SenseTime unveils SenseNova and the SenseChat chatbot

Background

Chatbots built on large language models were a major focus for AI companies in 2023. SenseTime is a Chinese AI company, and its 2023 event introduced its large models for conversation and content generation.

What happened

On 10 April 2023 SenseTime unveiled its SenseNova foundation model set at a Tech Day event ^[1]. The set included the ChatGPT-style chatbot SenseChat, a text-to-image generator and an avatar video generator ^[1].

Key details

SenseTime said SenseChat has more than 100 billion parameters ^[1]. CEO Xu Li gave a live demonstration of the chatbot, including code writing and medical advice ^[1].

Access to the products was initially limited to government and corporate clients ^[1].

The company's shares rose as much as 11.1% to HK$3.70 on the next trading day, before closing at HK$3.30 ^[2].

Impact

The event marked SenseTime's public introduction of SenseNova and SenseChat. It set the starting point for the SenseNova series, which later releases continued, and it recorded what the series contained and who could use it at launch.

Terms used here

A foundation model is a large model trained on a broad body of data and then used as a base for many applications. A chatbot built on such a model answers questions in conversation. A text-to-image generator produces a picture from a written description, and an avatar video generator produces a video of a digital person from input material.

A parameter is one of the adjustable numbers inside a neural network. A benchmark is a standard test set used to compare models, and its results depend on the test chosen. A chatbot demonstration shows how a model handles the tasks that were picked for the show, but it does not show how often the answers are correct.

Public access and enterprise access are different in practice. A product open to public testing reaches many more people than one limited to a small set of clients, so the access rules are part of what a launch describes.