井字棋

我们先实现一个最基本的使用控制台交互的井字棋游戏。
在这里插入图片描述

为了保持代码整洁，方便后续扩展，我们使用类Board来实现棋盘。除了常规的初始化方法__init__和字符串方法__str__，我们还要判断游戏的胜负、棋子位置的合理性。

在main中，我们在while循环中实现两个玩家的交替下棋，直到一方胜利或棋盘满了为止。

代码整体比较简单，只是在判断胜利时需要处理好多个条件。

class Board:def __init__(self, size=3):self.board = [['_' for j in range(size)] for i in range(size)]self.size = sizedef __str__(self):res = ""for row in self.board:temp_row = ""for col in row:temp_row = temp_row + col + '\t'res += str(temp_row) + "\n"return resdef display(self):print(self)def can_set(self, row, col):return 0 <= row < self.size and 0 <= col < self.size and self.board[row][col] == '_'def set_board(self, row, col, mark):if self.can_set(row, col):self.board[row][col] = markelse:print("Invalid input...,")def check_win(self):for row in self.board:if all(x == 'X' for x in row):return "Player 1 wins"elif all(x == 'O' for x in row):return "Player 2 wins"for col in range(self.size):if all(self.board[row][col] == 'X' for row in range(self.size)):return "Player 1 wins"elif all(self.board[row][col] == 'O' for row in range(self.size)):return "Player 2 wins"# Diagonalif all(self.board[i][i] == 'X' for i in range(self.size)) or all(self.board[i][self.size-1-i] == 'X' for i in range(self.size)):return "Player 1 wins"elif all(self.board[i][i] == 'O' for i in range(self.size)) or all(self.board[i][self.size-1-i] == 'O' for i in range(self.size)):return "Player 2 wins"return "No one wins"def is_full(self):return not any(self.board[i][j] == '_' for i in range(self.size) for j in range(self.size))if __name__ == '__main__':size = int(input("Please input the size of the board: "))board = Board(size)board.display()# player 1：X  player 2：  Omarks = ['X', 'O']player = 0while not board.is_full() and board.check_win() == "No one wins":index = input(f"Please player {player+1} input coordinates (row, col): ")try:row, col = index.split(',')row, col = int(row)-1, int(col)-1except ValueError:print("Invalid input. Please input two integers separated by a comma.")continueif board.can_set(row, col):board.set_board(row, col, marks[player])else:print("This cell is already occupied. Please choose another one.")continueplayer = (player + 1) % 2board.display()print(board.check_win())

使用tkinter创建图形界面

在这个程序中，使用了tkinter模块来创建窗口和按钮。每个按钮都由Button类创建，并通过grid()方法来定位。当按钮被点击时，clicked()方法被调用，将其设置为当前玩家的标记，并切换到下一个玩家。每次下棋后，程序会检查游戏是否已结束。如果游戏结束，程序会调用show_winner()方法来显示获胜者或平局信息，并禁用所有按钮。

import tkinter as tkclass TicTacToeGUI:def __init__(self):self.root = tk.Tk()self.root.title("Tic Tac Toe")# 创建9个按钮self.buttons = []for i in range(3):row = []for j in range(3):button = tk.Button(self.root, text="", font=("Helvetica", 24), width=5, height=2, command=lambda row=i, col=j: self.clicked(row, col))button.grid(row=i, column=j)row.append(button)self.buttons.append(row)self.current_player = "X"self.board = [["-", "-", "-"],["-", "-", "-"],["-", "-", "-"]]def clicked(self, row, col):"""当一个按钮被点击时，将其设置为当前玩家的标记，并切换到下一个玩家。"""if self.board[row][col] == "-":self.buttons[row][col].config(text=self.current_player)self.board[row][col] = self.current_playerif self.current_player == "X":self.current_player = "O"else:self.current_player = "X"self.check_game_over()def check_game_over(self):"""检查游戏是否已结束"""for i in range(3):# 检查行if self.board[i][0] != "-" and self.board[i][0] == self.board[i][1] and self.board[i][1] == self.board[i][2]:self.show_winner(self.board[i][0])return# 检查列if self.board[0][i] != "-" and self.board[0][i] == self.board[1][i] and self.board[1][i] == self.board[2][i]:self.show_winner(self.board[0][i])return# 检查对角线if self.board[0][0] != "-" and self.board[0][0] == self.board[1][1] and self.board[1][1] == self.board[2][2]:self.show_winner(self.board[0][0])returnif self.board[0][2] != "-" and self.board[0][2] == self.board[1][1] and self.board[1][1] == self.board[2][0]:self.show_winner(self.board[0][2])return# 检查是否有空格for row in self.board:for cell in row:if cell == "-":return# 如果没有空格，平局self.show_winner("Tie")def show_winner(self, winner):"""显示获胜者或平局信息，并禁用所有按钮。"""if winner == "Tie":message = "It's a tie!"else:message = f"Player {winner} wins!"for row in self.buttons:for button in row:button.config(state="disabled")self.root.title(message)def start(self):"""开始游戏。"""self.root.mainloop()# 运行游戏
game = TicTacToeGUI()
game.start()

