小程序
传感搜
传感圈

Google launches Bard chatbot in UK and US

2023-03-25
关注

Google has opened access to its long-awaited chatbot Bard to consumers in the US and the UK through a waitlist. It was previously only accessible to Google employees and a select group of testers. It’s the search giant’s response to the hugely successful ChatGPT from OpenAI and Microsoft’s Bing Search AI tool built on top of OpenAI’s GPT-4 large language model.

Bard is capable of responding to a range of queries and has live information access via Google Search
Bard is capable of responding to a range of queries and has live information access via Google Search. (Screenshot by Tech Monitor)

For now, Bard is available via a separate interface but Google is expected to integrate it into its main search page in a similar way to Microsoft and Bing. It has been described as an experiment for collaboration on generative AI and is branded as such. Bard itself declared at the end of many of its responses while being tested by Tech Monitor: “I am still under development, and I am always learning new things and may make mistakes.”

There were several instances where it made up information, including around its ability to analyse video – generating a fake quote and attributing it to the speaker in a YouTube video it was asked to review. It also struggled with information on other large language models, but Google says it is still in development and warns on every page it may make mistakes.

It was revealed amid a flurry of AI announcements in the past two weeks including Google opening up its massive 530bn-parameter PaLM large language model API to a select group of developers and Microsoft integrating its own ‘Copilot’ AI built on GPT-4 into its Microsoft 365 product range.

Bard is built on top of a lightweight and optimised version of Google’s smaller LaMDA language model but the aim is to update it to more powerful models over time. Google says this is to allow it to test how people interact with the model. “We’re using human feedback and evaluation to improve our systems, and we’ve also built in guardrails, like capping the number of exchanges in a dialogue, to try to keep interactions helpful and on topic,” the company wrote in a blog post.

Google Bard will be upgraded over time

Described as a “complimentary experience to Google Search”, Bard is a direct interface to a large language model in a way that is easier for humans to understand. It is designed so users can visit Search to check responses or explore sources quoted by the AI. After each generated response a list of search terms is displayed and opened in a new tab with relevant results.

“We’ll continue to improve Bard and add capabilities, including coding, more languages and multimodal experiences. And one thing is certain: We’ll learn alongside you as we go. With your feedback, Bard will keep getting better and better,” the company said.

It has tough competition, coming up against OpenAI’s own ChatGPT which is now powered by GPT-4 and reached over 100 million active monthly users within three months of being launched. It is also competing with other LLM chatbots from the likes of Anthropic with Claude+ and Microsoft with Bing Search AI and the Copilot-powered Business Chat.

Content from our partners

Fashion brands must seek digital solutions that understand the sector’s unique needs

Fashion brands must seek digital solutions that understand the sector’s unique needs

Banks must better balance compliance with customer outreach

Banks must better balance compliance with customer outreach

How one manufacturer transformed data capabilities by moving to the cloud

How one manufacturer transformed data capabilities by moving to the cloud

Speaking to Reuters, Google product director Jack Krawczyk said the focus was on users rather than the competition. He said Bard was a way to help users in “boosting their productivity, accelerating their ideas, really fuelling their curiosity”.

View all newsletters Sign up to our newsletters Data, insights and analysis delivered to you By The Tech Monitor team

Unlike ChatGPT, the Google tool is able to show three alternative “draft” responses that it considered before sending the final response. It also generates automatically rather than line-by-line as is the case with Bing and the OpenAI chatbot.

While it can generate code in a handful of programming languages, it doesn’t have the proficiency of ChatGPT and has limited memory of previous conversations. It can also only work in a handful of spoken languages although Google promises this will change as the tool evolves.

Google is encouraging users of the tool to click “thumbs down” on any inaccurate answers or even inaccurate responses such as producing more or fewer lines of text than asked for. “We know the limitations of the technology, and so we want to be very deliberate at the pace at which we roll this out,” said Krawcyzk.

Adnan Masood, chief AI architect at digital transformation company UST told Tech Monitor ChatGPT has “first-mover advantage”, with Google Bard emerging into a crowded market, but one with potential to grow. “With the advent of GPT-4, the fusion of text, images, and other data modalities through multi-modal AI heralds a new dawn of contextual intelligence, empowering businesses to make better-informed decisions and forge stronger connections with their customers. For example, in human resources, AI’s capability to assess resumes, cover letters, and even video interviews enables businesses to identify top talent efficiently,” Adnan said.

“Despite the competition from these alternative AI solutions, an even more significant challenge may arise from Stanford’s Alpaca-style models, which can run on-premises using cost-effective hardware and without the restrictions or safety boundaries associated with more established AI systems,” he said.

Testing the chatbots

Tech Monitor gave each of ChatGPT, Bing, Bard and Claude+, through the Quora chatbot tool Poe the same prompt: Write a one-paragraph summary of the global enterprise technology sector. The responses varied in detail, creativity and length but each gave a good attempt and provided what was asked. For the experiment, we used Bing on its “precise” setting.

