My observation, and its tainted by the fact that I lowered the AI trade value thing to 100% for Chieftan through Regent, and adjusted the tech rate for map sizes down alot)
On my two warlord games so far, it felt about right. I could get a tech lead, but I had to work at it. In one game, several ai's got behind and stayed that way like one would expect.
First game on Regent, it seems more like 1.17. The ai's trade techs almost as soon as they become available (though I wasn't left out, probally because of the 100% trade rate adjustment). However, the Americans did get left out of the loop, and fell about 5 or 6 techs behind for most of the game (they was the smallest empire and had no tradable resources)
All in all, I think its a nice balance no between realism and fun. They still trade a bit too easily at times (especially older techs), but its now at least possible to get a tech lead if you make it your primary goal (albeit a short lived one, they do catch up if given a chance) And its equally possible for an AI civ to fall behind. Unlike 1.21, if you find an isolated civ still working his way out of the stone ages when your industrial, that civ won't be in a tech lead 2 turns later. Unlike 1.16 and before, that civ won't stay stone aged for long though (in my regent game, the isolated zulu went from where the great library had left him to the end of the middle ages (military tradition) before he had nothing to trade to catch up totally (the other two leading civs: me & england) had refining & electricity) Presumably by the end of the game though Shaka will probally catch up since he has more cities than any other civ by far (most of which with german names, bad shaka bad shaka, well maybe good shaka if you go after england next

)
Most importantly though, unlike 1.21 the AI civs dont all always know the exact same list of techs as each other

I, at least, no longer really feel like its me against them in the research department.