AI has failed on many promises. With the ensuing scepticism of artificial intelligence, we are faced with a choice: choose to become too cynical and sit on the sidelines and wait for a winner, or find a way to filter the noise and identify business breakthroughs early enough to participate in this historic economy among the opportunities. There is a simple framework that distinguishes near-term reality from sci-fi. We measure it using the single metric that matters most to any technology – maturity: its ability to manage unforeseen events often referred to as edge cases. As the technology continues to improve, it becomes better at handling increasingly rare edge cases, and as a result gradually unlocks new applications. This is an important point: if today’s AI focuses on accuracy or retrieval, very high performance can be achieved. In other words, it optimizes one at the expense of the other (i.e. trades fewer false positives for more false negatives, and vice versa). But to achieve high performance on both fronts at the same time, AI models will encounter difficulties. Solving that problem is the holy grail of artificial intelligence.
Autonomous Delivery Mobile Robots (AMRs) are the first application of autonomous commercialization in cities, and robo-taxis are still waiting for unmatched high-fidelity AI performance. The pace of industry progress and the experience of the past five years reinforce our view that the best way to commercialize AI is to focus on supporting narrow applications with low-fidelity AI, with human intervention for high-fidelity performance when needed. In this model, low-fidelity AI leads to early commercialization, followed by incremental improvements that help drive business KPIs. By targeting more forgiving use cases, businesses can achieve early commercial success using low-fidelity AI, while maintaining a realistic view of high-fidelity features that will take years to implement. After all, sci-fi has no place in business planning.
This article is reprinted from: https://www.solidot.org/story?sid=72086
This site is for inclusion only, and the copyright belongs to the original author.