2023-Management-Ramakishnan

Conference Video|Duration: 49:50
March 8, 2023
Please login to view this video.
  • Video details
    To unlock value from today’s powerful AI models, large volumes of training data are considered necessary. For the consumer internet companies where many of today’s AI models originated, this hasn’t been difficult to obtain. But for companies in other sectors - industrial companies, manufacturers, healthcare organizations, educational institutions – curating labeled data in sufficient volume can be a significant and sometimes prohibitively expensive effort, resulting in a formidable barrier to putting AI to work.

    There’s good news on this front, however. AI practitioners and researchers have developed strategies over the last few years with the potential to significantly reduce the volume of labeled data needed to build accurate AI models. These approaches encompass ways to learn models with just unlabeled data, to transfer-and-adapt models across problems, as well as best practices around “iterating on data” to improve model performance. By using these approaches, it is often possible to build a good AI model with a fraction of the labeled data that might otherwise be needed.

Locked Interactive transcript
Please login to view this video.
  • Video details
    To unlock value from today’s powerful AI models, large volumes of training data are considered necessary. For the consumer internet companies where many of today’s AI models originated, this hasn’t been difficult to obtain. But for companies in other sectors - industrial companies, manufacturers, healthcare organizations, educational institutions – curating labeled data in sufficient volume can be a significant and sometimes prohibitively expensive effort, resulting in a formidable barrier to putting AI to work.

    There’s good news on this front, however. AI practitioners and researchers have developed strategies over the last few years with the potential to significantly reduce the volume of labeled data needed to build accurate AI models. These approaches encompass ways to learn models with just unlabeled data, to transfer-and-adapt models across problems, as well as best practices around “iterating on data” to improve model performance. By using these approaches, it is often possible to build a good AI model with a fraction of the labeled data that might otherwise be needed.

Locked Interactive transcript