A.I beats Another Human in solving a Puzzle


In addition to assessing the condition of singing coral reefs, artificial intelligence has been shown to outperform humans at solving crossword puzzles.

ads
 

Crossword puzzles are a favorite pastime for many people all around the world. It has long been believed that humans are superior to computers in solving crossword puzzles because of our exceptional ability. Participants put their vocabulary and general knowledge to the test as they try to decipher the cryptic clues and fill up the crossword grid with the correct answers.

The annual American Crossword Puzzle Tournament is regarded as one of the most elite crossword competitions. In 2017, a piece of software known as Dr. Fill allowed a machine to finish in 11th place, the most remarkable performance it has ever achieved in this competition. However, computers are currently making a comeback. In the year 2021, the Berkeley Crossword Solver, developed by a group of people at the University of California, Berkeley, successfully outscored every human player for the very first time.

This was no mean fit; it signaled that computers could be great at any specialized task if enough resources and time were dedicated to improving them.

The program, described in a post on arXiv, relies on “neural question answering, structured decoding, and local search,” which entails locating solutions to problems and refining them with databases and natural language artificial intelligence (AI). And even with that kind of development, there is always room for enhancement. Compared to earlier versions (57%), the Berkeley Crossword Solver now has an “exact puzzle accuracy” of 82% for New York Times crosswords. Crossworders beware.

Singing coral reefs

This week, we learned about another innovative use for artificial intelligence (AI), which involves evaluating the state of coral reefs to determine how healthy they are. A multinational team led by scientists from the University of Exeter and Lancaster University in the United Kingdom employed artificial intelligence to study the sounds reefs make. These “songs” are produced by the diverse population of fish and other organisms that call the reef home; the absence or absence of these organisms is directly correlated to the state of the ecosystem.

Team members used Indonesia’s Mars Coral Reef Restoration Project recordings to hone their artificial intelligence. Once taught, the algorithm accurately identified the reef’s health 92% of the time based on recordings taken from both damaged and healthy reefs.

According to Ben Williams of Exeter, “Coral reefs are facing many dangers, including climate change; thus, monitoring their health and the performance of conservation efforts is crucial.” “Coral reefs are facing various threats including climate change,” Moreover, he says it takes time and effort to evaluate a reef’s condition using traditional visual and auditory methods.

According to our research, a computer can recognize patterns inaudible to a human’s naked ear. Williams explains that this method can provide more timely and precise information about the reef’s condition.

Tim Lamont from Lancaster chimes in, “In many circumstances, it is quicker and cheaper to place an underwater hydrophone on a coral and leave it there than to have experienced divers visiting the reef periodically to survey it – especially in distant locations.”

 
ads