• News
    • Tech News
    • AI
  • Gadgets
    • Apple
    • iPhone
  • Gaming
    • Playstation
    • Xbox
  • Science
    • News
    • Space
  • Streaming
    • Netflix
  • Vehicles
    • Car News
  • Social Media
    • WhatsApp
    • YouTube
  • Advertise
  • Terms
  • Privacy & Cookies
  • LADbible Group
  • LADbible
  • UNILAD
  • SPORTbible
  • GAMINGbible
  • Tyla
  • FOODbible
  • License Our Content
  • About Us & Contact
  • Jobs
  • Latest
  • Topics A-Z
  • Authors
Facebook
Instagram
X
TikTok
Snapchat
WhatsApp
Submit Your Content
Expert explains how to spot AI-generated videos

Home> News> AI

Published 09:40 5 Mar 2024 GMT

Expert explains how to spot AI-generated videos

The errors defy the laws of physics.

Rebekah Jordan

Rebekah Jordan

Whilst chatbots are making it easier for students to write essays, AI video generators can create videos from single prompts - but they're far from perfect.

OpenAI recently launched Sora, an AI video generator.

With the new text-to-video tool, users can type in a single prompt and their words will be brought to life in hyper-realistic or animated scenes.

But take a closer look and you'll start noticing the flaws.

Advert

Co-founder of AI sales company, Collectivei, Stephen Messer, The Wall Street Journal shared some hints on how to spot AI-generated videos.

The first thing the narrator points out is the disappearance of things. One video shows an AI-generated 'cooking grandmother' who has a spoon one moment that's gone the next, and then back again.

Stephen Messer analyses the flaws in AI-generated videos / The Wall Street Journal
Stephen Messer analyses the flaws in AI-generated videos / The Wall Street Journal

A similar thing happens in another video showing a cat waking someone up. The cat's paws seem to move from one part of the cat's body to another in a way that doesn't make sense.

Advert

Messer, with more than a decade of experience in the AI industry, explains that it all comes down to the 'physics of the real world'. There are still things AI doesn't understand.

'Sometimes things may not feel quite right [...] So our senses are amazing at spotting weird things, things that just don't feel right,' he explained.

One video of a running man may look normal at first, but at a closer look, the way his arms are swinging in motion with his legs looks off. Messer describes the way his arms are doing a 'double take' which means his balance wouldn't work and the running motion wouldn't happen.

Additionally, simple aerial landscape shots might look like drone footage, but Messer points out the 'simple physics problem.' Things like waves moving in the wrong direction or cars driving backwards are things to look out for that obviously don't occur in our natural world.

Advert

If a user asks for a certain object, sometimes AI just throws it in in a way that doesn't make any sense, like staircases leading to nowhere in a landscape.

AI-generated videos sometimes defy the physics of our world / The Wall Street Journal
AI-generated videos sometimes defy the physics of our world / The Wall Street Journal

Historical footage looks seemingly realistic when it reflects the grainy texture of an old film camera, but a second look shows houses that are from 'all different generations' that obviously don't match the time period it's supposed to be filmed in.

It gets more difficult to detect faults when it comes to animated videos, however. Messer explains that with animations, 'you don't expect [them] to be perfect.

Advert

'In fact, part of the fun sometimes is that they do things that are physically impossible to do.'

But it is impressive that a user can create a single idea that could soon reflect the professionalism depicted in a full-blown Pixar movie.

However, the ease of creation has caused some legal trouble.

Comparing the story similar to that of Google, Messer states that when it comes to a company 'making money off the backs of other people's work' it's naturally going to 'lead to Copyright lawsuits'.

Featured Image Credit: The Wall Street Journal
AI
Science

Advert

Advert

Advert

Choose your content:

3 hours ago
5 hours ago
7 hours ago
  • 3 hours ago

    Sam Altman could be hitting Google with the biggest threat it's faced in a decade

    OpenAI has revealed a new major project

    News
  • 5 hours ago

    Users predict an 'apocalypse for websites' amid fears Google is about to 'destroy the web'

    The internet as we know it could be gone soon

    News
  • 5 hours ago

    McDonald's AI hiring bot exposes data of 64,000,000 applicants after hackers guess 'stupid' password

    The fast food giant isn't lovin' it

    News
  • 7 hours ago

    ChatGPT CEO Sam Altman shares eerie reality of what actually happens when you search with AI

    Some people are making sure they are polite to their AI chatbot

    News
  • AI company overtakes Apple to become world's first $4,000,000,000,000 company
  • Distressing signs of AI threatening its users highlights 'a sobering reality'
  • Woman scammed out of $850,000 after talking to AI-generated Brad Pitt for months
  • AI company confirm they're on the verge of a monumental breakthrough in mission to 'solve all diseases'