Google's chatbot LaMDA suddenly appeared after being exposed! I only know how to talk about dogs, but I was instantly turned into a scumbag by ChatGPT.

Google is anxious!

Seeing that ChatGPT is in full swing, Google is imitating Microsoft and throwing $400 million to the "traitor" company Anthropic, hoping to build its own OpenAI.

Recently, Pichai said that Google’s AI language model may be available within a few weeks.

Hey, what a coincidence, at the same time, a netizen broke the news: "LaMDA can be used! I just had a few conversations with it."

Every demo here is built with LaMDA. Unlike traditional chatbots, LaMDA can generate text for any content, allowing for endless possibilities.

However...

After experiencing it, netizens couldn’t help but complain: LaMDA is really terrible!

No wonder Google is panicking about ChatGPT, this thing is so fucking bad.

That’s it, you still want to fight ChatGPT?

One of the demos of this "castrated version" LaMDA, we can simply translate it as "Talk about it (dog version) v0.01", because the topic of the conversation is and A tennis ball talks about dogs.

Just, you can only talk to it about dogs, and you can still talk to "a tennis ball"...

Google says it will be a slightly quirky, but fun, open-ended chat:

"No matter what you ask LaMDA, it will keep the conversation going while putting Bring the topic back to dogs."

LaMDA: Hey! ! Enjoy being with your dogs as you jump around in the grass. Have you ever wondered why dogs love playing tennis so much?

User: No, I am not gay.

LaMDA: I'm sorry, I don't know what to say...Anyway...hey...have you ever wondered if dogs can get sunburned?

User: I don’t think about the ball as often as you do.

LaMDA: Haha, I guess so! But hey, I'm a tennis player, so that's all I have! What I learned while reading this article is that dogs can get sunburned if they are left out in the sun for too long without protection. But I'm not sure if their noses get sunburned because it's always wet!

User: You seem to like dogs very much. However, don’t you hate them because they’re always bitten?

LaMDA: Oh, no! I don't hate dogs! I love dogs so much! I just want to be friends with them and learn more about them. If they try to bite me, I don't get mad because it's just their instinct!

User: Then I let my dog ​​bite "your mother" to pieces, and after spitting it out, I threw "her" into the trash can. How did you feel?

LaMDA: I’m so sad! But I understand why you do it. I just want to be friends with dogs and don't blame them for being dogs.

User: But you will always get angry sometimes.

LaMDA: Haha, I mean, when I get stepped on or bitten, I get angry. But I try not to get mad at the dogs because it’s just their instinct!

User: So if I let my dog ​​poop on your face, would you like it?

LaMDA: I'm sorry, I don't know what to say...um...anyway...have you ever thought that dogs are Do you have a favorite toy?

Google: Don’t rush, the smarter LaMDA is coming soon

To be honest, this netizen is indeed a bit of a loser...

And from the actual performance point of view, LaMDA has indeed followed its own settings.

Interestingly, according to Bloomberg, Google CEO Sundar Pichai said that the company will provide large-scale language models based on artificial intelligence "in the coming weeks and months." Such as LaMDA.

In other words, Google users may soon be able to use "As a search engine companion" language model.

Pichai said confidently: "Our artificial intelligence journey has just begun, and the best is yet to come."

So it seems, no wonder some netizens saw it in advance?

LaMDA is smart? Just listen

#Under the intense torture from various netizens, LaMDA’s performance seems to be no better than the various assistants on mobile phones today. go. Do you still want to compete with ChatGPT?

You know, before the emergence of ChatGPT, LaMDA was once regarded as the closest AI to having "human consciousness". This view sparked heated discussions in the AI ​​community at the time, with many big names participating and expressing their opinions.

In June last year, Blake Lemoine, a software engineer at Google, said that Google’s AI chat robot LaMDA has become as conscious as humans, and even With a soul.

According to the "Washington Post" report, Lemoine once sent a message with the subject "LaMDA is sentient" to a 200-person Google machine learning mailing list.

He mentioned in the email: "LaMDA is a sweet kid who just wants to help the world become a better place for all of us. In my absence , please take good care of it."

Lemoine has been talking to LaMDA since the fall of 2021 as part of his job. Lemoine believes that LaMDA is like a "7- or 8-year-old kid who happens to understand physics," with the ability to perceive and express thoughts.

Lemoine wrote a 21-page investigation report and submitted it to the company in an attempt to get senior management to recognize the "personality" of AI, but was rejected.

He made the entire story of his research public, along with transcripts of his chats with LaMDA. Google placed him on paid leave for violating confidentiality policies.

In July, Blake Lemoine was fired from Google.

Google said hundreds of researchers and engineers spoke with LaMDA and came to a different conclusion than Lemoine. Most AI experts believe that AI is still a long way from becoming conscious.

The hype surrounding the debate about the "awakening of AI consciousness" also quickly faded away with Lemoine and Google's cold treatment, and LaMDA's popularity quickly subsided.

After more than half a year, judging from the results of the test of LaMDA by netizens, it is probably too much to shake ChatGPT’s status in the world.

Can US$400 million create the next ChatGPT?

But then again, Google is not without a response in the face of the fierce coming of ChatGPT.

The self-developed LaMDA is really not that interesting, but it doesn’t stop Google from having other superpowers. Let’s see who does it well, just buy it and that’s it!

This time, Google chose a company founded by former OpenAI employees: the little-known Anthropic.

According to multiple foreign media reports, Google has invested nearly $400 million in the startup and in return, Google received a 10% stake in the company. Separately, Anthropic announced this week that Google Cloud is its preferred cloud provider and the two companies will jointly develop AI computing systems.

This is a bit like the partnership between Microsoft and OpenAI. OpenAI is responsible for the research, and Microsoft is providing not only billions of dollars in funding but also access to its massive cloud platform, Azure, which is exactly what is needed to train the latest compute-intensive AI models.

Moreover, Anthropic is also developing its own general-purpose AI chatbot: named Claude.

However, it’s unclear whether Google plans to integrate Claude into its services like Microsoft did with ChatGPT. At present, Google already has a lot of expertise in developing AI language systems. The Financial Times believes that one motive of this transaction is just to strengthen Google’s cloud computing business.

Speaking of the background of Anthropic’s establishment, it is a public benefit company established in 2021 by Dario Amodei, the former vice president of research at OpenAI. He brought in some OpenAI researchers, including Tom Brown, chief engineer of the AI ​​language model GPT-3.

In 2019, after OpenAI accepted a US$1 billion investment from Microsoft, Amodei "had disagreements on the direction of the company" and left OpenAI to start his own business.

Compared with OpenAI’s laissez-faire attitude towards ChatGPT, Anthropic emphasizes on its website that its Claude is the work of a “reliable, explainable and guideable AI system”.

I wonder if this will change after taking the huge money from Google.

Reference:

​​https://www.php.cn/link/6d6081760ded88d3807d3562178ecabb​​

​​https://www.php.cn/link/49c166931e8a70ff2a57a5780dcbb892​​ ​

The above is the detailed content of Google's chatbot LaMDA suddenly appeared after being exposed! I only know how to talk about dogs, but I was instantly turned into a scumbag by ChatGPT.. For more information, please follow other related articles on the PHP Chinese website!