人机对战版

这是一个井字游戏的 Python 代码，分别有三个类：Board，AIPlayer，Game。

Board类
这个类代表了游戏的棋盘，有以下方法：

__init__(self, size=3)：构造函数，初始化棋盘为给定大小的二维列表，初始值为 ‘_’。
__str__(self)：返回当前棋盘的字符串表示形式。
display(self)：打印当前棋盘的字符串表示形式。
can_set(self, row, col)：判断给定位置能否放置棋子。
set_board(self, row, col, mark)：将给定位置放置给定的标记（‘X’ 或 ‘O’）。
check_win(self)：判断当前棋局是否已分出胜负，返回 “Player 1 wins”、“Player 2 wins” 或 “No one wins”。
is_full(self)：判断当前棋盘是否已满。
copy(self)：返回当前棋盘的副本。

AIPlayer 类
这个类代表了游戏中的 AI 玩家，有以下方法：

__init__(self, mark)：构造函数，初始化 AI 玩家的标记（‘X’ 或 ‘O’）。
get_best_move(self, board)：计算当前 AI 玩家应该下的最佳位置。
minimax(self, board, is_maximizing, alpha, beta)：计算给定棋盘状态下当前玩家的最大（或最小）得分。

Game 类
这个类代表了游戏本身，有以下方法：

__init__(self, size=3, ai_mode=False)：构造函数，初始化游戏的棋盘大小和是否启用 AI 模式。
start(self)：开始游戏。
在 start 方法中，游戏会循环进行，直到棋盘已满或有一方获胜。每次循环，根据当前玩家是否是 AI，分别提示玩家输入位置或计算 AI 玩家应该下的位置，然后更新棋盘，交换当前玩家。

