Some generative AI companies have been accused of scraping training data from the web containing copyrighted material without the consent of the creators or rights holders. As such, there can be a lot of creative and professional human work hidden behind the content produced.
(These issues are explored in three recent reports on IP at stake, focusing on the metaverse, visual arts, and music.)
On the other hand, “AI” (often those that actually generate derivative work) simulate intelligence by reusing the output of a skilled human. Some might think this is just a trick of self-confidence, despite other AI’s unmatched transformative ability to identify patterns in data. The latter can lead to real advances in science, medicine, healthcare, new materials, sustainability, climate technology, and more. Let’s hope so.
Unfortunately, there are a handful of more cynical companies scraping the web and waiting for lawsuits to bring the technology to the masses. Not only in the world of writing and music, but in other fields as well. In creative areas such as photography and video, many of us have unwittingly added training data by sharing content online.
But this generational mistake presents an opportunity for rival vendors to do a better job, highlight their ethical credentials and business models, and use AI to improve human well-being. their talent.
So can they achieve such a noble purpose? The jury is out yet.
One of the challenges is that ethical behavior is sometimes reflected in the eye of the beholder. As explored in another recent report, Dutch-based startup LaLaLand.ai’s sincere claim is that photorealistic AI-generated fashion models (trained using real online photos as a source) diversity), but certainly offset by employment opportunities. As a result, I was rejected by real-life plus-size, black, and Asian models.
If LaLaLand’s clients include wealthy global fashion brands such as Levi Strauss, Calvin Klein, and Tommy Hilfiger, who can afford to pay for physical models, who benefits from using virtual models? First, AI companies by taking up the income potential of those people. And second, mega-brands by saving money.
But what about ethnic minorities and plus-size citizens? and billboards staring at faces like ours – good thing. But they are not real. On the one hand, for these brands, the opportunity to work in the industry itself is closed. Hence the simulation of diversity. And definitely about ethics too.
Put another way, the vast majority of programmers are white men (studies show that STEM careers typically have 85-87% male employees and 91% white employees). These workers now receive money that would otherwise go to black and Asian models, mostly women. Is it Ethical AI?
Other players in the visual arts appear to be making more deliberate efforts to deploy AI ethically. For example, NVIDIA’s recent responsible AI partnership with Getty Images is part of a response to what Getty sees as a massive piracy of Stability AI (Stable Diffusion) in web scraping. It also attempts to inject AI into existing image search services to ensure rights owners get paid.
(Experimenters who have been using Stable Diffusion for a long time (I am one of them) will be accustomed to having what appears to be a watermark in the generated images. there is.)
And NVIDIA partnered startup Bria offers generative AI based on a library of licensed images. Again, the goal is to ensure that copyright owners are compensated. The challenge here is technical. In other words, how to output the best, most competitive images from systems trained on data sets much smaller than the web.
Meanwhile, software giant Adobe, which sees generative AI as a threat to its established image processing business, is gradually introducing AI across its Creative Cloud and is committed to using AI responsibly in its licensed content. I am aiming for
At first glance, all admirable efforts.
But what about AI in video (and Adobe has a presence, of course)?
This is the creative hotspot that rock star Peter Gabriel (always an early adopter of new technology) sought to encourage this month. He held a Stability AI video contest promoting songs from his upcoming album ‘i/o’ (input/output).
Gabriel certainly got more ideas than he intended. was forced to issue a statement clarifying his position on April 21.
In it, he said he was “plagued by negative reactions” to competition, explaining that it stemmed from a simple desire to be playful and creative with new AI tools.
He has reiterated his passionate commitment to artists’ rights and human rights in general (few can deny his track record on both counts).
AI is a product of our species, and we need to find ways to incorporate the ethics, compassion, and wisdom we hold so dear to our algorithms directly in order to protect and defend what is important to us.
Added my name to letters from Max Tegmark, Steve Wozniak, Elon Musk, and others. Pause new AI releases for 6 months and consider what to do. But if you don’t use this time to play and learn with what you’ve already built, how do you make sense of it?
fair point.
“New man, non-existent man”
Also in this space, Victor Riparbelli is co-founder and CEO of Synthesia, a UK-based AI start-up that generates video from text.
Its purpose is to bring boring documents to life by turning them into more engaging short movies. The service, he claims, is already used by his 35% of the Fortune 1,000. If true, that’s ample evidence of the rush to adopt generative AI among major companies.
At the recent Westminster Legal and Policy Forum, he explained the risks to intellectual property from AI, the metaverse and tokenization:
We live in a world where people want to see or hear that content. They don’t want to read anymore.
Perhaps not the triumph of humanity he thought when he said it. But to be fair to Riparbelli, he was primarily referring to technical manuals and his onboarding manuals rather than books and other texts in general.
Synthesia has raised $67 million in venture capital in just six years. And from day one, the CEO has viewed the company as a fully ethical company. He said:
Everything we do today is 100% fully agreed upon. We only work with fully consenting actors every day. [in which they appear].
On Data Mining and Data Analysis Problems […], the way we’ve thought about this question from the beginning is that we always want to have a dataset with 100% consensus. That’s how we built our business.
[By contrast] We know that many new text-to-speech systems are training by downloading 500,000 hours of audio from the Internet. [products].
At first glance, it is one of respectable and ethical businesses. One that honors the need for both human performers and voice actors on camera to make a living, as opposed to Internet scrapers.In fact, it creates new opportunities and markets for them.
However, Riparveri said:
Perhaps within the next 12-18 months, we will be able to create new people who don’t exist. That would certainly be an interesting part of the product.
It also opens up interesting questions such as: If I create a virtual person, can I own the likeness of that person? How to own the IP? What if one of our competitors tweets an exact replica of a non-existent person and uses that person’s likeness?
really what? However, the implication of these statements suggests travel direction, and perhaps isn’t as person-friendly as it first appeared.
If 100% of the data collected is consent-based, not collected from the internet, and the company pays real actors every time they appear in a video, then those experts are also the next generation. It sounds like it could be the source training data. “synthespians” – Virtual performers that may be wholly owned by the company.
Of course that’s an assumption. But AI-generated performers (effectively composites of real people) will likely be cheaper to hire than the human actors who trained them (such as LaLaLand’s models). And who knows how many companies will employ these synthspians in the future instead of real humans.
If true, the result – again – may be the potential income of struggling, diverse, and creative humans being usurped by heavily invested AI companies. is to help our Fortune 1,000 customers save money (because they have the right to do so).
Creative careers like most musicians, photographers, painters, illustrators, designers, writers, filmmakers, etc. already require risk, commitment, dedication, skill and talent. Successful, or viable and sustainable income.
Today, thanks to tools that claim to be AI, but are actually creative human outputs that generate trained derivative work, companies can just press a button.
Remember when tech industry CEOs said Industry 4.0 innovations would free them from tedious tasks and let them focus on their creativity? Well, bow down, humans. Studying AI awaits in the wings of your creative career. And it knows every part of you.
Certainly part of it is you.
my view
Welcome to Miller’s Hall, which increasingly hosts ethical debates about generative AI.
Is it your face staring at you from the screen? And did you deliberately agree that it was there? OK. Think again. Is it the AI output that looks a bit like yours and is the one that stole your career?