All of the responses are from that basic prompt without any further “engineering” to get a more detailed, more specific or more concise response. Some tools like Bard benefit from more detailed requests, whereas GPT-4 is better at understanding sentiment. It isn’t clear whether direct access to Claude+ would have generated a different response as we only had access via Poe. Here are the responses we received:

Bard:

The global enterprise technology sector is a rapidly growing and ever-changing field. It encompasses a wide range of technologies and services that are used by businesses of all sizes to improve efficiency, productivity, and profitability. Some of the key areas of the enterprise technology sector include cloud computing, artificial intelligence, big data, and cybersecurity.

The global enterprise technology sector is expected to continue to grow in the coming years, driven by the increasing adoption of cloud computing, artificial intelligence, and other emerging technologies. As businesses become more reliant on technology, the enterprise technology sector will play an increasingly important role in their success.

GPT-4:

The global enterprise technology sector has experienced significant growth in recent years, driven by rapid digital transformation, increasing adoption of cloud computing, and a relentless focus on data-driven decision-making. This dynamic industry encompasses a wide array of cutting-edge solutions, ranging from artificial intelligence (AI) and machine learning (ML) to cybersecurity, Internet of Things (IoT), and advanced analytics. As businesses across the world strive to streamline their operations, enhance customer experiences, and maintain competitive advantage, the demand for innovative enterprise technologies continues to surge. The sector has also seen notable consolidation, with major players like Microsoft, Amazon, Google, and IBM expanding their portfolios through strategic acquisitions and partnerships. As technology evolves and regulatory landscapes shift, the enterprise technology sector remains poised to redefine traditional business practices and drive global economic growth.

Claude+ from Anthropic:

The enterprise technology sector continues to grow rapidly as companies seek to become more digital, efficient, and innovative. This includes areas such as cloud computing, big data and analytics, mobility, security, and enterprise software. Major tech companies like Amazon, Microsoft, and Google are providing platforms and services to help other businesses transform themselves. Enterprise spending on technology is in the trillions of dollars annually and still rising. Emerging technologies around artificial intelligence, blockchain, virtual/augmented reality are starting to see increased adoption. At the same time, managing IT infrastructure and integrating new systems remains challenging. Overall, the enterprise technology sector is vibrant, competitive, and crucial for business and economic progress around the world.

Bing:

The global enterprise technology sector is a rapidly growing and changing industry that plays a central role in developing business standards and regulations. Technology continues to be a primary catalyst for change in the world, giving businesses governments, and social-sector institutions more possibilities to lift their productivity, invent and reinvent offerings, and contribute to humanity’s well-being. The tech sector is extending its reach into other industries using digital advancements to support innovation and transformation.

(Homepage image by gguy/Shutterstock)

