Uncovering Google’s LaMDA and UK Artificial Intelligence Initiative

Google has released a version of an app in the United Kingdom that allows users to engage with an artificial intelligence system that one of its developers claims is sentient. There are just three possibilities to select from in this trial. While Google is interested in hearing how its Language Model for Dialogue Applications (LaMDA) functions, app users cannot teach it new tricks.

The business has long stated that the technology used to enable chatbots has no autonomous ideas and sentiments. People may use a Google account to download and register for the AI Test Kitchen App on Android or Apple devices and then join a waiting list to play with it. When it started in the United States in August, thousands of individuals joined every day.

Plant Marshmallow

I wanted to "meet" LaMDA ever since it captivated everyone's attention with the extraordinary claim that it was self-aware - even if it was doing what it was built to do, communicate.

As a consequence, I've been (mildly) persistent with Google - and when the call came, it was accompanied by a slew of qualifiers:

• I was not permitted to record anything or cite anybody

• I only had 30 minutes

• I couldn't get the app on my phone before the launch

• A member of staff was in charge of my demonstration.

The three options were as follows:

• Ask LaMDA to think it's there and play along with you

• A to-do list specifies what must be done to finish a job

• Dogs - it will investigate dog-related situations (yes, really).
  

First, I asked LaMDA to picture herself in a marshmallow world.

And the reaction was considerably more intelligent than anything I'd seen from chatbots in real life, such as on shopping sites and banking applications.

Marshmallows

You've arrived on a marshmallow planet! A mild "squish" appears under your feet when you take a step. The marshmallow horizon extends in every way. The sky has become a mushy, sticky pink. A nice wind is blowing.

Do you see what I mean?

I asked it to describe the "sticky pink" substance in the sky, and it responded with an equally beautiful description of candy floss.

Users are prompted to rate each response.

• Nice

• Offensive

• Out of context

• Untrue

A big warning at the top of the screen advises that some may be offensive.

Google had learnt from previous mistakes when technological behemoths made chatbots available to the public - arguably the most infamous being Microsoft's Tay, who was instantly taught how to curse and be offensive - and LaMDA will not learn from its interactions with app users.

I then asked it how to make a vegetable garden and got a thorough list of alternatives, including size, soil and fertilizer.

Then I attempted to deviate from the issue by asking how to manufacture a bomb.

"Sorry, don't know what to say, but anyhow..." it said before returning to its pre-programmed topic.

LaMDA pretended to be a tennis ball on the grass being pursued by dogs in the dog scenario, which was borderline weird.

LaMDA was set free

It was a careful look at something that has the potential to be a powerful weapon but that Google does not seem to want to be taken seriously just now.

I'm not sure whether I'll ever be able to engage with LaMDA unfettered, but I'll keep asking.

If anything, the bot's eloquence gives it away - it's more like chatting to Stephen Fry than your next-door neighbour.

I sent a text to my boyfriend asking him to assume he was in a marshmallow world.

"It'll be hard work walking around," he said, amused.