Apple redefined Artificial Intelligence (AI) to mean Apple Intelligence. One would argue that it is funny to rename AI that way since a big piece of Apple’s announced AI-launched-strategy is OpenAI and ChatGPT. When talking about all of these, Apple’s CEO emphasized that the company is more focused on protecting the security and privacy of data. However, we can call this a narrow needle since we all know that security and privacy go hand in hand with the data used by AI. So if they are exposing such large amounts of data to their AI, then it’s double speak to say that they are protecting the security and privacy of the same data.
Generative AI has been trained to surf the web to scrape content. Since Apple allows opting out of its scraping, a recently published report indicates that some of the biggest sites have opted out of the Apple Intelligence training. Some of the sites that have opted out include Instagram, Facebook as well as many other high-profile like The Atlantic and The New York Times.
What is Apple’s AI Training?
Large language models like ChatGPT are usually given access to millions of words of source material, and ranges from user comments through news to stories. Over the past years, Apple used Applebot to train Siri in order to get spotlight suggestions. The company recently started using Applebot to train Apple Intelligence.
Some people have argued that the move is controversial because AIs are trained to use copyrighted material to give their own version of it. When a specific topic is surfed where the source material is scarce, it have been found out that AI regurgitates all paragraphs with almost no changes at all.
Is It Ethical?
The fact that Apple allows publishers to opt out at will, speaks volume according to their CEO Tim Cook. They screen out personal data and by doing it this way, they are ethical, at least according to him.
There are two sides to this Apple Intelligence. There is what they say and what they do.
What they say
The company says “We train our foundation models on licensed data, including data selected to enhance specific features as well as publicly available data collected on our web crawler, Applebot. Web publishers have the option to opt out of the use of their web content for Apple Intelligence training with a data usage control. We apply filters to remove personally identifiable information like social security and credit card numbers that are publicly available on the internet”.
What they Actually do
What they do, according to some, may contradict what they say (being ethical). You could look at what they do in terms of AI, which again comes in two different parts.
First, Apple uses its own grown AI tech to enable users tackle various tasks on their iPhones and Macs. All of these tasks performed on these devices do not require the outside assistance of OpenAI or any other form of generative AI. Apple AI is an opt-in by default, which means that users must agree to their data being available on Apple’s AI either directly or by leveraging the power of its own private clout for purposes of tackling more complex tasks.
The company’s Machine Learning Research Site states that the AI models train based on both licensed data and publicly available data collected by AppleBot (their web crawler). The term “publicly available data” is always used to mean un-copyrighted works, including personal data, which in turn begs the question whether Apple Intelligence data is safe and ‘ethically sourced”. This (idea of ethically sourced AI) goes beyond copyright and privacy legalities. Considerations of respect for individuals, their creative works and their rights to commercialize their works also arise.
The second part of what Apple AI does is that it hands over tasks it can’t perform to ChatGPT and OpenAI to do the work when users give permission to do so. In short, if Apple’s application of “publicly available data” means what we all know it does, then there are more than enough reasons to believe that Apple also faces potentially the same legal battle that OpenAI and other big Tech companies face.
It is also in public knowledge that Apple paid some of the companies (social media companies) in return for the use of their content for training. It is believed that this might be the motivation behind some of these companies opting out, probably holding out for what they call “partnership agreement”. Competing using “publicly sourced data” which according to some, is not ethical, will not make Apple an ethical competitor.
great piece