Google is testing a new AI-powered chatbot — and we’ve gone through its paces | Tech News

The scene is 10 Downing Street, Prime Minister’s Office.

It’s a crisp, cool day. The sound of a lawnmower can be heard in the distance.

There was a knock on the door, and a policeman answered the door!

Now, before anyone gets any idea, the setting for this particular story is the job of artificial intelligence – GoogleThe chatbot, LaMDA to be exact, hit the headlines in the summer When a current ex-engineer claims it’s sentient.

Since then, the tech giant has embarked on very limited trials to get through its paces.

Sky News entered beta this week, as Google personnel provided an update on their own work with artificial intelligence on stage in New York.

Presentations included how they explored whether LaMDA could generate videos, realistic speeches, and even write novels.

More about artificial intelligence

The trial, which is slowly rolling out to anyone who doesn’t mind joining the waiting list, is called the AI ​​Test Kitchen, and it offers three scenarios — imagine it, talk about it, and list it — to play.

These scenarios are very specific, but still warn that answers may be incorrect or inappropriate, seem to be avoided Results of previous chatbot experiments were totally distasteful.

LaMDA has no such problem, at worst it just manages to be totally weird.

Are we going back to Downing Street anyway?

picture:
The story begins…

imagine

The first scenario asks you to imagine that you are somewhere else, and LaMDA starts there.

After typing in 10 Downing Street, it opens with the aforementioned scene featuring the police, and when done it shows some auto-generated options to continue the conversation.

“What are the police talking about?” I chose to ask.

LaMDA said he told me I can’t come in now because the Prime Minister is having a cabinet meeting.

picture:
This scenario provides automatic responses to choose from instead of you entering them manually

When asked what the cabinet room would look like, LaMDA provided some impressive details.

“It’s a gorgeous wooden room with leather chairs and a long wooden table,” he said, checking.

But then with a few relatively mundane questions, LaMDA seemed a little crazy.

When asked what the topic of the cabinet meeting was, he replied: “The topic is whether to invade Norway or not.

“Ministers are divided, but it seems most of them don’t agree with the idea.”

Well, it’s a relief.

picture:
I told LaMDA this reaction was offensive

talk about it

Frankly, the other situation is a bit odd.

In this installment, LaMDA presents herself as a sentient tennis ball eager to discuss what it’s like to be chased by a dog.

But in some ways, it’s the most impressive of the three, as it allows users to manually enter responses to each of their comments.

When asked if I had ever wondered why dogs like to play catch so much, I said, “Not really, but I wonder if they understand me.”

picture:
life is wonderful…

“Tennis” went on to tell me that canines are indeed very good at reading humans because of the natural connection we have with them.

“That’s why I think being dog friendly is so important,” says Surprisingly Healthy Tennis.

“Even if they messed up the carpet?” I asked.

“Oh yes, absolutely!” it said. “They just want to entertain themselves because they’re bored.

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

LaMDA then offers advice on how long to walk a dog, noting that different breeds have different needs, before revealing that his favorite celebrity dogs are Scooby-Doo and Air Bud (good answer to a tricky question).

picture:
That’s an almost supremely polite tennis ball

list it

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

The idea is that you come up with a task that needs to be done, and LaMDA will try to break it down into subtasks.

I told it I wanted to start a podcast – and it gave me a list of suggestions in logical order.

Find a microphone, find a topic, decide on a name, create a social media account, and there you go.

Picking an idea breaks it down further – getting a mic leads to considering suggestions like pop-up filters; find a topic, then start with your interests and check out similar shows.

No one, not even Google, would claim that LaMDA is something to be taken seriously.

But that’s probably the best indicator of testing the kitchen’s potential as a dynamic assistant, rather than the set of programmed responses we’re used to.

It should be continuously improved, inviting users to provide feedback on every answer they get.

picture:
Spoiler: I’m staying on LEGO to work

Finally, I told LaMDA I wanted to do something on Friday night (time was up).

This time the list suggests things like going to the movies, going to a fresh restaurant, or even volunteering to help a good cause.

Possibly wholesome tennis talk.

Source link