Google hopes 'Bard' will outsmart ChatGPT, Microsoft in AI
Updated: 6:11 PM CST Feb 6, 2023
SOLEDAD: WELCOME BACK TO "MATTER OF FACT." ARTIFICIAL INTELLIGENCE IS SHAKING UP THE WAY WE CREATE ART, MUSIC, AND WRITING. IT鈥橲 NOW DIVIDING THE ACADEMIC WORLD AND THREATENING JOBS. CHATGPT IS POPULAR WITH STUDENTS WHO USE THE A.I. TECHNOLOGY TO CREATE ESSAYS AND OTHER WRITING ASSIGNMENTS. THE USER TYPES IN KEYWORDS OR ASKS A QUESTION, AND THE BOT GENERATES ARTICLES, PAPERS, AND EVEN POETRY. TEACHERS ARE CONCERNED ABOUT PLAGIARISM AND THE ETHICS OF LETTING A COMPUTER DO THE TYPING AND THE THINKING. A COMPUTER SCIENCE STUDENT AT PRINCETON UNIVERSITY BELIEVES HE HAS A SOLUTION. EDWARD TIAN CREATED GPTZERO. IT鈥橲 A SOFTWARE THAT DETECTS ESSAYS WRITTEN BY ARTIFICIAL INTELLIGENCE. EDWARD TIAN, IT鈥橲 SO NICE OF YOU TO JOIN US, THANK YOU. SO LET鈥橲 BEGIN WITH A LAYMAN鈥橲 EXPLANATION OF HOW CHATGPT ACTUALLY WORKS. EDWARD: HOW IT REALLY WORKS IS THAT IT鈥橲 INGESTING GIGANTIC PORTIONS OF THE INTERNET. LARGE, LARGE QUANTITIES OF DATA, AND IT鈥橲 LOOKING THROUGH IT FOR PATTERNS, AND IT鈥橲 SORT OF REGURGITATING THESE PATTERNS TO GENERATE TEXT. SO I WOULD CAUTION THAT IT鈥橲 NOT ACTUALLY COMING UP WITH ANYTHING ORIGINAL. IT鈥橲 READING, INGESTING A LOT, AND IT DOES KNOW A LOT AND IT鈥橲 REGURGITATING IT. SOLEDAD: TALK TO ME A LITTLE BIT ABOUT YOUR SOFTWARE, HOW DOES THAT WORK? EDWARD: GPTZERO, IT鈥橲 A SOFTWARE THAT DETECTS IF AN ESSAY IS A.I. OR HUMAN WRITTEN. IT USES TWO BIG VARIABLES RIGHT NOW. ONE WAS PERPLEXITY, WHICH IS ALMOST HAVING A LANGUAGE MODEL LIKE CHATGPT AND ASKING IT HOW FAMILIAR IT IS WITH A PIECE OF TEXT AND HOW LIKELY IT IS TO GENERATE IT ITSELF. AND THEN THERE鈥橲 A BIG PICTURE INDICATOR WE CALL BURSTINESS, WHICH IS MEASURING VARIANCE IN THIS WRITING OVER TIME. HUMANS HAVE CREATIVITY AND SUDDEN BURSTS IN CREATIVITY VERSUS THESE MACHINES ARE PRETTY CONSTANT OVER TIME. SOLEDAD: WHAT WAS THE IMPETUS FOR CREATING GPTZERO? EDWARD: HUMANS DESERVE TO KNOW THE TRUTH, WHETHER SOMETHING IS A.I. GENERATED OR HUMAN WRITTEN, NO ONE WANTS TO BE DECEIVED. A.I. GENERATIVE TECHNOLOGIES ARE LIKE A PANDORA鈥橲 BOX. ONCE WE OPENED IT, THERE鈥橲 NO DIALING BACK. AND WHEN WE鈥橰E BUILDING THESE INCREDIBLE NEW TECHNOLOGIES IT鈥橲 CRITICAL THAT AT THE SAME TIME, WE鈥橰E BUILDING THE TOOLS TO ADOPT THEM RESPONSIBLY. SOLEDAD: DOES YOUR SOFTWARE HAVE ANY LIMITATIONS? WE TRIED IT OUT AND WE FOUND FOR SOMETHING VERY SHORT, LIKE A LITTLE SONNET, IT WAS CHALLENGED. EDWARD: YES, DEFINITELY. IT鈥橲 NOT PERFECT. THERE鈥橲 A LOT OF LIMITATIONS WHEN WE鈥橰E DOING THIS, LIKE WITH SHORT SONNETS AS WELL AS WITH EDGE CASES, DIFFERENT LANGUAGES. AND WHAT WE鈥橰E DOING RIGHT NOW IS WE鈥橰E TRANSITIONING FOR, LIKE, TRAINING TOOLS FOR EDUCATION USE CASES, AND NOW WE鈥橰E HIGHLIGHTING PORTIONS OF ESSAYS THAT ARE MORE LIKELY TO BE HUMAN GENERATED, WHICH TURNS OUT TO BE SOMETHING TEACHERS WANT INSTEAD OF JUST HAVING THIS BLACK AND WHITE, YES OR NO, CATCH-ALL TOOL. SOLEDAD: WE OFTEN TALK ABOUT THE FAILURES OF A.I. BUT I鈥橫 CURIOUS IF YOU THINK CHATGPT HAS A VALUE JUST IN GENERAL. AND OF COURSE, YOU KNOW, SMALL BUSINESSES FOR, FOR TURNING OUT FLIERS OR EMAILS, YOU KNOW, I SORT OF THINK OF A LOT OF USES THAT AREN鈥橳 REALLY PLAGIARISM. EDWARD: IT鈥橲 GREAT AT GETTING PEOPLE STARTED. IN TERMS OF FINISHING THE JOB, I鈥橵E REALIZED THAT YOU REALLY NEED TO DO IT AT THE END, WHETHER IT鈥橲 FINISHING