[steve2005]

auction
2♣-2N-4♣ Gerber 4♦ (0 or 4 Aces) - 4♠
2NT is a slam killer but here gives me a chance to ceck for 1 ace and i will bid slam
Ben didnt show the ace though
My link

[2_1 luv GF]

steve2005, on 2026-April-11, 20:20, said:

auction
2♣-2N-4♣ Gerber 4♦ (0 or 4 Aces) - 4♠
2NT is a slam killer but here gives me a chance to ceck for 1 ace and i will bid slam
Ben didnt show the ace though
My link

Hello,

I don't know if I'm right but I think I found a discrepancy which explains why Ben actually made the "right" bid:

You must consider that, unlike GIB(BO), Ben learned Bridge from scratch solely by being trained on more than 100 million ACBL hands played in human pair games on BBO.

To quote the article "Parallel Universes" in the 05/2025 issue of the ACBL Bridge Bulletin (page 31):

"Remarkably, Ben has managed to learn bridge without understanding fundamental concepts such as what a trump is, what constitutes a trick, the order of bidding, the hierarchy of suits, the sequence of discards, or even the requirements for making a contract"
...
"The process began by loading 100 hands and their bidding sequences into the system. After this initial data, when we asked the AI to bid on a new hand, it produced nonsensical results, such as opening with 7NT or following it with 1d. When we added another 100 hands and bids, the AI learned to pass but still didn’t grasp that bidding ends after three passes – for example, it would pass up to seven times in a row."

lol

On the other hand, it "borrowed" its bidding explanations from GIB.

s. About Ben on BBO:

"For bidding explanation it uses GIB."

So, wouldn't it make sense if the experts in the ACBL games use the modern version of Gerber with 1430 responses (where "Quantitative replaces the Queen" i.e. e.g. 4♠ would mean 2 Aces but not enough HCP for a quantitative invite to 6NT).

But, since it uses GIBBO's bid explanations, it comes up with "Gerber 0 or 4" after it responded 4♦ to show 1 Ace.