Microsoft deploys neural network with 135 Billion parameters to improve Bing results
1 min. read
Published on
Read our disclosure page to find out how can you help MSPoweruser sustain the editorial team Read more
Microsoft Research has deployed a neural network nearly as big as the infamous GPT-3 to improve Bing results.
GPT-3 has 175 billion parameters and MEB (Make Every Feature Binary) has 135 billion parameters and is designed to analyse Bing search queries and connect them with the most relevant results on the web.
MEB improves results by preventing overgeneralization and offers more nuanced results by considering every possible outcome. It enables 100% coverage of all Bing searches and is able to learn from vast amounts of data continuously while reliably remembering facts.
Practically speaking MEB increases Bing clickthrough rates by 2% and yields a 1% reduction in users rewriting their queries because they did not receive any relevant results. 1.5% fewer users needing to click on the “next page” button means they didn’t find what they were looking for on the first page.
Read Microsoft Researches write-up for all the details here.
via MarkTechPost
User forum
0 messages