The hype machine becomes a reality with generative AI, Chat GPT, which seems to be everywhere in technology these days. So it’s no surprise that we’re starting to hear some chatter about the new and improved Siri. in fact, 9to5Mac You’ve already discovered a new natural language system.
do you speak my language
The tvOS 16.4 beta version of Siri claims to have a new “Siri Natural Language Generation” framework. As I explained, it’s not impressive because it seems to focus primarily on communicating (father?) Just kidding, you can also use natural language to set timers. The code name is “Bobcat”.
These Whispers Follow Recent Whispers new york times Apple’s February Report AI SummitAccording to that report, the event had some focus on the types of generated content and large-scale language models (LLMs) used in ChatGPT. He also said that Apple’s engineers are “actively testing” language generation concepts, trying out new language concepts every week, as Apple seeks to move his AI forward.
So are we building a ChatGPT competitor? According to bloomberg.
“Hey Siri, how do you spell ‘catch up’? ”
Siri seemed incredibly sophisticated, when it first appearedbecause development hasn’t kept up, Apple’s cheeky voice assistant is Mobile Me and pinLike both Apple failed, Siri promised to lag behind Google and Amazon’s Assistant, but even if it’s a little more private.
Siri’s lack of sense of context means it’s really only good at what it’s trained to do, which limits its capabilities. GPT seems to have left it in the dust. Recent GPT-4 updatesOpenAI is rapid innovationWe can already tell this has caught fire under big tech companies.Microsoft is ChatGPT adoption Inside Bing, Google is fast forwarding palm development and the Amazon Pushing AWS Chat Hard (The latter is now integrated into Microsoft Teams).
Apple and Siri appear to be in crisis.
not unique
Of course, Siri is not alone machine intelligence (MI) Apple is working on it. In some areas, such as accessibility and image enhancement, we have achieved very good examples of MI done right. But somehow Siri still makes mistakes.
I’m not entirely sure how Apple’s Steve Jobs handled it — I don’t see how he’s happy when HomePod says it can’t find Dylan’s tracks. The difference between the two voice-enabled AIs is that you can ask GPT to make a picture of Smart hurling his speaker at the wall.
Part of the reason is the way Siri is built.
How to make Siri
Siri is like a giant database of answers for different knowledge domains, complemented by Spotlight-sourced search results and natural language interpretation so you can talk to it. When a request is made, Siri verifies that it understands the question and uses deep/machine learning algorithms to identify the appropriate response. To get that response, we numerically assess the likelihood of having the correct answer (confidence score).
What this means is that when you ask Siri, it first asks if this is a simple request (“turn on the lights”), if it can be done quickly from what you already know, or if you need to consult it. It’s a quick way to see what’s going on. bigger database. Then do what you asked for (sometimes) to get the data you want (frequent(too often).
In theory, Siri is only as good as its database. In other words, the more answers Siri gets inside, the better and more effective it becomes.
However, there is a problem.As Former Apple Engineer John Barkey Explains, the way Siri is built means engineers would have to rebuild and upgrade the entire database. This is a process that can take up to 6 weeks.
According to Microsoft CEO Satya Nadella, this lack of real learning has made Siri and other voice assistants “stupid as rocks.” Of course, Microsoft invests billions in his ChatGPT, so you’d expect him to say something like that. woven into the product.
On the other hand, generative AI
Generative AI (the kind of intelligence used in ChatGPT, Midjourney, Dall-E, Stable Diffusion) also uses natural language, proprietary databases and search results, but leverages algorithms to create originals such as audio, images, etc. You can also create visible content. , or text.
When you ask a question, it examines all the available data to make some decisions and draw conclusions.
Now, as has been pointed out quite often since people started exploring this technology, their results are not necessarily great or originalbut they usually look convincing. Deepfake videos and photos Taking this further.
One way to see the difference between two AI models in use is to think about what they can accomplish.
So, with Siri, you might be able to ask for a map of Lisbon, Portugal, and enter directions to anywhere on that map, but with Generative AI, you can recommend which part of the city. You can ask more subtle questions, such as whether toYou can even create an action story based on that city, or an eerily accurate fake photo of you sitting in that really nice bar Largo dos Trigueiros.
It’s clear which AI is the most impressive.
what happens next?
You don’t have to do it this way.the developer is I was able to create an app to add ChatGPT to Apple’s productsThe recently renamed watchGPT Petey – AI Assistant For trademark reasons, is a good example.
It’s unlikely that Apple will want to hand over such competitively important technology to third parties, so it’s likely that it will continue to work towards its own solution, but this could take years. There is a nature. During that time, Siri may still be unable to open the cabin door.
however, GPT-4 costs ~12 cents per 1000 prompts, it is highly unlikely that Apple will weave it into their operating system in a direct sense. With over a billion users installed, it would cost a ton of money to do this, but Microsoft is already there.
In that context, Apple may simply bite the bullet for making it easier for developers to add support for OpenAI’s technology to their apps.
While this may help in the short term, I’m sure this will set Apple’s Machine Intelligence team on fire. They are twice as determined as he is now to further advance the innovations in natural language processing that are at the core of both technologies.
However, in terms of implementation, it seems to be lagging behind at this stage. However, as the GPT-generated image shows, the appearance can be deceptive.
follow me Mastodonor join Appleholic’s Bar & Grill and apple discussion MeWe’s group.
Copyright © 2023 IDG Communications, Inc.
