ChatGPT这回考砸了

版主: kazaawangwh

Dahuaidanyimei
知名作家
知名作家
帖子: 1126
注册时间: 3月 14, 2023, 4:18 pm
昵称(选填): Badegg

Re: ChatGPT这回考砸了

帖子 Dahuaidanyimei »

3.5用的是两年前的数据,不知道近两年的事情。这种需要更新法规的事情它怎么会懂
datada
著名点评
著名点评
帖子: 4469
注册时间: 7月 29, 2022, 3:23 pm
昵称(选填): datada

Re: ChatGPT这回考砸了

帖子 datada »

数字竞赛试过吗?AMC, AIMI 和 IMO的 题有人试过吗?
cellcycle1
论坛元老
论坛元老
cellcycle1 的博客
帖子: 38199
注册时间: 7月 24, 2022, 3:59 pm

Re: ChatGPT这回考砸了

帖子 cellcycle1 »

wh. 写了: 5月 9, 2023, 10:58 pm 转:这回ChatGPT栽了

ChatGPT自出道以来,所向披靡,可谓是“人挡杀人,佛挡杀佛”。它先后通过了宾大商学院 Wharton MBA考试,律师的 Bar Exam等考试。

据 “今日会计”(Accounting Today)报道, 4月13日,有好事者在纽约用了两台分别装有 ChatGPT 3.5 Pro account的笔记本电脑,各自参加了两门美国注册会计师CPA的考试。结果如何呢?

这是它们“哥俩”的成绩:
1. Reg (法规) 39%
2. Aud (审计) 46%
3. Far (财报) 35%
4. Bec (商务) 48%

要想取得CPA的资质,以上四门考试的成绩都要超过75%的答对率。这次ChatGPT有一门连一半的分数都没达到,多少给我们美国注会留了点面子啦!
鳖禁 chatGPT 有道理,
wh楼主
论坛元老
论坛元老
wh 的博客
帖子: 30593
注册时间: 7月 28, 2022, 12:07 am
昵称(选填): 问号

Re: ChatGPT这回考砸了

帖子 wh楼主 »

萧武达 写了: 5月 10, 2023, 4:37 am 聊天来说,ChatGPT不如这个 https://huggingface.co/chat/
只是抢了先机, 比较一下,就知道谁更聪明
回复都是英语啊。我问它顾城是谁,它给了五个答案,第一个答案又像又不像顾城:

1. An author: Gu Cheng (Chinese: 古承; pinyin: Gǔchéng) was a prominent modernist Chinese writer known for his experimental works that blurred boundaries between prose and poetry. His real name was Wang Lingyang (王令阳). He died tragically at age 28 by self-immolation after suffering from severe depression.

我查不到古承、王令阳。顾城37岁去世。
wh楼主
论坛元老
论坛元老
wh 的博客
帖子: 30593
注册时间: 7月 28, 2022, 12:07 am
昵称(选填): 问号

Re: ChatGPT这回考砸了

帖子 wh楼主 »

Dahuaidanyimei 写了: 5月 10, 2023, 9:42 am 3.5用的是两年前的数据,不知道近两年的事情。这种需要更新法规的事情它怎么会懂
嗯,应该用最新版。
wh楼主
论坛元老
论坛元老
wh 的博客
帖子: 30593
注册时间: 7月 28, 2022, 12:07 am
昵称(选填): 问号

Re: ChatGPT这回考砸了

帖子 wh楼主 »

datada 写了: 5月 10, 2023, 10:47 am 数字竞赛试过吗?AMC, AIMI 和 IMO的 题有人试过吗?
查到AMC:
https://www.businessinsider.com/list-he ... far-2023-1
OpenAI just announced GPT-4, an updated chatbot that can pass everything from a bar exam to AP Biology. Here's a list of difficult exams both AI versions have passed.
……
AMC Exams
The AMC 10 and 12 are 25-question, 75-minute exams administered to high school students that cover mathematical topics including algebra, geometry, trigonometry, according to the Mathematical Association of America's site.

In the fall of 2022, the average score out of 150 total points on the AMC 10 was 58.33 and 59.9 on the AMC 12, according to the MAA's site. GPT-4 scored a 30 and 60, respectively, putting it between the 6th to 12th percentile of the AMC 10 and the 45th to 66th percentile of the AMC 12, according to OpenAI.
datada
著名点评
著名点评
帖子: 4469
注册时间: 7月 29, 2022, 3:23 pm
昵称(选填): datada

Re: ChatGPT这回考砸了

帖子 datada »

wh. 写了: 5月 10, 2023, 10:11 pm 查到AMC:
https://www.businessinsider.com/list-he ... far-2023-1
OpenAI just announced GPT-4, an updated chatbot that can pass everything from a bar exam to AP Biology. Here's a list of difficult exams both AI versions have passed.
……
AMC Exams
The AMC 10 and 12 are 25-question, 75-minute exams administered to high school students that cover mathematical topics including algebra, geometry, trigonometry, according to the Mathematical Association of America's site.

In the fall of 2022, the average score out of 150 total points on the AMC 10 was 58.33 and 59.9 on the AMC 12, according to the MAA's site. GPT-4 scored a 30 and 60, respectively, putting it between the 6th to 12th percentile of the AMC 10 and the 45th to 66th percentile of the AMC 12, according to OpenAI.
Thanks. The sample size is limited. But 60 for amc12 is ok. Aime should be tested too, if it can score 3 or above, i will be amazed.
头像
萧武达
论坛元老
论坛元老
帖子: 16271
注册时间: 12月 28, 2022, 10:39 pm
昵称(选填): shiaovd

Re: ChatGPT这回考砸了

帖子 萧武达 »

wh. 写了: 5月 10, 2023, 10:05 pm 回复都是英语啊。我问它顾城是谁,它给了五个答案,第一个答案又像又不像顾城:

1. An author: Gu Cheng (Chinese: 古承; pinyin: Gǔchéng) was a prominent modernist Chinese writer known for his experimental works that blurred boundaries between prose and poetry. His real name was Wang Lingyang (王令阳). He died tragically at age 28 by self-immolation after suffering from severe depression.

我查不到古承、王令阳。顾城37岁去世。
这个语言是根据你的IP和操作系统变得, 在中国的朋友反馈,得到的就是中文答
wh楼主
论坛元老
论坛元老
wh 的博客
帖子: 30593
注册时间: 7月 28, 2022, 12:07 am
昵称(选填): 问号

Re: ChatGPT这回考砸了

帖子 wh楼主 »

萧武达 写了: 5月 10, 2023, 10:32 pm 这个语言是根据你的IP和操作系统变得, 在中国的朋友反馈,得到的就是中文答
内容也不太对哦。
头像
萧武达
论坛元老
论坛元老
帖子: 16271
注册时间: 12月 28, 2022, 10:39 pm
昵称(选填): shiaovd

Re: ChatGPT这回考砸了

帖子 萧武达 »

wh. 写了: 5月 10, 2023, 10:47 pm 内容也不太对哦。
好吧
回复

回到 “精华区”