Ventures, and Signalfire to help them make better investments. According to Gartner, by 2025, 75% of VCs will utilize AI to make investment choices, up from fewer than 5% currently. While some perceive the technology’s benefits, hazards are lurking beneath the surface. Harvard Business Review (HBR) published a study in 2020 that indicated that investing algorithms beat novice investors but had biases, such as favouring white and male entrepreneurs. This, according to HBR, mirrors reality, demonstrating AI’s propensity to accentuate existing prejudices.
In more upbeat news, researchers from MIT, Cornell, and Microsoft claim to have created a computer vision system called STEGO that can detect pictures down to the pixel level. While this may not seem like much, it is a huge step forward from the traditional way of “training” an algorithm to recognise and categorise objects in images and videos.