Massachusetts Institute of Technology’s (MIT’s) Alexey Radul is developing a conceptual framework for computing that could impact artificial intelligence research, parallel computing, and the design of computer hardware. In Radul’s system, multiple logic circuits and memory cells are arranged in a large network. Any logic circuit can exchange data with different memory cells, and any memory cell can exchange data with different logic circuits. However, contradictory data could confuse memory cells, leading to the overwriting of both sets of data. Radul solved this problem by developing memory cells that gradually accumulate information about data instead of just storing it. A programmer using Radul’s system can decide what kinds of information about data the memory cells will store. The cells can track where data comes from, a capability that could be useful in many applications, according to Radul. The system could determine the source of an error in the data, backtrack to that source, and correct all the resulting errors. A more developed version of Radul’s system would enable programmers to specify computational problems in a way that automatically takes advantage of parallelism.
View Full Article