class Board:def __init__(self, size=3):self.board = [['_' for j in range(size)] for i in range(size)]self.size = sizedef __str__(self):res = ""for row in self.board:temp_row = ""for col in row:temp_row = temp_row + col + '\t'res += str(temp_row) + "\n"return resdef display(self):print(self)def can_set(self, row, col):return 0 <= row < self.size and 0 <= col < self.size and self.board[row][col] == '_'def set_board(self, row, col, mark):if self.can_set(row, col):self.board[row][col] = markelse:print("Invalid input...,")def check_win(self):for row in self.board:if all(x == 'X' for x in row):return "Player 1 wins"elif all(x == 'O' for x in row):return "Player 2 wins"for col in range(self.size):if all(self.board[row][col] == 'X' for row in range(self.size)):return "Player 1 wins"elif all(self.board[row][col] == 'O' for row in range(self.size)):return "Player 2 wins"# Diagonalif all(self.board[i][i] == 'X' for i in range(self.size)) or all(self.board[i][self.size - 1 - i] == 'X' for i in range(self.size)):return "Player 1 wins"elif all(self.board[i][i] == 'O' for i in range(self.size)) or all(self.board[i][self.size - 1 - i] == 'O' for i in range(self.size)):return "Player 2 wins"return "No one wins"def is_full(self):return not any(self.board[i][j] == '_' for i in range(self.size) for j in range(self.size))def copy(self):new_board = Board(self.size)for i in range(self.size):for j in range(self.size):new_board.board[i][j] = self.board[i][j]return new_boardclass AIPlayer:def __init__(self, mark):self.mark = markdef get_best_move(self, board):best_score = -1000best_move = Nonefor row in range(board.size):for col in range(board.size):if board.can_set(row, col):board_copy = board.copy()board_copy.set_board(row, col, self.mark)score = self.minimax(board_copy, False, -1000, 1000)if score > best_score:best_score = scorebest_move = (row, col)return best_movedef minimax(self, board, is_maximizing, alpha, beta):result = board.check_win()if result == "Player 1 wins":return -1elif result == "Player 2 wins":return 1elif board.is_full():return 0if is_maximizing:best_score = -1000for row in range(board.size):for col in range(board.size):if board.can_set(row, col):board_copy = board.copy()board_copy.set_board(row, col, 'O')score = self.minimax(board_copy, False, alpha, beta)best_score = max(best_score, score)alpha = max(alpha, best_score)if beta <= alpha:breakreturn best_scoreelse:best_score = 1000for row in range(board.size):for col in range(board.size):if board.can_set(row, col):board_copy = board.copy()board_copy.set_board(row, col, 'X')score = self.minimax(board_copy, True, alpha, beta)best_score = min(best_score, score)beta = min(beta, best_score)if beta <= alpha:breakreturn best_scoreclass Game:def __init__(self, size=3, ai_mode=False):self.board = Board(size)self.ai_mode = ai_modeself.player_marks = ['X', 'O']self.player_turn = 0if ai_mode:self.ai_player = AIPlayer(self.player_marks[1])def start(self):print("Starting the game...")self.board.display()while not self.board.is_full() and self.board.check_win() == "No one wins":current_player_mark = self.player_marks[self.player_turn]if self.player_turn == 1 and self.ai_mode:print("AI is thinking...")row, col = self.ai_player.get_best_move(self.board)print(f"AI placed {current_player_mark} at row {row + 1}, col {col + 1}")self.board.set_board(row, col, current_player_mark)else:index = input(f"Please player {self.player_turn + 1} input coordinates (row, col): ")try:row, col = index.split(',')row, col = int(row) - 1, int(col) - 1except ValueError:print("Invalid input. Please input two integers separated by a comma.")continueif not self.board.can_set(row, col):print("This cell is already occupied. Please choose another one.")continueself.board.set_board(row, col, current_player_mark)self.board.display()self.player_turn = (self.player_turn + 1) % 2print(self.board.check_win())if __name__ == '__main__':game = Game(size=3, ai_mode=True)game.start()

说明

第二部分(人机对战)全部由ChatGPT生成（中间经过多次调整），第一部分的代码是ChatGPT在我的代码基础上优化后的结果。
对于简单的问题，ChatGPT通常能给出完美的回答。但对于复杂问题则相对差一些，需要多次修改。此外，ChatGPT无法保证内容的正确性，有时候会给出似是而非的误导性回答。值得一提的是，ChatGPT对代码的理解能力特别强，能很好地给自己的代码加注释（大概是因为代码比较规范，流程相对固定，比较容易学习）。

ChatGPT能提高写代码及其它文字的效率，可以用来

注释代码
写一段简单函数
提供灵感（比如起标题）
写格式比较固定的文字（邮件等）

在这里插入图片描述

但是，ChatGPT对事实类问题的回答比较离谱：
在这里插入图片描述

在这里插入图片描述

更多ChatGPT的案例可以看看：
ChatGPT评测观察之对话能力｜语义理解较准，尚难以摆脱知识整合和逻辑推理困境 https://mp.weixin.qq.com/s/ZjZgMZIXiD966hcFua_gew

ChatGPT是一种生成式预训练transformer(generative pre-trained transformer, GPT)，使用有监督学习和强化学习对GPT-3.5进行微调。在两种方法中，人类训练者都用于提高模型的性能。在有监督学习中，模型被提供了对话，训练者在其中扮演了用户和AI助手的双方角色。在强化学习步骤中，人类训练者首先对模型先前创建的响应进行排名。这些排名被用于创建“奖励模型”，…
此外，OpenAI继续收集来自ChatGPT用户的数据，这些数据可以用于进一步训练和微调ChatGPT。用户可以赞或踩他们从ChatGPT收到的响应；在赞或踩时，他们还可以填写一个文本字段以提供额外的反馈。
更多细节可以参考:
维基百科ChatGPT： https://en.wikipedia.org/wiki/ChatGPT
InstructGPT论文: Training language models to follow instructions with human feedback

最后提一下，井字棋的最优策略竟是先占角！
https://www.guokr.com/article/4754