Three games to mark DeepZenGo's retirement against top human players:

24th Marh, vs Mi Yuting
1st April, vs Park Junghwan
7th April, vs Iyama Yuta

It is unclear if the project simple ends or if DeepZenGo is going to be replaced by a stronger AI program.

Mi Yuting managed to win the first game. It's typical of DeepZenGo's performance: very good opening; building up a huge lead (Mi's chance was given at only ~20% by other AI programs); typical pruning moves that human-being will never play; then poor end-game performance. It ended up losing a ko fight with insufficient ko threats (think of those pruning moves).

Was this game played online or a real board? Do you know the time limits?

From the photo, we can see Mr Kato, the author of DeepZenGo played this game in person against Mi Yuting. Also from the photo, I guess the time limit is 3 hours each side.

Zen beat Park who kindly resigned. It might be somewhat cheap, but playing out a loss against a bot known to mess up winning won games world be reasonable to me. But hard to mess up such a big lead, about 15 points?

Today Zen beat Cho Chikun (not Iyama Yuta), in under 100 moves. I put a few comments in the sgf.

Tsk, tsk. AIs today. What's it been? Even two years? In my day, when a machine replaced a human, it was productive much longer. Why I was just chatting with the fuzzy logic circuit in the elevator the other day, who has been going strong for decades and his lot thinks it is a sorry state of affairs.