[丽莎]随着虚假信息担忧的加剧，语音克隆技术将为 2024 年政治广告提供动力

1 月 10 日与旧金山的领导人一起度过一个独特的社交之夜、见解和对话。

在此请求邀请。

人们对 2024 年美国大选中人工智能的使用的虚假信息担忧可能会日益加剧，但这并不能阻止人工智能语音克隆初创公司参与政治游戏。

例如，总部位于佛罗里达州博卡拉顿的人工智能音频/视频广告平台 Instreamatic 于 2021 年筹集了 610 万美元的 A 轮融资，该平台表示正在将其能力扩展到政治广告的狂野世界。

该解决方案使候选人竞选活动能够快速生成高度针对性的人工智能驱动的上下文视频和音频广告（以画外音为特色，而不是头部说话视频），以适应不断变化的事件或地点。

生成人工智能可以改变任何音频或视频政治广告

例如，这是 Instreamatic 分享的视频竞选演示，其中候选人（在本例中为巴拉克·奥巴马）可以通过复制声音来更改任何音频或视频政治广告，而无需进入工作室重新录制。

自去年 3 月以来，Instreamatic 已向品牌和代理商提供生成式语音 AI 产品。

它宣称，从一个完整的广告中，人工智能驱动的画外音可以自动创建无限的广告版本，其中可能包括受众的位置、一天中的时间、他们接收广告的应用程序或平台，或者最近的广告版本。店铺。

VB事件

人工智能影响之旅

制定 AI 治理蓝图 – 请求 1 月 10 日活动的邀请。

了解更多

然而，人工智能在 2024 年美国竞选活动中的使用预计将成为虚假信息雷区，并且已经发出了危险信号。

例如，美国广播公司新闻 (ABC News) 最近的一篇报道强调了佛罗里达州州长罗恩·德桑蒂斯 (Ron DeSantis) 今年夏天的竞选努力，其中包括人工智能生成的唐纳德·特朗普 (Donald Trump) 的图像和音频。

VentureBeat 最近采访了艾伦人工智能研究所的机器学习研究员 Nathan Lambert，他表示，无论是聊天机器人还是 Deepfakes，生成式 AI 都将使 2024 年美国大选变得“一团糟”。

Instreamatic 需要确认使用语音的权限

Stas Tushinskiy, CEO and co-founder of Instreamatic, insists the company has guardrails built in to make sure its product is not used for election disinformation.

“For any kind of campaign, whoever the client is, they have to confirm they have permission to use the voice,” he told VentureBeat. In addition, he said that the political advertising offering will not be available to everyone.

“You can’t just sign up,” he explained. “We will be engaged in campaign creation.” Instreamatic, he said, does not “want to get caught in the middle of something we didn’t intend the platform to be used for,” adding that if there were problems with political ads they would be “deleted immediately” on our hands ” and if necessary “we’ll make a public safety statement.”

Automating a manual process that already exists

Tushinskiy emphasized that Instreamatic is not reinventing the world of political ads to help candidates get elected. He described the company’s offering as automating a tedious manual process that already exists.

“This process involves somebody like a candidate or voice talent going to a studio to spend hours and hours in the studio, then someone else uploading them and someone else checking for human errors,” he said. “It’s an extensive and expensive process and we automated all of that,” compressing the process from six to eight weeks to a few minutes.

In addition, an ad campaign also requires a great deal of back and forth between the agency and the client, he explains, in which words may be changed, requiring new takes. But voice cloning allows an airline company, for example, to mention a variety of travel destinations in targeted ads, or car brands to cite local dealership locations. “Contextual ads always outperform generic ads, so it makes a lot of sense in terms of increasing the effectiveness of your ad spend,” he said.

Concerns about AI and election disinformation

Experts maintain the political ad landscape is fraught with potential AI-generated peril. For example, there are currently no federal rules for the use of AI-generated content, such as ads, in political campaigns.

Russell Wald, policy director at Stanford University’s Institute for Human Centered AI, told ABC News Live in November that “All campaigns can use this. So in that sense, who is setting the rules of the road as the campaigns themselves, as they go?”

But Tushinskiy said “if we were the ones that created misinformation, I wouldn’t want to be in this business process.” Instead, he maintained, “We’re just giving them the tools to be more effective.” And the moment Instreamatic catches somebody doing something unethical, “not only can we stop it, but we can also expose it.”

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.

热点：丽莎 提供