YOUR ESSAY OR FACT CHECKING. BECAUSE AGAIN, ANYTHING THAT鈥橲 WRONG IN THE TRAINING DATA WILL TRANSLATE UP TOWARDS THE END. SO IT鈥橲 NEVER GOING TO BE ABLE TO DO THE JOB OF A JOURNALIST, PER SE, IN CHECKING ALL THE FACTS. SOLEDAD: WHAT DO YOU THINK IS THE FUTURE FOR CHATGPT? EDWARD: I ALMOST THINK OF A FUTURE 10 OR 20 YEARS DOWN THE LINE WHERE EVERYBODY鈥橲 USING CHATGPT TO WRITE THEIR ESSAYS, AND THAT FUTURE SEEMS A LITTLE SAD TO ME. THERE鈥橲 SO MUCH BEAUTY IN HUMAN WRITING, AND THERE鈥橲 ASPECTS THAT THESE COMPUTERS CAN NEVER CO-OPT. I THINK IN THE FUTURE WHERE EVERYONE鈥橲 WRITING WITH THE MACHINE, IT WILL STILL REMAIN A VALUABLE SKILL TO WRITE ORIGINALLY. SOLEDAD: EDWARD TIEN, THANK YOU FOR TALKING WITH ME, A
Google hopes 'Bard' will outsmart ChatGPT, Microsoft in AI
Updated: 6:11 PM CST Feb 6, 2023
Google is girding for a battle of wits in the field of artificial intelligence with 鈥淏ard," a conversational service apparently aimed at countering the popularity of the ChatGPT tool backed by Microsoft.Bard will initially be available exclusively to a group of 鈥渢rusted testers" before being widely released later this year, according to a Monday blog post from Google CEO Sundar Pichai.Google's chatbot is supposed to be able to explain complex subjects such as outer space discoveries in terms simple enough for a child to understand. It also claims the service will also perform other more mundane tasks, such as providing tips for planning a party, or lunch ideas based on what food is left in a refrigerator. Pichai didn't say in his post whether Bard will be able to write prose in the vein of William Shakespeare, the playwright who apparently inspired the service's name.鈥淏ard can be an outlet for creativity, and a launchpad for curiosity," Pichai wroteGoogle announced Bard's existence less than two weeks after Microsoft disclosed it's pouring billions of dollars into OpenAI, the San Francisco-based maker of ChatGPT and other tools that can write readable text and generate new images.Microsoft's decision to up the ante on a $1 billion investment that it previously made in OpenAI in 2019 intensified the pressure on Google to demonstrate that it will be able to keep pace in a field of technology that many analysts believe will be as transformational as personal computers, the internet and smartphones have been in various stages over the past 40 years.In a report last week, CNBC said a team of Google engineers working on artificial intelligence technology 鈥渉as been asked to prioritize working on a response to ChatGPT." Bard had been a service being developed under a project called 鈥淎tlas," as part of Google's 鈥渃ode red鈥� effort to counter the success of ChatGPT, which has attracted tens of millions of users since its general release late last year, while also raising concerns in schools about its ability to write entire essays for students.Pichai has been emphasizing the importance of artificial intelligence for the past six years, with one of the most visible byproducts materializing in 2021 as part of a system called 鈥淟anguage Model for Dialogue Applications," or LaMDA, which will be used to power Bard.Google also plans to begin incorporating LaMDA and other artificial intelligence advancements into its dominant search engine to provide more helpful answers to the increasingly complicated questions being posed by its billion of users. Without providing a specific timeline, Pichai indicated the artificial intelligence tools will be deployed in Google's search in the near future.In another sign of Google's deepening commitment to the field, Google announced last week that it is investing in and partnering with Anthropic, an AI startup led by some former leaders at OpenAI. Anthropic has also built its own AI chatbot named Claude and has a mission centered on AI safety.
