Google Bard: An AI-powered chatbot will soon be based on a more capable language model – Google

Compared to OpenAI’s ChatGPT and Microsoft’s Bing Chat (which also uses the GPT-4 model), Google’s Bard is said to not be as precise or detailed as its competitors. However, this is expected to change soon according to Google CEO Sundar Pichai, who posted on the New York Times podcast “Hard Fork” Certain Over the next few days, Bard will shift from its existing LaMDA model to large-scale PaLM datasets.

When during the podcast I was askedOn how he felt about the response to Bard’s launch, Sundar Pichai commented: “We obviously have more capable models. Very soon, maybe by the time this (podcast) comes out, we’ll have upgraded Bard to some of our more capable PaLM models,” Which will provide more capabilities, both logical and programmatic. To highlight the difference, Google said it had trained LaMDA with 137 billion female teachers when it shared details about its language models last year. On the other hand, PaLM is said to be trained with about 540 billion parameters. Both models may have evolved and evolved since early 2022, but this difference between them probably explains why Google has continued to move Bard to PaLM with its larger data set and greater variety in its answers.

Sundar Pichai claims that he is not worried about the pace of development of Google’s artificial intelligence compared to its competitors. When Bard debuted in February, Sundar Pichai acknowledged that its reliance on LaMDA might have made it smaller, but it put less computing power as a feature, allowing more users to test it and provide relevant feedback.

Sundar Pichai also said that Google will do its own analysis of Bard’s safety and quality once it has more “factual” information (from its interactions with users, evolution of conversations, comments, etc.). From this perspective, Sundar Pichai said, “…we don’t want to launch a more capable model before we’re absolutely sure we can handle it right. We’re all in very early stages. We’ll have more capable models to ‘download’ as time goes on.” But I don’t want it to be just who got there first, but getting it right is very important to us.” This seems to be on the minds of more than 1,800 people (including tech figures, leaders, and AI researchers) who recently signed an open letter calling on companies to hold off for six months on developing AI technology that is “more powerful than GPT-4.” However, Sundar Pichai does not believe that this can be done effectively without government intervention, but agrees with the need for guidance: “Artificial intelligence is a very important field that cannot be regulated (and subject to rules and laws). It is also important not to properly prepare an area. That is why I am particularly pleased that such talks are underway.”