According to the report, the “Siri Natural Language Generation” framework was discovered in the tvOS 16.4 beta. It has been known to activate when trading jokes with Siri on an Apple TV with the tvOs 16.4 beta installed. earlier than that New York Times article on the Apple AI Summitargued that the event had a focus on generative content and large-scale language models (LLMs). He also cited sources saying that “many Apple engineers” are actively testing “the concept of language creation by introducing new language concepts every week.”
So, is Apple creating a competitor to ChatGPT? According to Bloomberg, no.
“Hey Siri, how do you spell ‘catch-up’?”
When it first appeared, Siri seemed surprisingly sophisticated. However, since the pace of Siri development has not kept pace with AI development, there were concerns that Apple’s voice assistant might follow in the footsteps of Apple’s MobileMe and Ping, which had faded. in history. Like Apple’s previously mentioned failures, Siri fell short of expectations, and now it lags behind assistants from Google and Amazon.
Siri’s lack of context awareness means this voice assistant is only good for the things it’s been taught to do. GPT seems to outperform Siri by a wide margin. OpenAI is innovating fast and recently released an update to GPT-4. It has also ignited competition among big tech companies. Microsoft has applied ChatGPT to Bing, Google is accelerating the development of Palm, and Amazon is pushing AWS Chat heavily (AWS Chat was recently integrated into Microsoft Teams).
Apple and Siri still seem to have a long way to go.
Of course, Apple is working on Artificial Intelligence (AI)It’s not just Siri. In some areas, such as accessibility and image enhancement, MI has been well implemented. But anyway, Siri always makes mistakes. This is partly due to the way Siri is built.
How Siri Was Created
Siri is a kind of huge database of answers on various fields of knowledge. When Siri receives a user’s request, it checks to see if they understand the question, then uses deep learning/machine learning algorithms to find an appropriate answer. Then, the probability of getting the correct answer is numerically evaluated (reliability score).
In other words, when you ask Siri a question, Siri first looks quickly to see if it’s a simple request that can be answered quickly with information it already knows (for example, “turn on the lights”), or whether it should refer to a larger database. It then does what the user asked (sometimes), retrieves the data it needs (often), or tells the user it didn’t understand the user’s request, or changes a hidden parameter somewhere in the system (too often).
In theory, of course, Siri can provide better answers and be more efficient with more databases. But there is a problem. According to former Apple engineer John Bucky, The Siri build method means engineers have to rebuild the entire database to bring it up to speed. This process can take up to 6 weeks. The resulting lack of learning has made Siri and other voice assistants “stupid as a rock”, said Microsoft CEO Satya Nadella.
On the other hand, generative AI (ChatGTP, Midjourney, Dali, a type of intelligence used for stable broadcasting) uses natural language, its own database and search results, but leverages algorithms to create a content that resembles the original (such as audio, images or text). Additionally, when a user asks a question, they sift through all available data to make decisions and derive results.
As has often been noted since people started exploring this technology, these results are not always brilliant or original. But generally the crab looks convincing. The ability to request the creation of deepfake videos and photos pushes the technology further.
Here’s one way to see the difference between Siri and generative AI. Just think about what the two AI models can achieve. With Siri you can ask for a map of Lisbon, Portugal, or directions to somewhere on a map, but with generative AI you can ask more nuanced questions, such as what area of town it recommends , and in turn tell a story set in that area. It can create creepy fake photos of you writing or sitting in a fancy local bar.
It seems pretty clear which AI is awesome.
What will happen next?
Developers have created applications that add ChatGPT to Apple products. For recent brand reasons ‘Petey – AI Assistant’‘Watch GPT’, which has been renamed ‘Watch GPT’, is that.
Apple will likely continue to work on developing its own solutions, as it won’t want to hand over competitively important technology to third parties. But that could take years, and in the meantime, Siri could continue to fall behind. However GPT-4 costGiven that these prompts cost up to 12 cents per 1,000, it seems unlikely that Apple will introduce the technology directly into its operating system. Because with over a billion users worldwide, that would be prohibitively expensive.
In this context, Apple can effectively pass the costs on to developers and customers by allowing developers to easily add OpenAI technology to the applications they create. It might help in the short term, but I think Apple’s AI team caught fire. The artificial intelligence team will work hard to advance natural language processing.
At the current stage of implementation, Siri seems to be lagging behind. However, as you can see from the images created with GPT, looks can be deceiving.
editor@itworld.co.kr
