world

Google’s Latest AI Model Can Be Taught How to Solve Problems

This Article was last updated 5 days ago by Adesola

Also Checkout This Trending Articles on TmZ Blog Plastic Surgeries and Implants

Google’s Latest AI Model Can Be Taught How to Solve Problems #Googles #Latest #Model #Taught #Solve #Problems Welcome to TmZ Blog, here is the new story we have for you today:

apps

Click Here to Start Making Money With Your Smart Phone

You Can Click Here To View Restricted Videos/Images in this Article

Google’s latest AI model, called the Pathways Language Model, uses chain-of-thought prompting to learn how to solve problems, the company announced at its Google I/O 2022 event Wednesday.

“Chain-of-thought prompting allows us to describe multistep problems as a series of intermediate steps,” Google CEO Sundar Pichai said. “Kinda like when your teacher gives you a step-by-step example to help you understand how to solve a problem.”

Pichai said usually you would give an AI a question and an answer to use as a key for any future questions you might ask. In chain-of-thought-prompting you still give an AI a question and answer, but you also show the steps you took to reach that answer. Like when you had to show your work on a math-word problem in school.

chain-of-thought-prompting
With a question, answer and steps showing how the answer was reached, Google’s latest AI can solve this math-word problem.


Google

This technique can also be used to teach the AI different languages, so someone who speaks a less-represented language can still access information. In an example, Pichai showed how the AI can learn Bengali from English the same way it learned how to solve math-word problems.

“Just like before we prompt the model with two examples of questions, in Bengali with both Bengali and English answers,” Pichai said. “That’s it. Now we can start asking questions in Bengali.”

Follow Us For More

chain-of-thought-prompting
Teaching Google’s latest AI Bengali. 


Google

While this technique increases the chances of AI getting an answer correct, it’s not perfect. Research from Cornell University found this technique improves an AI’s chances of answering a complex problem to about 58%. Previous models that didn’t use this technique were expected to give a correct answer about 18% of the time. 

“We’re so optimistic about the potential for language models,” Pichai said. “One day we hope we can answer questions on more topics in any language you speak, making knowledge even more accessible in search across all of Google.”


Now playing:
Watch this:

Google Updates Conversational AI at I/O



6:49

Click Here To Continue Reading from Source

1 2Next page

Related Articles

Leave a Reply

Back to top button

AdBlocker Detected

Please disabled your AdBlocker to enable you use our website seamlessly