🇧🇷 First-class complaint against an AI raises the issue of copyright
Powered by billions of images collected from the web, generative AIs like DALL-E 2, Midjourney or Imagen are ready to revolutionize the graphics industry. Unless the outcome of a trial decides otherwise.
Emily Turrettini – journalist specializing in new technologies
Before getting into the process, let me digress for a moment to reflect on the euphoria of recent months over generative AI, to put the issues in perspective.
Despite the emotion generated by these stunning visuals, it’s the writing software, powered by GPT-3 (the largest model ever trained with 175 billion parameters), which will have the biggest impact on all sectors of the economy.
The ability to generate language – written and spoken – will be far more transformative than the ability to produce visuals. Every line of business, every business, every business transaction depends on this form of communication.
the first real killer app Large Language Models (LLM), in terms of commercial adoption, have proven to be text writing. They already feed websites, press articles, marketing or advertising campaigns and academic work🇧🇷
So AI-assisted writing has exploded this year. Jasperone of the biggest players in this space, launched just 18 months ago, is expected to hit $75 million in revenue by 2022. Making it one of the fastest growing software startups of all time.
But one of the most promising business applications involves not natural language but computer language, ubiquitous and indispensable in the modern world.
In a recent study, Google found that employees who used their AI autocomplete tool reduced their coding time by 6%. And recent data from the GitHub hosting platform is even more compelling: Using Microsoft’s latest tool, Copilot, a GPT-3 AI model, can reduce the time it takes a software engineer to spot a spot.
Legal action against an AI for copyright abuse
But a few days ago, a collective of programmers presented a legal action against GitHub, its parent company Microsoft and technology partner OpenAI, for allegedly violating its open source licenses without proper attribution. They accuse them of using their work to train the programming assistance software, Copilot.
The latter is an artificial intelligence capable of generating lines of code and proposing entire functions or complex algorithms.
This is the first major process involving AI. Potential damage could exceed $9 billion.
The legal aspect of using AIs is now on very worrying ground. DALL E-2, Stable Diffusion and Jasper AI training is also done using billions of data collected from the web, not all of which are copyright free.
The companies involved assumed that the exploitation of big data to train AI is covered in the United States by the doctrine of fair use (fair use). And it is on this point that justice will have to decide – with consequences for the future development of artificial intelligence.
Sources: ICT Journal 🇧🇷 the computer world 🇧🇷 new scientist 🇧🇷 On the edge 🇧🇷 TechDirt 🇧🇷 vice 🇧🇷 Forbes
Did you find an error?Please inform us.