OpenAI, the maker of ChatGPT, has revealed a new artificial intelligence tool. Known as Sora, it's designed to convert text prompts into short high-quality videos. Reporters Yang Chengxi and Liu Jiaxin ask: what could the new technology mean for the future of video journalism?
YANG CHENGXI Beijing "In journalism, we have a saying: report the story, don't become part of it. But this time, it feels a bit different. As AI disruption takes aim at our industry, many of us reporters ask ourselves: what impact could AI videos have on journalism, and people's understanding of whether what we show them, is or isn't real. And could we be out of a job one day?"
It's not often that we become the interviewees. I first asked our in-house cameraman about his thoughts on these Sora-generated videos.
WANG LEI Cameraman "From a professional point of view, some of the perspectives in the videos are illogical. However, for the majority of the audience, these images can already pass as authentic."
I then sat down with fellow tech reporter Jiaxin.
"So what do you make of it?"
LIU JIAXIN Beijing "I feel like, for the first time, AI videos are becoming capable of challenging perceptions of reality. Sure, you can still detect flaws, like over here: AI still can't get text characters right. It's all still gibberish. But look at the overall image! Considering how previous AI videos from just a few months ago are just so creepy and bad. I think Sora could be the start of something big."
YANG CHENGXI Beijing "That could be a fake news nightmare, no? Remember those 'AI Trump arrest' images that went viral? Now imagine you have videos from multiple angles of the same 'incident': TV livestreams, shaky phone recordings and drone footage... All as seemingly as real as they could be."
LIU JIAXIN Beijing "Yeah, and it gets even more terrifying when you consider that one of Sora's highlights is the option to alter videos and artificially extend them forward or backward in time. So now an 'AI Trump arrest' video could even have multiple fake endings."
YANG CHENGXI Beijing "I'm sure Sora will open a new can of worms for the world's tech regulators. I talked to some of them last year, and they've been stressing the urgency of joint AI governance. And now with Sora, I guess they'll have much more to talk about."
LIU JIAXIN Beijing "I think that's why OpenAI is not making the tool available to the public just yet. The company says it's working on combatting misinformation, hateful content and biases online. It says the new video model will reject prompts that request the creation of celebrity lookalikes. So, we may not actually see Sora-generated Trump videos anytime soon or will we?"
YANG CHENGXI Beijing "I just hope we won't be proven wrong in a year or so. So, what do you think this could mean for video journalism?"
LIU JIAXIN Beijing "I'll be honest, it's hard not to feel some kind of existential crisis at first."
YANG CHENGXI Beijing "So does everyone in our industry, I guess."
LIU JIAXIN Beijing "But it may be too early to be pessimistic. Because journalism, especially live reporting, is about recording things in the real world as they happen, and telling stories that have not been told. So how can AI generate what's not even in the database yet?"
YANG CHENGXI Beijing "That's really kind of a paradox, isn't it? Sora trained itself by studying and imitating the work of humans and for the longest time, it's the random sparks of human creativity that have forged new techniques and advancements. If AI were to eventually put all artists and story-tellers out of business, would it have the ability to continue pushing the boundaries like we humans do?"
LIU JIAXIN Beijing "There are always more questions than answers."
"Ask ChatGPT?"
"Sure."