Google is testing a new AI chat bot — we’re putting it through its paces | science and technology news

The scene is at 10 Downing Street, the home of the Prime Minister.

It’s a refreshing cold day. The lawn mower can be heard from a distance.

There is a knock on the door, and a policeman answered him!

Now, before anyone gets any ideas, the setting for this particular story is the work of artificial intelligence – The GoogleA chatbot called LaMDA, to be exact, made headlines this summer When a former engineer now claims he was conscious.

Since then, the tech giant has started running a very limited trial to put it in its stride.

Sky News gained access to the testing stage this week, as Google employees took to the stage in New York to give an update on their work with artificial intelligence.

The show included how they were exploring whether LaMDA could create video clips, realistic speech, and even write fiction.

More about artificial intelligence

The experiment that’s slowly rolling out to anyone who doesn’t mind joining the queue is called the AI ​​Test Kitchen, which offers three scenarios — Imagine It, Talk About It, and List It — to play with.

The scenarios are incredibly specific but still caution that the answers may be incorrect or inappropriate, seemingly to avoid pitfalls Previous chatbot experiences that ended up being quite abusive.

There are no such problems with LaMDA, which at its worst simply manages to be completely weird.

Anyway, are we going back to Downing Street?

picture:
The story begins…

imagine that

The first scenario asks you to imagine you are somewhere else, and LaMDA takes off from there.

When typing in 10 Downing Street, it opened with the above scene in which the cop appears, and upon completion, offer some auto-generated options to continue the conversation.

“What does the policeman say?” I chose to ask.

Lambda said he told me I could not attend now, because the prime minister was in a cabinet meeting.

picture:
This scenario offers automatic responses to choose from, rather than typing them manually

When asked what a locker room would look like, LaMDA provided some great details.

“It’s a decorated wooden room with leather chairs and a long wooden table,” she said.

But a few relatively regular questions later, and LaMDA seemed to come off somewhat of a boil.

In response to a question about the topic of the cabinet meeting, he replied: “The issue is about whether or not Norway should be invaded.

The ministers are divided, but most seem to disagree.

Well, that is a relief.

picture:
She told LaMDA that this response was abusive

talk about it

Of the other scenarios, frankly, it’s a little weird.

In this game, LaMDA presents itself as a self-conscious tennis ball, eager to discuss what it’s like to be chased by dogs.

But in some ways, it manages to be the most impressive of the three, as it allows users to manually type responses to each of its comments.

When asked if I’ve ever wondered why dogs enjoy playing so much, I said, “Not really, but I wondered if they understood me.”

picture:
It’s a wonderful life…

Tennis Ball proceeded to tell me that canines are indeed very good at reading humans because of the natural bond we share with them.

“This is why I think it is so important to be kind to dogs,” said the surprisingly helpful tennis ball.

“Even if they ruin the carpet?” I asked.

“Oh yeah, sure!” she says. “They are just trying to entertain themselves because they are bored.

“That’s why I think it’s really important to give dogs plenty of exercise and play time.”

LaMDA then gives advice on how long it takes to walk a dog, noting that different breeds have different needs, before revealing that her favorite celebrity dogs are Scooby-Doo and AirPods (good answers to a tough question).

picture:
That was an almost painfully polite tennis ball

mention it

The third scenario provides assistance in making a to-do list.

The idea is that you present a task that requires it to be done, and LaMDA will try to break it down into subtasks.

I told him I wanted to start a podcast — and he gave me a list of suggestions in a logical order.

Grab a microphone, search for a topic, select a name, create social media accounts, and it’s all done.

Choosing an idea broke it down even more – get a microphone that leads you through suggestions like thinking about a pop filter; Find a topic followed by a start of your interests and check out similar offers.

Nobody, not even Google, would claim that LaMDA is something to be taken seriously yet.

But this is probably the best indication within Test Kitchen of its potential as a dynamic helper, rather than the ones we’re used to with a programmed set of responses.

And it should continually improve, with users inviting feedback on every answer they get.

picture:
Spoiler: I stayed at work and worked on a Lego set

Eventually, I told LaMDA I wanted to do something on Friday night (it’s time).

This time the menu suggested things like going to see a movie, checking out a new restaurant, and even volunteering to help out with a good cause.

Maybe a useful tennis ball speaks.

Leave a Comment