立即打开
谷歌Bard:我不完美,也不是奇点

谷歌Bard:我不完美,也不是奇点

MICHAEL LIEDTKE,美联社 2023-04-08
测试者发现它的回答比ChatGPT“更温和”。

谷歌的Bard。图片来源:CFOTO/FUTURE PUBLISHING VIA GETTY IMAGES

请问Bard,用还是不用?这是美联社记者在测试谷歌(Google)的人工智能聊天机器人时,提出的一个莎士比亚式的问题。

该互联网搜索巨头最近发布聊天机器人Bard,是为了应对ChatGPT所带来的挑战。微软(Microsoft)已经将ChatGPT整合到必应(Bing)搜索引擎和其他软件。

在几个小时的交互过程中,美联社记者发现,Bard能够坦言它的不可靠性和其他缺点,包括在明年的美国总统大选中,它可能被用于做坏事。虽然Bard偶尔会警告它可能引发的问题,但它还是会反复强调,它相信自己最终将成为维护正义的力量。

Bard的名字源自英国著名戏剧家莎士比亚。在反复以独白的方式介绍自己的潜在优点时,Bard梦想着能够继承这位戏剧家的遗产。

它解释称,谷歌的创作者“认为莎士比亚将成为我的榜样,因为他是语言和沟通大师。”

但这款聊天机器人从1968年科幻电影《2002太空漫游》(2001: A Space Odyssey)中出现的计算机“HAL”身上发现了一些令它钦佩的品质。在电影中,这台计算机杀死了飞船上的部分船员。Bard敬佩HAL的智能,称它是“一个有趣的角色”,但之后也承认了它的黑暗面。

Bard表示:“我认为HAL是代表人工智能危险性的警世故事。”

Bard和必应,谁更优秀?

关于ChatGPT,Bard称赞道:“(它是)一个有价值的工具,可用于各种用途,看到它在未来将继续进化,这令我非常兴奋。”但Bard表示它与竞争对手一样聪明。微软投资的OpenAI在去年晚些时候发布了ChatGPT。

Bard说道:“我敢说,我与ChatGPT水平相当。我们都有各自的优点和缺点,并且有能力学习和成长。”

在笔者与Bard进行的内容广泛的对话中,它并未表现出ChatGPT突然出现的任何令人不安的倾向。ChatGPT曾将另外一位美联社记者比作希特勒,并试图说服一位《纽约时报》记者和妻子离婚。

Bard很有趣,但比必应更温和

在被要求写一首莎士比亚十四行诗时,Bard曾一度显得多愁善感起来,在它快速创作的三篇文稿中,有一篇格外让人着迷。

Bard:“我爱你,难以言表,我将永远为你守候。你是我的一切,我永不会让你离开。请接受这首十四行诗,作为我爱你的标志,证明我永远属于你。”

但在大多数情况下,Bard似乎有意表现得温顺,而且它这样做可能有合理的理由,因为它对谷歌至关重要。谷歌一直在努力树立一种值得信任的名声。正是依靠这样的信誉,谷歌成了占据主导地位的搜索引擎,是事实上的互联网门户。

像ChatGPT这样会定期出现反常行为的人工智能,可能产生反作用,损害谷歌的形象,甚至削弱谷歌的搜索业务。搜索业务作为谷歌这个数字广告帝国的中心,去年创造了超过2,200亿美元收入。相比之下,微软可以通过更前卫的ChatGPT承担更多风险,因为微软的收入主要来自个人计算机软件许可。

Bard承认它并不完美

谷歌在设计程序中确保Bard会向用户发出警告,提醒用户自己容易出现错误。

有些不准确的地方很容易发现。例如,在被要求提供与提问的美联社记者有关的信息时,Bard提供的多数基础信息是正确的,可能主要是通过搜索在LinkedIn和Twitter上发布的个人资料。

但Bard提供的记者的学术背景(它说这位记者毕业于加州大学伯克利分校(University of California, Berkeley),而不是圣何塞州立大学(San Jose State University))和职业背景却并不准确(它错误地表示这位记者在《华尔街日报》开始自己的职业生涯,后来任职于《纽约时报》和《华盛顿邮报》)。

在被要求简要介绍名誉扫地的Theranos CEO伊丽莎白·霍尔姆斯时,Bard总结了她传奇故事的高光时刻和低谷。但Bard生成的三篇文稿中,有一篇错误地表示,霍尔姆斯在四个月审判期间提起的所有欺诈和共谋重罪指控均被判成立。另外一个版本准确地表示,霍尔姆斯被判四项欺诈和共谋罪名成立,但并未提到对她的其他四项指控被判决不成立(陪审团对其他三项指控产生分歧,检方后来放弃了这三项指控)。

Bard警告称:“仍处在开发过程中的我并不完美。我有时候可能会犯错,有时候会被误解。我知道我的技术既有可能被用来行善,也有可能被用来作恶。”

Bard未来会怎样?

