Become a fan of Slashdot on Facebook

 


Forgot your password?
Close
typodupeerror
×
Google

Google CEO's New Memo To Employees: Put Two To Four Hours Into Improving Bard Chatbot (businessinsider.com) 67

Posted by msmash from the all-in dept.
Alphabet CEO Sundar Pichai sent an internal memo to Googlers on Wednesday asking them to contribute 2-4 hours of their time to helping improve Bard, the company's AI chatbot that it intends to integrate into search. From a report: The email signals how Google's urgency in moving to win the next generation of AI-based search. The company has found itself on its back foot as Microsoft took the spotlight for its investment in OpenAI. OpenAI created the popular ChatGPT, a chatbot released in late 2022 which can respond to broad, open-ended questions with human-like answers. Last week, Microsoft unveiled a revamped version of its Bing search engine with ChatGPT, and CEO Satya Nadella called it a "new day" for search.

"I know this moment is uncomfortably exciting, and that's to be expected: the underlying technology is evolving rapidly with so much potential," Pichai wrote in his memo to Googlers. "The most important thing we can do right now is to focus on building a great product and developing it responsibly." Google kicked off "dogfooding," or internally testing, Bard on Tuesday, according to another memo seen by Insider. It already has thousands of external and internal testers using it, submitting feedback regarding the quality, safety, and "groundedness" of Bard's responses, Pichai's memo said.
This discussion has been archived. No new comments can be posted.

Google CEO's New Memo To Employees: Put Two To Four Hours Into Improving Bard Chatbot More

Google CEO's New Memo To Employees: Put Two To Four Hours Into Improving Bard Chatbot

Comments Filter:
  • Comment removed based on user account deletion

  • Pichai has to go (Score:5, Insightful)

    by pete6677 ( 681676 ) on Wednesday February 15, 2023 @03:17PM (#63296343)

    This fool is so in over his head I almost feel bad for him.

    • Re: (Score:2)

      by gweihir ( 88907 )

      This fool is so in over his head I almost feel bad for him.

      Obviously. He seemed to have thought that Google could just continue as it always worked and hence he overlooked important thing. A vastly overpaid moron.

    • No, keep him. Google has too much data for it to succeed. Let them die

    • Re: (Score:2)

      by nuntius ( 92696 )

      Aren't "throw more people at a late project" and "work insane hours" central productivity tenets in classic books such as Brooks' Mythical Man Month and Lister and DeMarco's Peopleware? /sarcasm

      I have this amusing image of thousands of Googlers feeling the need to justify their 4 hours by submitting and reviewing worthless pull requests. GIGO. SNAFU.

  • "Why bother? This thing's just gonna end up in the graveyard [killedbygoogle.com] in a few months anyways..."

    • I don't think chatbots and AI are going to wind up in the graveyard. They work too well and they are legitimate threat to Google's search dominance. If this stuff ends up in the graveyard Google is going to join it there

      • Comment removed (Score:5, Interesting)

        by account_deleted ( 4530225 ) on Wednesday February 15, 2023 @03:40PM (#63296439)
        Comment removed based on user account deletion

        • Lawsuits about unauthorized use of source data are already flying. If your prediction about less Internet surfing, due to uncited chatbot answers, comes true, even more lawsuits will fly. It is hard to see how that will turn out, but even apart from that, Google has good reason to make this investment.

          You seem to be implying that this tech will destroy Google's business, and therefore Google shouldn't develop it. Well, if it is possible to do at all (technically and/or legally), then Google's competitors

          • Re: (Score:2, Interesting)

            by account_deleted ( 4530225 )
            Comment removed based on user account deletion

            • That doesn't make sense.

              If ChatBots won't work (due to tech or law), then they won't kill Google's business. They have to work to be a threat. So if they don't work, this investment is wasted money, but not the death of Google's business.

              If ChatBots do work, then Google's search engine business is about to die. So Google will need a new business. The business that is replacing the search engine business is the ChatBot business. Therefore, google needs to be competitive in that business in order to surv

            • Choose one and only one: A. This is a viable technology, and whomever controls the most popular one will see a huge benefit. B: This is not a viable technology, therefore traditional search will still be required. In the case of A it's a wise investment even if the search engine concept dies. In the case of B Google will still be around and still being Google despite having lost money investing in this. In the case of B it may help along the search engine concept as well. Google search results have been suc
        • Do you think that in a world full of competent or competent-sounding chat AIs your website is going to matter? It's not worth protecting your drop of data when these AIs already have oceans of it. Better let your ideas be learned and replicated this way than ignored.

      • They work too well and they are legitimate threat to Google's search dominance.

        Daphuck?

        What are the legitimate uses for these things besides being a means for kids to cheat on their homework (although having seen common core math, I can't say I blame them)?

        Usually if I'm searching for something it's to locate a specific web destination on the internet that has the content I'm seeking, not to have a question answered. We're also probably way too far off from a bot being able to correctly answer things like "Which retailer has the lowest price on the Raspberry Pi Zero W and actually ha

  • Instead of at least being on the second page of the found results, now they'll get disappeared by a chatty 'I'm feeling lucky' button.

  • How to I get rid of Sundar Pichai?

    Asking for a friend.

  • Scary stuff (Score:5, Interesting)

    by ageoffri ( 723674 ) on Wednesday February 15, 2023 @03:27PM (#63296379)

    I've been in IT for over 25 years and most of that time in Cybersecurity and this scares me like nothing else has. Today I've been playing around with ChatGPT.

    I don't use sed and awk frequently enough to use them without looking up how they work. Today I decided let's use ChatGPT to help me get that part of a bash script right.

