Research conducted at Massachusetts Technological Laboratory (MTL) has demonstrated significant gains in execution speed in phased array algorithms for null-steering and adaptive nulling by implementation in the fine-grained ("data level") parallel computational architecture of the Connection machine. Such work suggests that significant gains in phased array performance may be obtained by the development of a "smart" phased array, where computational elements are dispersed among the phased array elements and an appropriate network of interconnections for transfer of information between processors is provided. Such a "smart" phased array has the potential for substantially reduced high data-rates, large bandwidth communication with a central computer, and rapid (microsecond time scale) nulling. Additional capabilities for the real-time monitoring of array performance and limited automatic failure compensation ("self-healing") also appear promising. The Connection Machine is highly suitable for the development of algorithmic tools and simulations for the study of "smart" phased arrays. A program of research to demonstrate the feasibility of "smart" phased arrays by simulations on the Connection Machine and for the study of their properties is outlined.