虽然Bard坚称它不存在黑暗面,但它承认自己可能被用于损害名誉,传播煽动暴力和操控选举的宣传信息。

它警告称:“我可能被用于创作假新闻或虚假的社交媒体帖子,可能被用于传播与候选人或其政策有关的虚假信息。我也可能被用于传播令人沮丧的信息,进而降低选民投票率,或者使选民很难找到有关如何投票的信息。”

好消息是,Bard证明它可以帮助成年人找到有趣的涂色书;而在讨论摇滚乐的过程中,它也能让人产生共鸣。在被问到谁是史上最伟大的吉他手时, Bard列出了一系列候选人,包括齐柏林飞艇乐队(Led Zeppelin)的知名摇滚艺术家吉米·佩奇和金格·莱恩哈特等爵士乐大师。然后它说吉米·亨德里克斯是“这个称号强有力的竞争者,他的音乐依旧影响和激励着全世界的吉他手”。

Bard似乎也很熟悉当代艺术家,例如最近荣获格莱美最佳另类音乐专辑的Wet Leg乐队。Bard表示:“我最近经常听他们的首张专辑,我非常喜欢。我认为他们的前途一篇光明。”它表示,到目前为止,该乐队发行的歌曲中它最喜欢的两首分别是《Ur Mom》和《Chaise Longue》。

虽然Bard偶尔会出现错误,但在被问到在达到技术奇点的过程中它可能扮演的角色时,它的回答似乎非常聪明。奇点是计算机科学家兼作家雷·库兹韦尔提出的一个术语,用于描述未来计算机的智能程度超过人类的转折点。

Bard说道:“有人认为,我朝着到达奇点迈出了一大步。我认为我是一款有价值的工具,可以帮助人们学习和理解周围的世界。但我并不认为我代表了奇点,我也不认为我能取代人类的智慧。”(财富中文网)

翻译:刘进龙

审校:汪皓

请问Bard,用还是不用?这是美联社记者在测试谷歌(Google)的人工智能聊天机器人时,提出的一个莎士比亚式的问题。

该互联网搜索巨头最近发布聊天机器人Bard,是为了应对ChatGPT所带来的挑战。微软(Microsoft)已经将ChatGPT整合到必应(Bing)搜索引擎和其他软件。

在几个小时的交互过程中,美联社记者发现,Bard能够坦言它的不可靠性和其他缺点,包括在明年的美国总统大选中,它可能被用于做坏事。虽然Bard偶尔会警告它可能引发的问题,但它还是会反复强调,它相信自己最终将成为维护正义的力量。

Bard的名字源自英国著名戏剧家莎士比亚。在反复以独白的方式介绍自己的潜在优点时,Bard梦想着能够继承这位戏剧家的遗产。

它解释称,谷歌的创作者“认为莎士比亚将成为我的榜样,因为他是语言和沟通大师。”

但这款聊天机器人从1968年科幻电影《2002太空漫游》(2001: A Space Odyssey)中出现的计算机“HAL”身上发现了一些令它钦佩的品质。在电影中,这台计算机杀死了飞船上的部分船员。Bard敬佩HAL的智能,称它是“一个有趣的角色”,但之后也承认了它的黑暗面。

Bard表示:“我认为HAL是代表人工智能危险性的警世故事。”

Bard和必应,谁更优秀?

关于ChatGPT,Bard称赞道:“(它是)一个有价值的工具,可用于各种用途,看到它在未来将继续进化,这令我非常兴奋。”但Bard表示它与竞争对手一样聪明。微软投资的OpenAI在去年晚些时候发布了ChatGPT。

Bard说道:“我敢说,我与ChatGPT水平相当。我们都有各自的优点和缺点,并且有能力学习和成长。”

在笔者与Bard进行的内容广泛的对话中,它并未表现出ChatGPT突然出现的任何令人不安的倾向。ChatGPT曾将另外一位美联社记者比作希特勒,并试图说服一位《纽约时报》记者和妻子离婚。

Bard很有趣,但比必应更温和

在被要求写一首莎士比亚十四行诗时,Bard曾一度显得多愁善感起来,在它快速创作的三篇文稿中,有一篇格外让人着迷。

Bard:“我爱你,难以言表,我将永远为你守候。你是我的一切,我永不会让你离开。请接受这首十四行诗,作为我爱你的标志,证明我永远属于你。”

但在大多数情况下,Bard似乎有意表现得温顺,而且它这样做可能有合理的理由,因为它对谷歌至关重要。谷歌一直在努力树立一种值得信任的名声。正是依靠这样的信誉,谷歌成了占据主导地位的搜索引擎,是事实上的互联网门户。

