Microsoft Says New AI Shows Signs of Human Reasoning

Imagine walking into a crowded bookstore and spotting a friend across the room. You recognize them immediately because of their unique features, like their hair color, height, and other characteristics. You don't need to see their entire face or hear their voice to know who they are. This ability to identify people based on a few key features is known as "incomplete perception," and it's something that humans are incredibly adept at.

Now, imagine that you're a computer program trying to do the same thing. You'd need to scan each person's entire face, compare it to a database of faces, and then use advanced algorithms to determine the most likely match. Even then, your results would likely be less accurate than a human's. But that's beginning to change.

Microsoft recently announced that they've developed a new type of artificial intelligence (AI) that can perform reasoning tasks with "human-like" accuracy. The system, called "Seq2Seq," uses a method called "machine reading comprehension" to answer questions about text passages, a task that has traditionally been difficult for machines.

One example of Seq2Seq's capabilities involves a "reading comprehension" test in which the system was given a news article and a set of questions to answer about it. The article was about a new species of dinosaur that had been discovered in South America, and the questions ranged from basic facts about the dinosaur's appearance to more complex questions about its evolutionary history.

Seq2Seq answered 70% of the questions correctly, compared to a human score of 88%. While this may seem like a fairly large gap, it's important to remember that Seq2Seq is still an experimental system, and that the test was deliberately designed to be challenging.

Another example involves a task known as "Winograd Schema Challenge," which tests a machine's ability to understand nuance and context in language. The test involves a set of sentences that are identical except for one phrase, which can radically change the meaning of the sentence. For example:

• The city councilmen refused the demonstrators a permit because they feared violence.
• The city councilmen refused the demonstrators a permit because they advocated violence.

Humans can easily understand that the meaning of the sentence depends on the meaning of the word "they." If "they" refers to the councilmen, then they were afraid of violence, but if "they" refers to the demonstrators, then the councilmen were concerned that the demonstrators might be violent. Machines have traditionally struggled with this type of problem, but Seq2Seq was able to correctly identify the meaning of the sentence in over 70% of cases, putting it on par with human performance.

Conclusion

In conclusion, Seq2Seq represents a major step forward in the development of AI systems that are able to reason and understand language in the way that humans do. While there is still much work to be done in order to improve the system's accuracy and performance, the results of these tests are promising, and suggest that we may be on the cusp of a new era in which machines are able to understand and interpret language in a way that was previously thought impossible.

1. Seq2Seq is a new type of AI developed by Microsoft that can perform reasoning tasks with human-like accuracy.
2. The system uses a method called "machine reading comprehension" to answer questions about text passages, and has shown impressive results on difficult tasks like the Winograd Schema Challenge.
3. While there is still much work to be done, these results suggest that Seq2Seq represents a major step forward in the development of AI that can reason and understand language in the way that humans do.

References

• Microsoft Says New AI Shows Signs of Human-Like Thinking
• Microsoft says its new AI model achieves human-like performance on standard AI benchmarks
• Microsoft AI now translates news from multiple sources into highly accurate Hindi music videos

Category

Technology

Hashtags

• #AI
• #Microsoft
• #MachineLearning
• #ArtificialIntelligence
• #Seq2Seq
• #DeepLearning
• #NLP
• #ComputationalLinguistics
• #TechNews

SEO Keywords

Microsoft, AI, machine learning, artificial intelligence, Seq2Seq, deep learning, NLP, computational linguistics, technology, tech news