Google is girding for a battle of wits in the field of artificial intelligence with 鈥淏ard," a conversational service apparently aimed at countering the popularity of the ChatGPT tool backed by Microsoft.
Bard will initially be available exclusively to a group of 鈥渢rusted testers" before being widely released later this year, according from Google CEO Sundar Pichai.
Google's chatbot is supposed to be able to explain complex subjects such as outer space discoveries in terms simple enough for a child to understand. It also claims the service will also perform other more mundane tasks, such as providing tips for planning a party, or lunch ideas based on what food is left in a refrigerator. Pichai didn't say in his post whether Bard will be able to write prose in the vein of William Shakespeare, the playwright who apparently inspired the service's name.
鈥淏ard can be an outlet for creativity, and a launchpad for curiosity," Pichai wrote
Google announced Bard's existence less than two weeks after Microsoft disclosed it's into OpenAI, the San Francisco-based maker of ChatGPT and other tools that can write readable text and generate new images.
Microsoft's decision to up the ante on a $1 billion investment that it previously made in OpenAI in 2019 to demonstrate that it will be able to keep pace in a field of technology that many analysts believe will be as transformational as personal computers, the internet and smartphones have been in various stages over the past 40 years.
In a , CNBC said a team of Google engineers working on artificial intelligence technology 鈥渉as been asked to prioritize working on a response to ChatGPT." Bard had been a service being developed under a project called 鈥淎tlas," as part of Google's 鈥渃ode red鈥� effort to counter the success of ChatGPT, which has attracted tens of millions of users since its general release late last year, about its ability to write entire essays for students.
Pichai has been emphasizing the importance of artificial intelligence for the past six years, with one of the most visible byproducts materializing in 2021 as part of a system called 鈥淟anguage Model for Dialogue Applications," or LaMDA, which will be used to power Bard.
Google also plans to begin incorporating LaMDA and other artificial intelligence advancements into its dominant search engine to provide more helpful answers to the increasingly complicated questions being posed by its billion of users. Without providing a specific timeline, Pichai indicated the artificial intelligence tools will be deployed in Google's search in the near future.
In another sign of Google's deepening commitment to the field, Google announced last week that it is investing in and partnering with Anthropic, an AI startup led by some former leaders at OpenAI. Anthropic has also built its own AI chatbot named Claude and has a mission centered on AI safety.