Alibaba’s artificial intelligence model managed to outdo a human in a reading comprehension competition organized by the Stanford University in the United States, marking a major milestone for the technology which so far wasn’t able to match humans in the Stanford Question Answering Dataset. The competition saw human testers score 82.304 on the ExactMatch (EM) scale, with Alibaba’s SLQA+ model achieving an 82.44 rating. Microsoft’s r-net+ solution shared the first place with the Chinese company, posting a slightly higher score of 82.65, but the historic achievement still belongs to Alibaba’s tech due to the fact that its scores were finalized faster. Microsoft’s AI also recorded an incremental improvement over the 82.136 score posted by the model last month.
While Microsoft and Alibaba now outdid human performance, most other competitors still have some way to go, with Samsung being one of them as its MAMCN+ model placed 14th with a score of 77.436. The South Korean tech giant still managed to beat out 83 other participants but its placement doesn’t put it anywhere near the top performers in the field. Samsung has been committing significant resources to AI advancements in recent years and is now also reportedly working on new a new chip meant to mimic the manner in which a human brain operates in order to provide a better, more efficient basis for its future AI solutions, yet such creations are still multiple years away and aren’t even guaranteed to outdo existing leaders in the field, let alone whatever the industry comes up with next.
The current state of even the very peak of the technology still isn’t noticeably better than a human, with Alibaba and Microsoft’s models barely beating a human score, yet the development is still indicative of the rapid progression the AI industry has been delivering for years now. Alibaba’s AI technology is first expected to benefit its e-commerce solutions but should eventually power a wide variety of products and services, with one of the company’s focus areas being self-driving cars. Microsoft has similarly broad ambitions in the segment but remains vague about its plans that are presumably yet to take a more defined form. In the long term, AI technologies should allow for everything from improved healthcare and analytics to better home automation, driverless vehicles, and on-demand cities.