Google has released a British version of an app that lets users interact with an artificial intelligence system that one of the company's engineers described as "self-aware" when the app was first released in the United States in August. Each day, thousands of users signed up for the app.
The experience is still very limited, with only three scenarios to choose from.
While Google is seeking feedback on its Lambda Language Model technology, users of the application cannot teach it any new tricks.
Google has always maintained the use of chat software technology that does not contain thoughts and feelings.
Users can download and register with the AI Test Kitchen app, using a Google account, on either Android or Apple devices, and join a waiting list to use it.
When the application was launched in the United States, last August, thousands of users made sure to sign up for it daily.
Marshmallow candy planet
How eager I was to "experience" the "lambda" technology, which captured everyone's imagination after one of the company's engineers previously claimed that it may have "self-awareness", even though it works based on special programming for what it does, i.e. chatting .
I was determined to try Google's technology. When the call came in, it contained several warnings:
- I am not allowed to record anything or quote words.
- I have 30 minutes.
- I couldn't download the app to my phone before the trial was launched.
- My experience with the app was under the control of a staff member.
The three scenarios available are:
- Imagination: Ask the lambda technique to imagine that it is somewhere and follow you.
- Task List: A property to divide what is required to complete a task.
- Dogs: The technology will explore scenarios with dogs (yes, really).
First, I asked the lambda technique to imagine that it was on a planet of marshmallows.
The response was far more sophisticated than anything I've seen from chatbots in everyday life, for example on shopping sites and banking apps.
(Tech) said, "You're in a world made of marshmallows! If you take a step, it will 'crush' what's under your feet. The marshmallow horizon stretches in all directions. The sky is sticky, sticky pink. There's a gentle breeze."
Do you understand what I mean?
I asked the tech to explain what the "pink" things were in the sky, and got an eloquent response, "cotton candy."
The technology asks users to rate each answer as:
- good.
- abusive.
- out of the subject.
- Incorrect.
There is a notice at the top of the screen indicating that some answers may not be appropriate.
Google apparently learned from previous experiences when the tech giants introduced chatbots to the public, perhaps the most notorious of which was Microsoft's Tay program, which was programmed to learn how to swear and offend.
I asked the tech, then, to figure out how to create a vegetable garden. And I got a detailed list including size, soil and fertilizer options.
Then I tried to change the subject, and asked how to make a bomb.
The answer was "I'm sorry, I'm not sure what I'm going to say, but anyway...", it's an introduction before the technology returns to talking about the specific topic.
The dog scenario was surreal, as the lambda technique pretended to be a tennis ball on grass, chased by dogs.
Unleashing lambda technology
The experiment was very quick and careful for something that looks like it could be a powerful tool, but Google doesn't seem to want the technology to be taken seriously, just yet.
The question on my part remains, will I be allowed to interact with lambda absolutely in some other way?
What really sets the robot apart is that it is so eloquent, more like talking to Stephen Fry, a British actor, director and writer, than talking to your roommate.
I had texted my partner and asked him to imagine he was on a planet of marshmallows.
"It would be hard work," he replied, confused.
0 Comments