First off, you need to look at the differential amplifier configuration of an op-amp, not just the op-amp itself.
NOT the article you had before. Yes, the op-amp is a differential amplifier, but its internal gain can and often times is more than a million. You must control it with negative feedback. I won't blurt out the equations for the control loop theory behind it, but your current configuration is causing the op-amp to function with all its gain. Therefore, a 1microvolt (1 millionth of a volt) difference between the + and - terminals could easily translate to 1 volt or more. The thermal noise ALONE from a 10K resistor could generate upwards of 6uV (pk-pk) of noise (@10KHz bw). Plus there is noise from the op-amp itself. Now, you've got 6uV times a million, that's 6V. You get a randomly swinging signal of 6V that you can't even control because it's noise.
Long story short, you have to have feedback resistors.
What are you trying to accomplish with that circuit? Are you trying to do strain gauge measurements?