Paul Stenhouse: OpenAI's latest tool can complete tasks for you such as booking flights or restaurants

OpenAI's latest tool can complete tasks for you.
  
The biggest thing is that the operator can take actions for you using its 'computer vision'. Basically, it takes screenshots of the webpage, then can click with its virtual mouse and keyboard to complete things, like book a flight, or a restaurant. It's a big step forward because traditionally to do these things you needed to use APIs - structured data messages with explicit details. Now you can write a sentence and the AI will just do it. No preparation needed.  
  
Excited? Well, it's available now, but only for those with the $200-a-month ChatGPT Pro service. 

LISTEN ABOVE 

See omnystudio.com/listener for privacy information.

Episode Transcript

Speaker 1 (00:07):
You're listening to the Saturday Morning with Jack Tame podcast
from News Talks at me.

Speaker 2 (00:12):
Open Ai has just released a new tool. This is
the company, of course, behind chat GPT, and this tool
can go one step further our textbit Paul Steenhouse is
here with the details. Got to Paul, what can it do?

Speaker 3 (00:25):
Yeah, yeah, I feel like we're getting one step closer
to Rosy the robot from the Jetsons. Although I guess
this thing can't move around. Maybe that's the next step
it'll take. Because this is Rosy the Robot effectively for
your computer and your digital version of that assistant. But yeah,
it's called Operator Jack and basically it's going to be

(00:45):
able to start doing things for you on the Internet.
So let's imagine you want to book a flight. You
can simply type in and this is the cool part.
You just use natural language. You say, I want to
book a flight to Hawaii, and I want to do
three nights, and I don't want it to be rainy season,
and I want to do this, this, and this activity.

(01:06):
It will then just go away and start browsing the
Internet just like you or I would, And it uses screenshots.
They call it computer vision, but really it's screenshots of
the web page, It analyzes those, figures out what fields
it needs to fill in, figures out what buttons it
needs to click with its virtual mouse, and what to
type with its virtual keyboard, and actually just starts doing

(01:28):
things for you. So what's really interesting about this though,
is we have kind of like assistance and things, and
they can do things for us, but typically they all
need to be pre programmed, right, because they need to
use what we call in the digital world APIs effectively
structured data messages that different services send each other, and
we will be able to be able to do things.

(01:50):
This is interesting though, because it actually just like a human.
It basically looks at the screen, figures it out, it
doesn't need any pre programming, and it can just start
doing things. So if a new restaurant popped up, it
could theoretically go to its website and make a booking
for you without ever having seen that website before. And
so it's a big step forward because we're actually going
to start getting towards some assistance that are smart and

(02:12):
might actually be able to do some things for us.

Speaker 2 (02:15):
Yeah, so I mean how.

Speaker 3 (02:16):
The price deck? Oh yeah, go on, Okay, So it's
part of the chet GPT's pro plan, which is just
two hundred dollars two hundred US dollars a month. They
do say that they're going to start rolling it out
to the other plans. Do you know Sam Oltman, the
CEO of chap of open Ai, he said that he
chose that price and he thought they'd make money off

(02:38):
that price. Turns out they're not making any money off
that price because people are using it so heavily this
chet GPT pro feature that it's it's making a loss.
Oh really, the next one's probably going to be Yeah.

Speaker 2 (02:50):
So it's not that they've priced people out of the market.
It's just that it's actually being used so much and
it's using so much computing power that they're Yeah. So
my question though, is like, honestly, how usable is this?
Like how much stuff do you really need to be
done by? How many times are you booking a flight
to Hawaii and needing that kind of research downe you

(03:11):
know what I mean. It's kind of like with voice
assistance and Siri and stuff, it's like, well, how often
do you really need to time a set? How often
do you really need to know the temperature?

Speaker 1 (03:19):
You know?

Speaker 3 (03:20):
Well, this one's different because it can do basically anything.
Right now, it still is in its research beta phase,
so they're going to say there will be hiccups, but
we're well on our way to it being able to
craft a Facebook post for you or browsing Facebook and
telling you what might be interesting, or you know, you
could probably even send it if you were gaming your
bank details and tell it to pay your bills and

(03:44):
do things like. That's where it's headed, right, Like you
could actually say pay my contact energy bill and it
could go away and start to figure out how to
do that. So we're not there yet, but that's what
they wanted to be able to do.

Speaker 1 (03:55):
Yeah.

Speaker 2 (03:55):
Yeah, Oh it sounds amazing. Okay, thank you so much. Paul,
sounds amazing if expensive. Paul Stenhouse our texpert.

Speaker 1 (04:02):
There for more from Saturday Morning with Jack Team. Listen
live to News Talks ed B from nine am Saturday,
or follow the podcast on iHeartRadio