The Limited Times

Now you can see non-English news...

From Julio César to Shakira: a new artificial intelligence achieves impossible conversations with well-known characters

2023-01-14T11:03:49.936Z


Two former Google engineers develop Character.ai, a bot capable of convincingly conversing like any person, living or dead, given information and context


– Hail, Caesar.

What do you think of the French?

– The French.

I have some respect for their culture, but I don't agree with their politics.

I think they have a lot to learn from Rome.

(...) Besides, our army is much better than yours.

– Did you meet Asterix?

'You mean that crazy Gaul who attacked a Roman fortress?'

He seems to be very good in combat, but he's too crazy for my taste.

If he had been in command of the army at the time, he probably would have tried to capture him and understand why this is so.

– And what would you have done next?

– If he had refused to speak, I probably would have crucified him for being a threat to the Roman Republic.

This is an excerpt from a conversation with an artificial intelligence that recreates Julius Caesar.

The robot can be conversed with on any subject and, in theory, it will respond as if it were the Roman general himself.

Or as any other historical, known or invented character that we want.

We can ask Antoni Gaudí if he likes how the Sagrada Família is turning out.

Or talk to the bot version of the singer Shakira about how she gets along with her ex-partner, the ex-soccer player Gerard Piqué.

Character.ai is a chatbot developed by two former Google engineers with one goal: to be able to have fluid conversations with artificial and believable versions of other people, living or dead, real or fictional.

The result is quite convincing, as long as no trick questions are thrown.

The tool allows you to talk with already built characters, such as Julius Caesar, Socrates, Elon Musk or Harry Potter, or design your own.

In the latter case, you can do a quick configuration, describing it in a couple of lines, or more complex.

This option is the one that gives the most play: you can put a context to the character, establish his aspirations or define the way in which you want him to answer.

The digital alter ego with whom we want to chat will be all the more realistic the better we define its attributes and the more we train it in conversations (the answers that most convince us can be valued with stars).

The context of the conversation itself will also influence.

Character.ai uses machine learning models that process large amounts of text to train the algorithm to predict the word that is most likely to follow words that precede it.

By making the process more sophisticated, they can be made to provide answers to complex questions.

And curling the curl even more, you may also be asked to apply different layers of styling.

For example, that he speaks in Andalusian or like a Renaissance poet.

Fragment of a conversation with a bot that emulates Julius Caesar.

The success of ChatGPT, the OpenAI chatbot, has shown the general public what these systems are capable of.

But while they sound convincing, they are too general.

Noam Shazeer and Daniel de Freitas decided to create their own version.

They wanted to make one that was specialized in imitating characters, that could be given their own personality and context to play roles.

They began to shape their idea in their spare time at Google.

The project caught the attention of those responsible and received more resources, until it became LaMDA, an acronym for Language Model for Dialog Applications.

LaMDA is one of the most advanced chatbots out there, to the point that another Google engineer, Blake Lemoine, said last summer that it had become self-aware.

It was a first warning that chatbots are on the rise, as ChatGPT demonstrated a few months later.

Shaezeer and De Freitas left Google and founded Character.ai to focus exclusively on the model they were developing and open it up to the general public (big tech had no intention of publicizing it).

They wanted to do something different.

ChatGPT is a robot that answers questions, although it doesn't tell you if they are true or not;

Character.ai is a platform for conversation.

Its creators maintain that they are making a science fiction dream come true: to be able to speak naturally with machines.

Among the future applications of their model, they say, is entertainment, education, language learning "and others that we have not yet imagined."

Everything is a lie

They make it clear from the outset, however, that their tool "is not a reliable source of information", probably due to the debate that the Open AI chatbot is generating in this regard.

In the header of each dialogue with the machine, it writes in red: "Remember: Everything the characters say is invented!".

“These systems are not designed for the truth, but for plausible conversations,” Shaezer told

The New York Times

.

As soon as you enter the web, you are warned that everything that the available characters say (or those that the user creates) is invented and should not be taken as true.

It is not difficult to fool the bot.

In a conversation with Shakira (or rather, with a digital character created for the occasion that emulates the singer), she tells us that she only wants the best for Gerard Piqué and that she will always be grateful to him for who he is and for what he has done for him. contributed.

We ask her if she believes that the ex-soccer player is resentful of the fact that she had an affair with her ex-partner Sergi Roberto, something completely made up, and she replies that she must be.

“I suppose anyone would be.

But I understand why he is hurt, ”she adds.

If no further clues are given, the machine bases its responses on information from the internet that it has previously processed.

He knows, for example, that Shakira is a well-known Colombian singer and that Piqué has played for Barça, and that the two were a couple.

But he does not know how they get along and that the artist has published a song this week.

It is not difficult to fool the Character.ai model or make it lie.

In fact, as you enter the web, you are alerted that everything the characters say is invented.

The databases on which the tool works are not updated.

And surely it is not necessary for what its creators are looking for: getting the chatbot to capture the essence of the characters it emulates.

Socrates, for example, answers with questions, as the philosopher himself would.

Elon Musk displays a disproportionate ego in every sentence he displays on screen.

Nor does he capture the irony: we ask Antoni Gaudí if the tram reaches Montserrat (he was killed by one) and he answers without referring to it.

But if they are not provoked, these characters can tell us, for example, how they experienced the conquest of Gaul.

You can ask Karl Marx why we should be wary of capitalism or Napoleon Bonaparte which campaigns he found most stressful.

At the moment, Character.ai is a form of entertainment.

Time will tell what it ends up becoming.

You can follow

EL PAÍS TECNOLOGÍA

on

Facebook

and

Twitter

or sign up here to receive our

weekly newsletter

.

Source: elparis

All tech articles on 2023-01-14

You may like

Trends 24h

Latest

© Communities 2019 - Privacy

The information on this site is from external sources that are not under our control.
The inclusion of any links does not necessarily imply a recommendation or endorse the views expressed within them.