AI Content Chat (Beta) logo

Technology Vision 2023 | When Atoms meet Bits #TechVision Constructing a new AI strategy Every year AI plays a bigger role in companies’ be the better choice. Beyond that, some applications innovation agendas, operations, customer require special characteristics that other approaches interactions, and more. But now, organizations’ are better suited to. On-device AI may require carefully constructed strategies need to be extreme low power operation or the ability to run rethought with foundation models in mind. an AI model fully of昀氀ine so data can stay with the This means discovering the most appropriate device rather than being sent to the cloud. 98% applications for foundation models in your organization, planning to access the broader What’s more, foundation models have some of global executives foundation model ecosystem, and building the characteristics that—for now—make them agree AI foundation talent pipeline that will be needed to create, questionable for certain situations. This is adding maintain, or integrate these models. an entirely new dimension to enterprises’ security models will play an efforts. For instance, bias in foundation models is a important role in First to identify the right role for foundation models common concern due to homogenization as well as their organizations’ in your organization, it’s necessary to understand the fact that many are trained on large datasets from strategies in the their best use cases. There are many AI applications the internet. Historic datasets that excluded certain that work with data types no foundation model populations, people, and demographics can lead to next 3 to 5 years. can handle yet. Additionally, some use cases that undesirable outcomes. And some companies’ early a foundation model could feasibly attempt are still efforts even found misinformation was affecting the fundamentally better served by narrow AI, which is outputs of these algorithms. When the same few trained speci昀椀cally for a task rather than across a models are used as the basis for many downstream 216 modality. Clearly de昀椀ned problems like predictive applications, then any issues in the original models maintenance and medical image analysis, for may propagate throughout the rest. There have been instance, are areas where narrow intelligence may efforts from the makers of foundation models to

When Atoms meet Bits - Page 76 When Atoms meet Bits Page 75 Page 77