SmartNews’ Global Trust and Safety Head prioritizes user transparency and uses artificial intelligence (AI ) are calling for new regulations.
“We need guardrails,” said Arjun Narayan. “If humans fail to think through all the things that can go wrong, such as biases creeping into models and large language models falling into the wrong hands, there can be very serious consequences for humanity.”
Narayan, who was previously the trust and safety officer at Bytedance, which runs Google and TikTok, said it is essential for businesses to be aware of opt-ins and opt-outs when using Large Language Models (LLMs). said. By default, whatever is fed into LLM is considered training data and collected by the model. Narayan said many companies, especially start-ups, will have to opt out to avoid leaks and ensure confidentiality.
AI Around the World: How the US, EU and China Plan to Regulate AI Software Companies
Arjun Narayan, Head of Global Reliability and Safety at SmartNews, previously worked on reliability and safety at Google and ByteDance. (Fox News Digital)
Narayan also said regulators should be transparent about how algorithms are trained. If an algorithm uses his SmartNews data or another company’s data, that company needs to know if that’s happening and if they’ll be compensated.
“You don’t want the world to know what your next new product launch is,” he says.
As new applications of AI continue to be developed, Narayan emphasized the need to improve media and consumer literacy. In recent months, several countries, including China and Kuwait, have released AI news anchors that use natural language processing and deep learning to create lifelike sounds and actions to deliver news to viewers.
Narayan warned that scammers and other malicious actors could abuse the technology to use Catfish to gain access to personal information. Cases like this highlight that the lines between AI and human-generated media are disappearing.
What is Google Bard? How AI chatbots work, how they are used, and why they are controversial
Illustration of users participating in social media. (St. Petersburg)
“It’s kind of a fascinating space, and as long as users know what they’re building, who they’re talking to, what’s real and what’s not, I think that’s fine. But if users don’t know, we are in big trouble,” Narayan said.
He also talked about some other use cases for AI in media and news. For example, journalists can use AI for investigations or extract data from existing data. Additionally, some companies ask AI to create partial content, and the model creates a draft, which humans edit and publish.
Other companies, such as SmartNews, use AI to curate and select articles that matter to users based on unique signals and reader categories. Narayan said the recommender system’s process often uses a combination of different algorithms for “hyper-personalization.”
In the extreme case of social media, these AI-curated filter bubbles are set entirely to the user’s interests. Narayan used soda as an example of this process. Suppose the user likes soda. The AI will give the user a soda whether it’s good or bad. As the user keeps drinking that soda, the recommender’s algorithm in his system keeps pushing that product. Narayan also said the long-term psychological effects of this type of model are unknown.
What are the dangers of AI? Find out why people fear artificial intelligence
Close-up of a software engineer working on an AI algorithm. (St. Petersburg)
Narayan has spent years ensuring that reputation and user safety are top priorities within the various organizations he has worked for. Part of that responsibility includes assessing fairness, accountability and bias. He emphasized that it is humans who set the rules for AI models. If a company wants emotional or criminal stories for engagement, the system will pick just that.
“In some cases, AI obviously optimizes engagement, but that also means biasing towards certain types of stories,” Narayan said. “It will set a certain tone for your platform. And the question worth asking is, ‘Is that the right tone for your media outlet? Is it the right tone for your platform?’ I don’t have the answer, but this is a question that companies and editorial boards need to take initiative on. ”
As for “robot reporters,” editors who use AI to create their content entirely, Narayan said the train had already left the station. A quick Google search reveals that Amazon Books and online news platforms already have hundreds of his works and novels written entirely by AI.
Speaking to SmartNews, Narayan said AI isn’t being used for fact-checking not because we don’t trust it, but because it’s not accurate enough right now. He added that the human element is needed to proofread the report for accuracy and ensure that all checks related to journalism values, diversity and political bias are checked.
Schumer announces bipartisan senator-only briefing on ‘amazing’ advances in AI
“I’ve seen other platforms that claim to be fully digital and AI-driven, and while they may be, I don’t think society is ready for that,” Narayan said. rice field. “AI is error prone. We still believe in human ingenuity, creativity and human oversight.”
AI is sometimes prone to quantitative errors, but Narayan said the technology works mostly with precision. In his experience, AI has been very helpful in conducting research, extracting facts, and evaluating large datasets. However, AI models are often trained using data from several years ago, so the model cannot answer questions about recent or current events. In that regard, latency issues make this model great for evergreen topics, but inefficient when it comes to breaking news, such as the recent mass shooting.
Narayan believes AI needs certain ground rules, such as a system that prioritizes user transparency so readers can judge the material fairly. When an article is completely written by an AI, there needs to be a way for publishers to indicate it so that the information is conveyed to consumers.
“It’s either watermarking or technology that can’t be exploited,” he said. “Then at least you will know that this content is AI-generated or that the student wrote an essay using ChatGPT and therefore has this watermark. We need professors who are competent, and I think these things are definitely what we do.” We should advocate, or at least promote, when it comes to regulation. ”
CLICK HERE TO GET THE FOX NEWS APP
