Insights and updates from our agency

Image GPT provided a research lens for understanding how sequence modelling could extend beyond text. This is a continuing-trend review, not a claim of an August launch.

The CASP14 assessment period made protein structure prediction a useful example of AI working on scientific problems. This article discusses the assessment period, not the November result announcement.

OpenAI announced an API, creating a route for developers to build products around its language models.

The GPT-3 paper demonstrated that a large language model could perform a range of tasks from instructions and examples.

Jukebox explored generating music with singing, showing both the creative possibilities and the limitations of early audio generation.

CORD-19 brought together coronavirus research literature for machine-assisted discovery during the pandemic.

Microsoft introduced Turing-NLG, illustrating how larger language models could support summarisation and text generation.

Conversational systems were being assessed on whether replies sounded sensible and specific, a useful distinction for customer service teams.