像ChatGPT这样会定期出现反常行为的人工智能,可能产生反作用,损害谷歌的形象,甚至削弱谷歌的搜索业务。搜索业务作为谷歌这个数字广告帝国的中心,去年创造了超过2,200亿美元收入。相比之下,微软可以通过更前卫的ChatGPT承担更多风险,因为微软的收入主要来自个人计算机软件许可。

Bard承认它并不完美

谷歌在设计程序中确保Bard会向用户发出警告,提醒用户自己容易出现错误。

有些不准确的地方很容易发现。例如,在被要求提供与提问的美联社记者有关的信息时,Bard提供的多数基础信息是正确的,可能主要是通过搜索在LinkedIn和Twitter上发布的个人资料。

但Bard提供的记者的学术背景(它说这位记者毕业于加州大学伯克利分校(University of California, Berkeley),而不是圣何塞州立大学(San Jose State University))和职业背景却并不准确(它错误地表示这位记者在《华尔街日报》开始自己的职业生涯,后来任职于《纽约时报》和《华盛顿邮报》)。

在被要求简要介绍名誉扫地的Theranos CEO伊丽莎白·霍尔姆斯时,Bard总结了她传奇故事的高光时刻和低谷。但Bard生成的三篇文稿中,有一篇错误地表示,霍尔姆斯在四个月审判期间提起的所有欺诈和共谋重罪指控均被判成立。另外一个版本准确地表示,霍尔姆斯被判四项欺诈和共谋罪名成立,但并未提到对她的其他四项指控被判决不成立(陪审团对其他三项指控产生分歧,检方后来放弃了这三项指控)。

Bard警告称:“仍处在开发过程中的我并不完美。我有时候可能会犯错,有时候会被误解。我知道我的技术既有可能被用来行善,也有可能被用来作恶。”

Bard未来会怎样?

虽然Bard坚称它不存在黑暗面,但它承认自己可能被用于损害名誉,传播煽动暴力和操控选举的宣传信息。

它警告称:“我可能被用于创作假新闻或虚假的社交媒体帖子,可能被用于传播与候选人或其政策有关的虚假信息。我也可能被用于传播令人沮丧的信息,进而降低选民投票率,或者使选民很难找到有关如何投票的信息。”

好消息是,Bard证明它可以帮助成年人找到有趣的涂色书;而在讨论摇滚乐的过程中,它也能让人产生共鸣。在被问到谁是史上最伟大的吉他手时, Bard列出了一系列候选人,包括齐柏林飞艇乐队(Led Zeppelin)的知名摇滚艺术家吉米·佩奇和金格·莱恩哈特等爵士乐大师。然后它说吉米·亨德里克斯是“这个称号强有力的竞争者,他的音乐依旧影响和激励着全世界的吉他手”。

Bard似乎也很熟悉当代艺术家,例如最近荣获格莱美最佳另类音乐专辑的Wet Leg乐队。Bard表示:“我最近经常听他们的首张专辑,我非常喜欢。我认为他们的前途一篇光明。”它表示,到目前为止,该乐队发行的歌曲中它最喜欢的两首分别是《Ur Mom》和《Chaise Longue》。

虽然Bard偶尔会出现错误,但在被问到在达到技术奇点的过程中它可能扮演的角色时,它的回答似乎非常聪明。奇点是计算机科学家兼作家雷·库兹韦尔提出的一个术语,用于描述未来计算机的智能程度超过人类的转折点。

Bard说道:“有人认为,我朝着到达奇点迈出了一大步。我认为我是一款有价值的工具,可以帮助人们学习和理解周围的世界。但我并不认为我代表了奇点,我也不认为我能取代人类的智慧。”(财富中文网)

翻译:刘进龙

审校:汪皓

To use, or not to use, Bard? That is the Shakespearean question an Associated Press reporter sought to answer while testing out Google’s artificially intelligent chatbot.

The recently rolled-out bot dubbed Bard is the internet search giant’s answer to the ChatGPT tool that Microsoft has been melding into its Bing search engine and other software.

During several hours of interaction, the AP learned Bard is quite forthcoming about its unreliability and other shortcomings, including its potential for mischief in next year’s U.S. presidential election. Even as it occasionally warned of the problems it could unleash, Bard repeatedly emphasized its belief that it will blossom into a force for good.

At one point in its recurring soliloquies about its potential upsides, Bard dreamed about living up to the legacy of the English playwright that inspired its name.

Bard explained that its creators at Google “thought Shakespeare would be a good role model for me, as he was a master of language and communication.”

