Combining　Artificial　Intelligence　with　control　theory　will　get　high　performance　regulators.　This　paper　discusses　a　kind　of　intelligent　adaptive　regulators　in　which　both　the　PI　compensation　in　control　theory　and　the　knowledge　base　are　used.　When　the　parameters　of　the　controlled　system　have　changed,　this　regulator　can　automatically　update　its　own　PI　parameters　so　that　the　stability　and　performance　of　the　system　can　be　maintained　in　spite　of　process　variations.　The　key　problem　is　how　to　apply　not　perfect　experience　of　human　experts.　This　paper　describes　the　principles　about　this　type　of　regulators　and　its　reliability　problems.