In general I am happy to have them around and to have them improve. There are of course some dangers from such machines making harmful decisions, but probably no more dangers than with humans making such decisions.
Having such machines will not answer the questions about the world that are most important to me and many others. What constitutes the dark matter of the universe? Is supersymmetry really a symmetry of nature that provides a foundation for and extends the highly successful Standard Model of particle physics we have? These and similar questions can only be answered by experimental data.
No amount of thought will provide such answers. More precisely, perhaps given all the information we have about nature some machine actually would come up with the right answers. Indeed, perhaps some physicists already have come up with the answers. But the true role of data is to confirm which answers are the correct ones. If some physicist, or some machine, figures it out they have no way to convince anyone else they have the actual answer. Laboratory dark matter detectors, or the CERN Large Hadron Collider, or possibly a future Chinese collider, might get the needed data, but not a thinking machine.