Technology Vision 2023 | When Atoms meet Bits #T#TechVechVisionision Face closed a $100 million funding round at a The changes foundation models have brought $2 billion valuation in May 2022 to help it become to the AI landscape—and their expected impact the premier location for building with pretrained on businesses—shouldn’t be understated. With is is a fast-moving AI models. platforms making it increasingly simple to adapt pre-trained models for various AI needs, companies ‹eld, so it’s time to A “middle layer” is also starting to appear alongside may deploy AI solutions far more readily than start imagining how foundation model platforms, offering services before. And with foundation models expanding these models may like 昀椀ne-tuning the models for more custom use. into new data modalities, companies can build truly transform—or disrupt Though foundation models are pretrained and highly novel applications that wouldn’t have been possible adaptable, some downstream tasks may require before. This is a fast-moving 昀椀eld, so it’s time to start —your business. more speci昀椀c training, and 昀椀ne-tuning means imagining how these models may transform—or training it on new, task-speci昀椀c data to change its disrupt—your business. output. Microsoft, for instance, is packaging and selling GPT-3 capabilities to customers, integrating it into its low-code app development platform Power Apps, and offering customers access to the model along with the enterprise-grade security, compliance, reliability, and data privacy assurance of Azure.215