Facebook’s AI mathematician can solve university accounting problems
- December 30, 2019
- William Lewis
Machines are improving mathematics – artificial intelligence has learned to solve university-level calculus problems in seconds.
François Charlton and Gillum Sample at Facebook AI Research unravelled the problems of tens of millions of calculus through a computer. Problems were mathematical impressions that included integration, a common calculus technique for finding the area under a curve.
To find solutions, AI used natural language processing (NLP), a computational tool commonly used for language analysis. This works because the math in each problem can be thought of as a phrase, with variables, usually defined as x, playing the role of nouns and processes, such as finding the root of the square, of the function. To perform a task The AI then turns the problem into a solution.
Read more: Google has developed a math AI that has already proven 1200 theorems
When the pair tested AI on 500 computational problems, it found a solution with 98% accuracy. A comparable standard program for solving math problems had only 85% accuracy on the same issues.
The team also gave various AI distinctions to solve, which are other equations that require integration as well as other techniques to solve them. For these equations, AI was not good enough, which answers them correctly for a type of discrimination, at 81 per cent and 40 per cent strictly.
Even so, it can still correctly answer questions that confuse other math programs.
Read more: DeepMind developed a math AI that can increase up to 6 but goes wrong 7
Charting says doing computer calculations is not particularly practical, but more training may one day be able to tackle math problems for humans. Suffer from severe issues.
Nicos Eletres of the University of Sheffield, UK, says AI performance can save people time in other math tasks, for example, when proving theories.