BENGALURU: Open-source models will remain competitive with proprietary models, offering more choices to customers and putting pressure on industry prices. In December 2023, French startup Mistral launched an open model that equalled or beat GPT3.5, the top GenAI model in early 2023, on most benchmarks, Moody’s said in its latest report on 2024 AI Outlook.
Open-source developers also offer solutions that simplify the creation of AI-powered applications. The report says advances in artificial intelligence are still at an early stage and will not significantly affect credit quality over the next 12 months.
AI innovation will develop rapidly in 2024, drawing more users and fostering a vibrant network of hardware suppliers and app developers. Risks will increase, too, as organisations start rolling out AI applications at scale for the first time.
The report says the transformation of the banking, insurance and manufacturing sectors will accelerate. Over the next two years, Moody’s expects a material impact only on certain software and semiconductor firms. From 2026, the effects will be far broader. Sectors with large datasets and sizable staff costs will benefit the most.
On regulation, the Moody’s report says regulators and policymakers will aim to balance risk mitigation with innovation. Legislation in the US will progress but extensive national regulation is unlikely. The European Union will pass the AI Act during the first half of 2024. China’s AI rules are the most comprehensive, although government oversight may hold back its AI industry.
The report says that rapid innovation in AI technologies will continue in 2024, drawing more users and fostering a vibrant network of hardware suppliers, model providers, and application developers. “We do not foresee AI significantly affecting credit quality in the next 12 months, but as use cases gradually broaden across industries, many issuers will start feeling the effects over the next two to three years,” it adds.
Edge computing, a means of processing data near its source, such as a personal computer or a smartphone, rather than sending it to a distant cloud server, will also facilitate the development of AI applications. It enables the creation of more responsive applications that can work offline and comply with data protection regulations while lowering computing costs for application providers.