Apple has made recent claims that it may have used caption data from YouTube videos to power one of its synthetic intelligence (AI) tools, but says the tool is not used in Apple Intelligence.
Apple showed a report from Proofnews that used this YouTube knowledge to exercise one of its models. The company explained that it did this to take advantage of open source OpenELM versions released earlier this year. The data included in a larger collection is controlled through the nonprofit EleutherAI corporation that supports AI research.
However, Apple told 9to5Mac that models trained with this data do not force any of its own artificial intelligence or device learning tools, Apple Intelligence added. It was a study task originally created through Apple’s AI groups and then shared, added through the company’s own device learning studies. place.
Importantly, this shows how much Apple should be noticed to deliver on its promise that Apple Intelligence models are trained with authoritative data.
But this is not the full picture. As mentioned earlier in the week, Apple Intelligence also trains its models “publicly to gather insights through our internet crawler. ” This admission reflects the extent to which generation corporations use data published online to create new AI products from which they profit.
The challenge is that by turning other people’s artwork into data and then profiting from that data, tech corporations are playing fair.
Speaking to Proofnews, Dave Farina, host of “Professor Dave Explains,” put it this way: “If you take credit for the paintings I’ve made [to create a product], that’s going to leave me jobless or out of work. ” . If other people like me being out of a job, then there needs to be a verbal exchange about reimbursement or some kind of regulation.
To some extent, knowledge on YouTube distracts from this critical argument: that the generative AI (genAI) machines commonly used today were likely trained with data created by humans and shared online. This is the type of data that is recovered through webcrawlers, including that of Apple.
But the quality of knowledge is an important factor here, and localizing knowledge inherently means that knowledge resources are the strongest fuel to drive educational AI.
Let’s consider two of the demanding situations AI researchers face.
This means that in their search for information, AI corporations inevitably seek out sources of knowledge. When this translates into activities retrieved from the open public web, this in itself implies that the creatives who have lately been fighting tech corporations for reimbursement for the use of their hardware in educational AI systems are right.
Because the most productive and up-to-date information they create is worth something, both to the creators, who consume it, but also to the other people who own and use the machines that derive their knowledge from it. Given that AI, by nature, becomes a tool directly available to everyone and in all supported languages, it is believable to think that the price of this data could increase once it is used to exercise an AI model.
So even if Apple doesn’t use YouTube knowledge for its Apple Intelligence models, it will use other knowledge found on the public web. And while Apple can at least try to avoid information that it shouldn’t do in this way (and it’s fair enough to have responded to the ongoing controversy over YouTube), not all AI companies do the same. And once the device is trained, it cannot be untrained.
Stay with me on Mastodon or sign up for the AppleHolic Bar & Grill groups and Apple Discussions on MeWe.
Hello and thank you for preventing. Nice to meet you. My name is Jonny Evans and I’ve been writing (mostly about Apple) since 1999. Today I write my AppleHolic blog on Computerworld. com, where I explore Apple’s developing corporate identity. You can also follow my work on AppleMust and follow me on Mastodon, LinkedIn, and (maybe) Twitter.
This study is written from an article about the article beyond the article on podigee. com. If you are interested, you can do it with your protection.
External content of podigee. com anzeigen
By clicking on “External Inhalte von podigee. com anzeigen”, I will make it transparent that you will see the Inhalt angezeigt wird. Dadurch können personenbezogene Daten an podigee. com et autres Drittanbieter übermittelt werden. You can find more information in our protection data and in https://www. podigee. com/en/about/privacy.
This article is written with the aim of writing an article from external articles published in reddit. com. If you are interested, you can do so with your protection.
External content from reddit. com anzeigen
With one click on “External Inhalte von reddit. com anzeigen”, I will show you how you can see the Inhalt angezeigt wird. Dadurch können personenbezogene Daten an reddit. com andere Drittanbieter übermittelt werden. You can find more information in our protection data and at https://support. google. com/reddit/answer/2801895?hl=de.