We do not know how to send information back in time, but we know how to change the past without knowing what the past was that we're changing, and we do know we're changing it and what it changes to. Therefore the past can be changed, and we should expect time machines will exist in the future and therefore in the past and present if their operators choose to.
In the double-slit experiment, a laser sends 1 photon or electron or other particle/wave at 2 holes in a wall. The particle/wave goes through the left slit, the right slit, both, or bounces off the wall. Put a detector past the slit wall, which is set up to be close enough to detect or far enough not to detect as controlled by a choose-to-detect switch that is activated after the particle goes through the slits but before its time that it may or may not be detected. If choose-to-detect is true, then the sum of 2 bell curves will be observed on the farthest back wall. If choose-to-detect is false, then a wave interference pattern will be observed on the farthest back wall. The pattern is a statistical distribution over time that converges exponentially toward certainty, not for individual tests.
By changing choose-to-detect from true to false or from false to true, after the particle/wave has gone through the slit(s) but before where it would be detected, the past can be changed between: (1) went through only 1 slit, resulting in the sum of 2 bell curves on the farthest back wall, or (2) went through both slits, resulting in a wave interference pattern on the back wall.
This is a fact of quantum physics. Its been proven. The past has been changed many times in labs this way, except I don't know if they've added such a switch, but the switch is a trivial change and the behavior of the experiment depends only on if its observed or not. Time travel exists. Its a fact.
Also, information may be sendable to the past by changing the angle and position that detection is done (if choose-to-detect is true it detects) so that a small amount of statistical information can be known about the previous position of the particle/wave before it went through the slit(s) (and if it goes through one slit or both depends on what happens in the future, setting choose-to-detect or not), so by running many of these experiments simultaneously with the same choice of choose-to-detect, information maybe can be exponentially-reliably sent to the past, and by chaining any number of these experiments together, information could maybe be sent as far back in the past as you have machines, where each machine gets you a very small fraction of a second. Maybe a loop made of mirrors would allow re-using the same machines?