Read more: UK AI taskforce launch by government to look at foundation models

Topics in this article : AI , Google

参考译文
谷歌在英国和美国推出巴德聊天机器人
谷歌已通过等待名单向美国和英国消费者开放了其期待已久的聊天机器人巴德(Bard)。它以前只对谷歌的员工和一组选定的测试人员开放。这是这家搜索巨头对OpenAI的ChatGPT和微软基于OpenAI GPT-4大型语言模型开发的必应人工智能搜索工具取得巨大成功的回应。目前,Bard可以通过一个单独的界面访问,但谷歌预计将以类似于微软和必应的方式将其集成到其主搜索页面中。它被描述为生成式人工智能合作的一个实验,并被贴上了这样的标签。在接受Tech Monitor的测试时,巴德自己在许多回复的结尾宣称:“我仍在开发中,我一直在学习新东西,可能会犯错误。”有几个例子是它编造信息的,包括关于它分析视频的能力——生成一个虚假的引用,并将其归因于它被要求审查的YouTube视频中的演讲者。谷歌在其他大型语言模型的信息方面也遇到了困难,但谷歌表示它仍在开发中,并在每一页上警告可能会出错。这是在过去两周一系列人工智能公告中披露的,包括谷歌向一组精选的开发人员开放其庞大的5300亿参数PaLM大型语言模型API,以及微软将其自己的基于gbt -4的“Copilot”人工智能集成到其微软365产品系列中。Bard构建在谷歌较小LaMDA语言模型的轻量级优化版本之上,但目标是随着时间的推移将其更新为更强大的模型。谷歌表示,这是为了测试人们如何与模型交互。该公司在一篇博客文章中写道:“我们正在利用人类的反馈和评估来改进我们的系统,我们还建立了护栏,比如限制对话中的交流次数,以努力保持互动的有益性和正题性。”Bard被描述为“谷歌搜索的免费体验”,它以一种更容易理解的方式直接连接到大型语言模型。它的设计让用户可以访问搜索来检查回复或探索人工智能引用的资源。在每个生成的响应之后,将显示一个搜索词列表,并在一个新选项卡中打开相关结果。“我们将继续改进Bard并增加功能,包括编码、更多语言和多模式体验。有一件事是肯定的:我们会和你们一起学习。有了你们的反馈,巴德会变得越来越好。”它面临着激烈的竞争,对手是OpenAI自己的ChatGPT,后者现在由GPT-4提供支持,在发布三个月内就拥有超过1亿的月活跃用户。它还与其他LLM聊天机器人竞争,比如Anthropic的Claude+,微软的必应搜索人工智能和副驾驶驱动的商务聊天。谷歌产品总监杰克·克劳奇克在接受路透社采访时表示,公司的重点是用户,而不是竞争对手。他说,巴德是一种帮助用户“提高他们的生产力,加速他们的想法,真正激发他们的好奇心”的方式。与ChatGPT不同,谷歌工具能够在发送最终响应之前显示三个备选的“草稿”响应。它还能自动生成,而不是像必应和OpenAI聊天机器人那样逐行生成。虽然它可以用少数几种编程语言生成代码,但它不具备ChatGPT的熟练程度,而且对以前对话的内存有限。它也只能在少数几种口语中工作,尽管谷歌承诺,随着工具的发展,这种情况将会改变。 谷歌鼓励该工具的用户对任何不准确的答案,甚至是不准确的回答,比如产生的文本行数比要求的多或少。克劳西茨克说:“我们知道这项技术的局限性,所以我们希望在推广这项技术的速度上非常谨慎。”数字转型公司UST的首席人工智能架构师Adnan Masood告诉Tech Monitor, ChatGPT具有“先发优势”,谷歌Bard进入了一个拥挤的市场,但有增长潜力。“随着GPT-4的出现,通过多模态人工智能融合文本、图像和其他数据模式,预示着上下文智能的新曙光,使企业能够做出更明智的决策,并与客户建立更强的联系。例如,在人力资源领域,人工智能评估简历、求职信甚至视频面试的能力使企业能够有效地识别顶级人才。尽管这些替代人工智能解决方案存在竞争,但斯坦福大学的羊驼式模型可能会带来更大的挑战,该模型可以使用具有成本效益的硬件在内部运行,并且没有与更成熟的人工智能系统相关的限制或安全边界,”他说。通过Quora聊天机器人工具Poe, Tech Monitor给ChatGPT、Bing、Bard和Claude+每个人都给出了同样的提示:写一段全球企业技术领域的摘要。这些回答在细节、创意和长度上各不相同,但每个人都做出了很好的尝试,并提供了所问的问题。在实验中,我们使用必应的“精确”设置。所有的回复都来自这个基本的提示,没有任何进一步的“设计”来获得更详细、更具体或更简洁的回复。一些工具,如Bard,受益于更详细的请求,而GPT-4更擅长理解情绪。目前尚不清楚直接访问Claude+是否会产生不同的反应,因为我们只能通过Poe访问。以下是我们收到的回答:全球企业技术领域是一个快速增长和不断变化的领域。它包含广泛的技术和服务,被各种规模的企业用于提高效率、生产力和盈利能力。企业技术领域的一些关键领域包括云计算、人工智能、大数据和网络安全。在云计算、人工智能和其他新兴技术的日益普及的推动下,全球企业技术部门预计将在未来几年继续增长。随着企业越来越依赖技术,企业技术部门将在其成功中发挥越来越重要的作用。近年来,全球企业技术部门在快速数字化转型、云计算越来越多的采用以及对数据驱动决策的不懈关注的推动下,实现了显著增长。这个充满活力的行业包含了广泛的前沿解决方案,从人工智能(AI)和机器学习(ML)到网络安全、物联网(IoT)和高级分析。随着世界各地的企业努力简化运营、增强客户体验并保持竞争优势,对创新企业技术的需求持续激增。该行业也出现了显著的整合,微软、亚马逊、谷歌和IBM等主要参与者通过战略收购和合作扩大了他们的投资组合。随着技术的发展和监管环境的变化,企业技术部门仍然准备重新定义传统的商业实践,并推动全球经济增长。 随着企业寻求更加数字化、高效和创新,企业技术部门继续快速增长。这包括云计算、大数据和分析、移动性、安全性和企业软件等领域。亚马逊、微软和谷歌等大型科技公司正在提供平台和服务,帮助其他企业转型。企业每年在技术上的支出高达数万亿美元,而且还在不断增长。围绕人工智能、区块链、虚拟/增强现实的新兴技术开始越来越多地采用。与此同时,管理IT基础设施和集成新系统仍然具有挑战性。总的来说,企业技术部门充满活力,具有竞争力,对世界各地的商业和经济发展至关重要。全球企业技术部门是一个快速增长和不断变化的行业,在制定业务标准和法规方面发挥着核心作用。技术仍然是世界变革的主要催化剂,为企业、政府和社会部门机构提供了更多的可能性来提高生产力、发明和改造产品,并为人类福祉做出贡献。科技行业正在利用数字技术的进步来支持创新和转型,将其触角延伸到其他行业。(主页图片来自gguy/Shutterstock)
您觉得本篇内容如何
评分

评论

您需要登录才可以回复|注册

提交评论

techmonitor

这家伙很懒,什么描述也没留下

关注

点击进入下一篇

机器人可以代替人类的哪些工作呢?

提取码
复制提取码
点击跳转至百度网盘