Remember the flood of uncanny AI-generated pictures in 2022? Brace yourselves for a similar wave, but this time with video. OpenAI’s new model, Sora, creates video based on prompts. While not publicly available yet, it’s sparking debate.
Some fear “disinformation waves,” but past predictions of AI-driven chaos haven’t always held true. Others point out Sora’s flaws, echoing similar critiques of early image generators like DALL-E 2. But history tells us these models can rapidly improve.
DALL-E 2, released two years ago, caused a similar buzz. The concerns then were: is it real art? A threat to artists? A disinformation machine? Two years later, let’s revisit these questions to inform our discussions about Sora.
DALL-E 2 wasn’t alone. Midjourney and Stable Diffusion followed, each with their strengths. Collectively, they democratized AI art creation. Just like with images, the impact of Sora might not be solely from itself, but from its imitators and competitors.
Many feared DALL-E and its kin would unleash a flood of deepfakes. While that threat remains, it hasn’t materialized yet. Deepfakes are expensive, and traditional propaganda techniques remain more common. It’s important to be skeptical of claims that Sora will be the turning point.
Instead of focusing on Sora’s current limitations, it’s crucial to remember how image generation models like DALL-E 2 and Midjourney rapidly improved. Early versions struggled with compositionality and basic physics. But both models and user prompting techniques have evolved, leading to more complex and detailed images.
DALL-E 3, for example, addressed many of these issues, and today’s image generators produce impressive crowd scenes and human hands. AI image generation has grown from a party trick to a major industry in just two years.
Therefore, when considering Sora’s future, remember: we’re witnessing early steps in a powerful new technology. It has the potential for good or harm, but overselling it or underestimating its potential are both easy mistakes.
Instead of clinging to specific predictions, let’s embrace the uncertainty. It’s safer to say this technology will improve rapidly than to guess exactly how. The key is to be vigilant, adaptable, and open to the possibilities, both good and bad, that Sora and its successors may bring.