Transformer-based deep learning models, such as GPT-3, have received a lot of attention in the machine learning world. These models excel at understanding semantic relationships and have contributed to major improvements in Microsoft Bing’s search experience. However, these models can fail to capture more subtle relationships between query terms and documents beyond pure semantics.
See also: Pentagon: Its AI technology can predict events
Microsoft researchers have developed a neural network with 135 billion parameters, which is the largest “universal” artificial intelligence they have in production. The large number of parameters makes this one of the most sophisticated artificial intelligence ever publicly reported to date. OpenAI’s GPT-3 natural language processing model has 175 billion parameters and remains the world’s largest neural network built to date.
Microsoft researchers call their latest AI project MEB (Make Every Feature Binary). The 135 billion-parameter machine is built to analyze queries entered by Bing users. It then helps identify the most relevant pages from the web with a set of other machine learning algorithms built into its functionality, and without performing any tasks on its own.
See also: Bitmain Co-Founder Turns Crypto Business into Unicorn
MEB is a great complement to Transformer-based deep learning models because it can map individual events and features, which allows MEB to gain more granular understanding. For example, many DNN (deep neural network) language models may generalize when filling in the blanks of this sentence: “(blank) can fly.” Since most training cases result in “ birds” being able to fly, some DNNs may fill in the word “bird” only for these blanks. However, MEB feature mapping helps by not just relying on one or two examples, but instead by paying extra attention to each possible outcome.
MEB ultimately enabled 100% coverage across all Bing searches. Unlike other models that can be a bit rigid and static with features, this one is able to learn from huge amounts of data continuously while remembering reliable facts represented by binary features.
See also: NVIDIA AI hub now available to North American customers
The Microsoft Bing team discovered that adding MEB to the search engine resulted in a 2% increase in click-through rates and a greater than 1% reduction in users retyping queries because they received no relevant results.
Information source: marktechpost.com
