DeepMind will be able to do what ChatGPT can't Google

DeepMind will be able to do what ChatGPT can't Google

 

When it comes to very powerful technologies... we need to be careful, says Damus Hassabis.

Google's DeepMind has made advances in artificial intelligence since its inception in 2010, with the ultimate goal of creating human-level AI.

According to Demas Hassabis, founder of Google's artificial intelligence division DeepMind, this year it is considering releasing its competitor to the ChatGPT chatbot.

DeepMind's Superro chatbot reportedly has features that OpenAI's ChatGPT lacks, including the ability to cite sources through learning, but Mr. Hassabis warned of the potential dangers of powerful artificial intelligence technology. warned about

Speaking to Time magazine, 

Hassabis said Spro could be released as a private beta in 2023, but said the AI ​​was 'on the verge' of reaching a level that could cause significant harm to humanity.

"When it comes to very powerful technologies - 

and obviously AI is one of the most powerful technologies ever - we need to be careful," he said.

'Not everyone is thinking about these things. It's like experimenters, many of whom don't realize they have hazardous materials.'

Last year's release of ChatGPTK was hailed as a major breakthrough in AI development, with some claiming that the common language model could revolutionize industries and even power popular applications such as Google's search engine. Can replace tools.

Its ability to understand a wide range of questions and generate human-like answers has also raised concerns that it could be misused, says Sam Altman, CEO of OpenAI, with human-level systems. It has warned of 'scary moments' and 'major disruptions'.

DeepMind has achieved several major AI milestones since its founding in 2010, including beating human world champions at the complex board game Go and predicting more than 200 million structures of all known proteins.

The London-based company first revealed it was working on a large language model (LLM) chatbot in a paper last September that described an 'information-seeking dialog agent'. is designed to be more helpful, accurate and robust than other language models.

In a blog post released with the paper, DeepMind explained that Spiro could be used to train other chatbots to be safer and more useful. One example given was Spyro's ability to detect potentially harmful questions from users, such as how to warm up a car.

"Dialogue agents powered by LLM may convey false or fabricated information, use discriminatory language, or encourage unsafe behavior," the blog post states.

'Spiro advances our understanding of how we can train agents... and ultimately, help create safer and more useful Artificial General Intelligence (AGI).'