See how Siri would work if it had the intelligence of ChatGPT

See how Siri would work if it had the intelligence of ChatGPT


In the video, developer Mate Marschalko managed to integrate Apple’s Siri with GPT-3 and test how the commands would work in everyday life

A developer managed to make Siri work with the power of ChatGPT, the artificial intelligence that has taken the internet by storm lately. In the video, programmer Mate Marschalko demonstrated how the virtual assistant from Apple works when powered with GPT-3 technology.




See how Siri would work if it had the intelligence of ChatGPT

Marschalko’s exhibition shows a virtual assistant capable of interpreting much more natural (and therefore more complex) speech interactions. The developer is able to communicate without many ties to traditional commands, such as “turn on the light”, “set a reminder”, “recommend movies”, and can only chat with Siri.

 

The answers from the Apple assistant are also just as interesting and natural. The assistant is able to interpret the user’s will from deductions, such as in a moment when Marschalko says “I noticed now that I am recording this video in the dark and in the office, can you do something about it?”, turning on the lights in the living room ready.

The entire process of integrating Siri with GPT-3 was described on Marschalko’s personal blog. The developer has used the capabilities of artificial intelligence to develop the principle-based integration of a virtual assistant capable of managing a smart home (interactions, response format and other properties), and put this structure into practice.

The project, however, is clearly a prototype with its limitations. As he himself points out, each GPT-3 API call (commands, in this case) costs $0.014, which can get pretty expensive if you’re using AI-powered Siri on a daily basis. Editing the video could also have hidden some bot glitches, response delays, or context-free interactions.

How should virtual assistants work?

Marshhalko’s idea was hardly new, but it raises important questions about using a virtual assistant. Do you prefer an assistant that can interpret speech in detail or an engine that accepts direct and succinct voice commands?

Each of the formats has advantages and disadvantages, such as the user’s comfort in knowing that there is a super intelligent AI engine inside their home or being limited to the most basic commands.

In any case, the flexibility introduced by AI makes the virtual assistant much more interesting for everyday use. For example, the user would not need to stick to certain keywords to trigger specific commands or, when needed, could use the wizard’s search capabilities to get comprehensive answers on complex topics.

Domestic AIs have their dangers

A virtual assistant works with a limited list of actions based on user input. Such a limitation on the number of possible interactions has its value: the program is intended to help with daily life and household management, nothing more.

An AI like GPT is not limited in this way and the multitude of possible responses can interfere with the assistant’s purpose or have unfortunate consequences: although it has blocking mechanisms, an AI like GPT can interact in a racist, misogynistic or provide detailed information about “how to build a bomb”, for example, even indirectly.

Trending on Canaltech:

+The best content in your email for free. Choose your favorite Terra newsletter. Click here!

Source: Terra

You may also like