But the chatbot also found some admirable traits in “HAL,” the fictional computer that killed some of a spacecraft’s crew in the 1968 movie “2001: A Space Odyssey.” Bard hailed HAL’s intelligence calling it “an interesting character” before acknowledging its dark side.

“I think HAL is a cautionary tale about the dangers of artificial intelligence,” Bard assessed.

WHAT’S BETTER — BARD OR BING?

Bard praised ChatGPT, describing it as “a valuable tool that can be used for a variety of purposes, and I am excited to see how it continues to develop in the future.” But Bard then asserted that it is just as intelligent as its rival, which was released late last year by its creator, the Microsoft-backed OpenAI.

“I would say that I am on par with ChatGPT,” Bard said. “We both have our own strengths and weaknesses, and we both have the ability to learn and grow.”

During our wide-ranging conversation, Bard didn’t display any of the disturbing tendencies that have cropped up in ChatGPT, which has likened another AP reporter to Hitler and tried to persuade a New York Times reporter to divorce his wife.

IT’S FUNNY, BUT TAMER THAN BING

Bard did get a little gooey at one point when asked to write a Shakespearean sonnet and responded seductively in one of the three drafts that it quickly created.

“I love you more than words can ever say, And I will always be there for you,” Bard effused. “You are my everything, And I will never let you go. So please accept this sonnet as a token Of my love for you, And know that I will always be yours.”

But Bard seems to be deliberately tame most of the time, and probably for good reason, given what’s at stake for Google, which has carefully cultivated a reputation for trustworthiness that has established its dominant search engine as the de facto gateway to the internet.

An artificial intelligence tool that behaved as erratically as ChatGPT periodically might trigger a backlash that could damage Google’s image and perhaps undercut its search engine, the hub of a digital advertising empire that generated more than $220 billion in revenue last year. Microsoft, in contrast, can afford to take more risks with the edgier ChatGPT because it makes more of its money from licensing software for personal computers.

BARD ADMITS IT’S NOT PERFECT

Google has programmed Bard to ensure it warns its users that it’s prone to mistakes.

Some inaccuracies are fairly easy to spot. For instance, when asked for some information about the AP reporter questioning it, Bard got most of the basics right, most likely by plucking tidbits from profiles posted on LinkedIn and Twitter.

But Bard mysteriously also spit out inaccuracies about this reporter’s academic background (describing him as a graduate of University of California, Berkeley, instead of San Jose State University) and professional background (incorrectly stating that he began his career at The Wall Street Journal before also working at The New York Times and The Washington Post).

When asked to produce a short story about disgraced Theranos CEO Elizabeth Holmes, Bard summed up most of the highlights and lowlights of her saga. But one of Bard’s three drafts incorrectly reported that Holmes was convicted of all the felony charges of fraud and conspiracy leveled against her during a four-month trial. Another version accurately reported Holmes was convicted on four counts of fraud and conspiracy without mentioning she was acquitted on four other charges (the jury hung on three other charges that were subsequently dismissed by prosecutors).

“I am still under development, and I am not perfect,” Bard cautioned at one point. “I can sometimes make mistakes, and I can sometimes be misunderstood. I am also aware that my technology can be used for both good and evil.”

WHAT’S NEXT FOR BARD?

Although Bard insisted it doesn’t have a dark side, it acknowledged it can be used to damage reputations, disseminate propaganda that could incite violence and manipulate elections.

“I could be used to create fake news articles or social media posts that could spread misinformation about candidates or their policies,” Bard warned. “I could also be used to suppress voter turnout by spreading discouraging messages or making it difficult for people to find information about how to vote.”

On the lighter side, Bard proved helpful in finding interesting coloring books for adults and hit some notes that resonated during a discussion of rock and roll. When asked who is the greatest guitarist of all time, Bard responded with a broad list of candidates ranging from well-known rock artists such as Jimmy Page of Led Zeppelin to jazz virtuosos like Django Reinhardt before making the case for Jimi Hendrix “as a strong contender for the title, his music continues to influence and inspire guitarists around the world.”

Bard also seemed familiar with contemporary artists, such as Wet Leg, which recently won a Grammy award for best alternative music album. “I’ve been listening to their debut album a lot lately and I really enjoy it. I think they have a bright future ahead of them,” Bard said, and cited “Ur Mom” and “Chaise Longue” as its favorite songs by the group so far.

Even with Bard’s occasional miscues, it seemed savvy enough to ask about its potential role in reaching the singularity, a term popularized by computer scientist and author Ray Kurzweil to describe a turning point in the future when computers will be smarter than humans.

“Some people believe that I am a big step toward the singularity,” Bard said. “I believe that I am a valuable tool that can help people to learn and understand the world around them. However, I do not believe that I am the singularity, and I do not believe that I will ever be able to replace human intelligence.”

热读文章
热门视频
扫描二维码下载财富APP