With Generative AI and ChatGPT, which appear to be omnipresent in tech these days, the hype mill is in full swing. It's hardly surprising, therefore, that there's talk of a new, better Siri. Indeed, 9to5Mac has discovered a new natural language system.
Do you understand what I'm saying?
Siri on tvOS 16.4 beta is said to include a new "Siri Natural Language Generation" foundation. As explained, it doesn't seem particularly amazing, since it appears to be mostly focused on cracking (dad?) jokes, but it may also allow you to set timers using natural language. The codename is "Bobcat."
These rumors are in response to a recent New York Times piece on Apple's February AI meeting. According to the report, the event focused on the type of generative material and large language models (LLM) employed by ChatGPT. It further stated that Apple's engineers are "actively testing" language-generation concepts by throwing new language concepts around every week as part of Apple's efforts to advance AI.
[How AI can help businesses improve their DEI]
So, is it constructing a ChatGPT competitor? Bloomberg claims that this is not the case.
"Hey, Siri, how do you spell "catch-up"? ”
While Siri looked to be extremely intelligent when it initially debuted, progress has lagged, giving Apple's witty voice assistant memories of MobileMe and Ping. Siri, like both Apple failures, made promises it never kept and currently trails behind Google and Amazon assistants, despite being a bit more private.
Siri's lack of contextual understanding implies that it is only competent at what it has been trained to do, limiting its powers; GPT appears to leave it in the dust. OpenAI is developing quickly, as seen by the current GPT-4 version. We can see that this has sparked a fire under the giant tech companies. Microsoft has incorporated ChatGPT into Bing, Google is working hard on Palm development, and Amazon is pushing hard on AWS Chat (the latter is now integrated within Microsoft Teams).
Of course, Siri isn't the only machine intelligence (MI) project Apple is working on. It has produced extremely fantastic instances of MI done well in various sectors, such as accessibility and picture augmentation. Siri, though, continues to make errors.
I'm not sure how Apple's Steve Jobs would have handled it - I can't imagine him being pleased when his HomePod informs him it can't find his Dylan tunes. The difference between the two voice-capable AIs is that I could ask GPT to draw a picture of him hurling the smart speaker against the wall.
This is due, in part, to the way Siri was designed.
Siri is a massive collection of answers for many knowledge disciplines, complemented with Spotlight search results and natural language interpretation so you may speak to it. When a request is made, Siri checks to ensure that it understands the inquiry before employing deep/machine learning methods to determine the right response. To obtain that response, it computes a numerical estimate (confidence score) of the likelihood that it has the correct answer.
This means that when you ask Siri a question, it first looks to see if it is a basic request ("turn on the lights") that it can complete quickly from what it already knows, or if it needs to search the broader database. Next, it performs what you want it to do (sometimes), provides you the data you need (often), tells you it doesn't understand you or requests that you alter a setting buried someplace on your system (too often).
In principle, Siri is only as good as its database, so the more answers it has, the better and more successful it gets.
However, there’s a problem. As explained by former Apple engineer John Burkey, the way Siri is built means engineers must rebuild the entire database to upgrade it. That’s a process that can take up to six weeks.
This lack of real learning makes Siri and other voice assistants “dumb as a rock,” according to Microsoft CEO Satya Nadella. You’d expect him to say something like that, of course, as Microsoft has billions invested in ChatGPT, which it is weaving inside its products.
Generative AI (the type of intelligence utilized in ChatGPT, Midjourney, Dall-E, and Stable Diffusion) likewise employs natural language, its databases, and search results, but it may also use algorithms to produce original-looking material such as audio, pictures, or text.
You may ask it a question, and it will go through all available facts and make a few judgments to get an answer.
Now, as has been noticed regularly since individuals began investigating the technology, the outcomes aren't necessarily fantastic or novel, but they typically appear believable. The ability to instruct it to make deep fake films and photographs push this even further.
While Siri may allow you to ask for a map of Lisbon, Portugal, or even source directions to a location on that map, Generative AI allows you to ask more nuanced questions, such as what parts of the city it recommends, to write a story with the action set in that city, or even to create a spookily accurate fake photo of you sitting in that really lovely bar in Largo dos Trigueiros.
It's quite evident which AI is the most amazing.
This does not have to be the case. Apps for adding ChatGPT to Apple's products have been created by developers. A good example is watchGPT, which was recently renamed Petey - AI Assistant for trademark considerations.
Apple is unlikely to hand up such a competitively crucial technology to a third party, so it will likely continue working on its solution, which may take years – during which time Siri may still fail to unlock the cabin door.
But, considering that GPT-4 can cost up to 12 cents per thousand prompts, Apple is unlikely to incorporate it directly into its operating systems. It would be very costly to do so with a user base of over a billion, and Microsoft is already there.
In that scenario, Apple may just bite the bullet and make it simple for its developers to include support for OpenAI's technology in their applications, thereby passing the expense on to them and their users.
It may assist in the near run, but I'm confident Apple's machine intelligence teams are on fire. They will be much more driven to innovate in the natural language processing that is at the heart of both technologies.
Yet, they appear to have slipped behind in terms of execution at this point. Nevertheless, as GPT-generated photos demonstrate, looks may be deceiving.
As a leading independent research provider, TradeAlgo keeps you connected from anywhere.