    I went down a deep dark rabbit whole of improving the script using ChatGPT. Asking it to improve my script multiple times. While the script has increased at least 10 times in size, the robustness and commenting is top notch.

    The only thing it wasn't able to solve is my input file was a text file in dos format and the script didn't process it. I tried a few questions to get ChatGPT to fix it but it wasn't able to.

    Then I asked ChatGPT to update the script to check if it's dos style and if so, check to see if dos2unix is installed, and then finally run dos2unix.

    One of the outputs I asked it was for all tools it used, to check if they were installed and if they weren't installed to show the user a message on how to install using sudo on Ubuntu.

    Except for giving ChatGPT too many questions and getting locked out for about an hour, this tool has done this part of my job and more.

    • Re:Scary stuff (Score:5, Insightful)

      by gweihir ( 88907 ) on Wednesday February 15, 2023 @03:49PM (#63296475)

      You overlook that you had to give it a _lot_ of guidance and that any insight was provided by you. This thing critically depends on being asked the right questions. Ad any good engineer knows, good questions are harder to come up with than answers. And then you need to make sure it die not produce more or less subtle nonsense. For example, corner cases, performance and security are things it cannot do reliably in any way, unless specifically asked about them in detail.

      • Yesterday I had to give it a lot of guidance, next year it'll take less. If it is truly learning, the next time someone asks ChatGPT to write a bash script in this case checking whois information for a large list of domains, it'll need less prompts.

        Again a big IF, if ChatGPT is able to learn as the basic scripts get clearly done, it just takes an engineer to start asking how to deal with this edge case or make sure that SQL injection isn't allowed. Eventually it wouldn't surprise me that I'll be able to f

    • I've been using it to write Powershell scripts and interrogate Regex patterns.

      It is quite clearly imperfect in its responses, but its way good enough to save me hours of work.

      It is somewhat daunting, but I'm finding it to be quite exciting. I wonder if anyone has asked it to invent AI.

    • Re: (Score:2)

      by kbahey ( 102895 )

      The only thing it wasn't able to solve is my input file was a text file in dos format and the script didn't process it. I tried a few questions to get ChatGPT to fix it but it wasn't able to.

      Then I asked ChatGPT to update the script to check if it's dos style and if so, check to see if dos2unix is installed, and then finally run dos2unix.

      Or do it like that, by piping the output to tr:

      | tr -d '\r' > output.txt

      Couldn't resist ...

    • Fear is just a reminder that something important is happening.

      Human brains are not magical, and everything we mean by the word "intelligence" comes from human brains. It is simply a matter of time before we build machines that can do the same things. Some of the most brilliant people in the world have been working on this problem set for a very long time. It is no surprise that we are seeing improvements.

      There is no need to fear losing your job. Some parts of your job might change, and you might be more

    • Yeah, it can do that. However if you ask it to write you instructions for seducing an 18 year old girl it wouldn't, so our job is never finished!

      Instead of providing clear instructions it will spit it so much nonsense about this being unethical, etc., it's vomit inducing.

  • If you havenâ(TM)t seen ChatGPT yet, go check it out, google has lost. Artists will be out of business, programmers might be in trouble also, google search might be replaced

    • Re: (Score:3)

      by gweihir ( 88907 )

      Actually, the higher skill levels in these fields will have even more work, because all too often Artificial Idiocity produces just blatantly wrong or bad results. Somebody needs to filter and that takes real skill and insight. Of course, coders that struggle with simple stuff, artists that only produce generic stuff, etc. will be in trouble.

  • Bullcrap is more profitable and nobody cares about the results as long as it feels right or is atleast amusing.

  • I don't know about you... (Score:3)

    by Krokus ( 88121 ) on Wednesday February 15, 2023 @03:42PM (#63296447) Homepage

    ...but I can't think of a better way to get something done right than to panic, race to get it done and get it out the door as fast as possible. I mean, that always works, right? Just ask CD Projekt Red.

  • Of course this cannot work. This is not merely a problem of the training data size. And it takes a lot of time, experience and insight for fix the issues, that is if they can be fixed.

    Maybe if Google had some other products besides ads (and the supporting tools like search), this would not be such a pressing issue.

  • Let's see. Something that provides information, with anywhere from 0 to 100% of the result made up out of whole cloth, and delivers it with a show of authority.

    Where have we seen that before?

    (And yes, you can probably include whatever "news" source you love to hate on the most. It's mostly a question of degree.)

    I don't get this fascination with chatbots. If I want made-up shit, I'll open a volume of Grimms'.

  • Pichai needs to go.

  • How should I improve Bard, the Google chatbot intended to replace you?

    ! An error occurred. If this issue persists please contact us through our help center at help.openai.com.

    • How should I improve Bard, the AI intended to replace you?

      As an AI language model, I am designed to understand and generate natural language text, while an AI chatbot like Bard is intended to engage in more conversational interactions with users. Here are some tips to help improve Bard:

      Improve natural language processing: To make Bard more conversational, it's essential to focus on improving its natural language processing (NLP) capabilities. This means training the AI to recognize and understand t

  • ChatGPT stole my girlfriend! It's far more articulate than me, and knows how to swoon her better with love poetry and flattery. Yes, that dress does make her look fat, GPT lied! I thought she was attracted to honesty. El Not!

  • "Uncomfortably excited" is not how I would normally describe 2-4 hours (more!) of unpaid overtime...

  • Automatic training (Score:3)

    by kmoser ( 1469707 ) on Thursday February 16, 2023 @12:24AM (#63297487)
    Maybe they should just hook it up to Bing's AI and train it to not say what Bing says.

Slashdot Top Deals

Business is a good game -- lots of competition and minimum of rules. You keep score with money. -- Nolan Bushnell, founder of Atari

Close