A web developer has created a Siri GPT-3 mashup to showcase the potential of a truly smart voice assistant.
The AI system was able to understand the meaning of spoken language, including one in which the required action was implied rather than directly stated …
Mate Marshalko wrote a blog on Medium in which he talks about what he did, as well as a YouTube video (below) with examples of the results. He said his goal was to effectively replace Siri with GPT-3.
Ever since I tried ChatGPT and GPT-3, everything else seems painfully stupid and useless: Siri, Alexa, Google Home, and all the other “smart” assistants. […]
I thought it would be great to replace Siri with GPT-3. I asked GPT-3 to pretend to be the smart brain of my house, explaining in detail what it can access around the house and how to respond to my requests.
More impressively, he didn’t have to write any code to achieve this. Instead, he explained to GPT-3 what he wanted to do. He asked him to interpret spoken English and provide an appropriate response in JSON format.
As you can see, I explained everything in simple language. I described the types of requests, the exact structure of the response, and asked him to behave like a sentient AI, giving advice even on personal matters. I also provided some details about the time, location, devices, and rooms in the house. From this we will get a perfectly structured answer. And that’s all there is to programming!
Marshalko used a Siri shortcut to turn this into a real HomeKit command.
He then tested it and said he couldn’t believe how well it worked.
Our example request was: “Just noticed I’m recording this video in the dark at the office.
Can you do something about it?” And we got the answer:{
“action”: “command”
“location”: “office”
“target”: “light”
“value”: “enabled”,
“comment”: “I turn on the light for you.”,
“ScheduleTimeStamp”: “”
}
He also relied on GPT-3 to sort out some other things.
Set the bedroom temperature to what you think will help me sleep better.
The AI then set the thermostat to 19°C. By the way, this sounds too high for sleep to me (I choose 16°C myself) and illustrates one potential danger of this approach: GPT AIs are not always good at getting the right facts and consistent advice.
Just yesterday, for example, CNET had to announce a review of all of its articles written about AI after a really stupid mistake was discovered.
Futurism went through one of the articles highlighted by Guglielmo in the post, namely the article titled “What is Compound Interest?” and found several serious errors. Although the article has since been corrected, the original version stated that “you will earn $10,300 at the end of the first year” – instead of $300 – if you deposit $10,000 into an account that earns 3 percent per annum. The AI also made mistakes in explaining loan interest payments and CDs or CDs.
However, this is an impressive demo that illustrates the potential of a truly smart version of Siri. Hopefully Apple has a team playing with the Siri GPT-3 combo or will create one after seeing this.
Across the Edge