Think about a rising star in the executive suite that your company recently hired that was so appealing that a competitor firm simply employed a doppelganger. They are surrounded by an exhilarating buzz. Everyone appears to agree that this person is the company's future, from the CEO to the stockholders.
Then you find out that the executive has a "hallucination problem" (Opens in a new tab). There is a 15 to 20% probability that they may simply make up sentences every time they speak (Opens in a new tab). A Princeton professor refers to the individual as a bullshit producer (Opens in a new tab). They are unable to distinguish between fact and fiction (Opens in a new tab). In five minutes, they'll take the stage to introduce a brand-new item. Do you still make them the center of attention?
This week, Microsoft and Google's responses were in the affirmative. After two months of ChatGPT's popularity, which saw 100 million monthly active users, Microsoft organized a last-minute surprise event to reveal that OpenAI will be bringing ChatGPT-style search to the Bing search engine and Edge browser. The day before and the day after it was launched in Paris, Google announced Bard, an AI search tool of its own. However, Bard encountered its own hallucination issue.
Microsoft CEO Satya Nadella told media gathered at the Redmond, Washington headquarters on Tuesday, "A new race starts today." Yes, it is lovely to believe that (Opens in a new tab). Microsoft, the chronically uncool kid on the IT block, would like you to believe that Google and Bing are competing for supremacy in all areas of search.
Condescension oozed from Google's pre-response announcing Bard(Opens in a new tab): Google CEO Sundar Pichai tweeted, "Six years ago, we re-oriented the firm around AI."
The "hallucination problem" and Google
It's a telling point, really. Even while Google, the market leader in search, has had years to add AI, ChatGPT's rival Bard is only just entering the beta stage with a small number of testers. The Bard unveiling had an unanticipated messiness to it despite Pichai's hipster act. By all the ChatGPT noise, Google also appears to have been caught off guard.
What other way is there to account for the embarrassing Bard error that was prominently displayed at launch—not at the actual event, where some demo blunders are anticipated, but in a pre-made GIF? A user is seen asking Bard for information about the James Webb Space Telescope that he may share with his 9-year-old.
It is incorrect that the JWST captured the first image of an exoplanet, which is one of those "facts." Bard was having delusions (Opens in a new tab). (UPDATE: A Financial Times reporter claims Bard's statements were technically accurate(Opens in a new tab), but that requires understanding terminology that no human would ever use, which is still another issue with AI search.))
It's understandable why parent firm Alphabet saw a loss of up to 8% in the value of its stock on the day of the Bard launch. Google highlighted the primary issue with AI search and further implied that the corporation cannot use its enormous data repository to fact-check itself.
Given that it already had a "hallucination problem" with its featured snippets(Opens in a new tab) at the top of search results back in 2017, Google ought to be aware of better practices. The snippets algorithm seemed to take special pleasure in making up lies about American presidents. What more could possibly go wrong?
Never rely just on one source: The updated guidelines for online learning
In other words, if you start using an AI search tool too soon, you run the risk of embarrassing yourself. Microsoft was fortunate in that there were no glaring mistakes at its launch ceremony. But why is ChatGPT-based search in such a hesitant beta stage if it weren't rife with errors? A sign-up sheet is available if you're interested in performing unpaid AI quality assurance work for Bing (Opens in a new tab).
When asked by Wired about ChatGPT's hallucination issue, Sarah Bird, Microsoft's Head of Responsible AI (a revealing title! ), responded, "There's still more to do there" (Opens in a new tab). Yes, seriously: the 15% hallucination rate was provided by a business that is competing to develop a ChatGPT fact-checker (Opens in a new tab). (UPDATE: A New York Times columnist's breathless investigation into New Bing(Opens in a new tab) discovered that it couldn't even produce a list of nearby kid-friendly activities or simple math queries correctly.))
Bird noted that although this functionality had been deactivated, earlier iterations of the software might have assisted users in plotting a school massacre. Nice to know! What possibly could go wrong next? There can't possibly be any other unexpected effect in this hallucinogenic beta search tool that might put a massive and legally weak internet behemoth in the hot seat.
Clippy. Zune. Fresh Bing.
Naturally, Microsoft is aware of this embarrassment because it produced Clippy, one of the worst software failures in history. The paperclip assistant had a reputation for giving out bad advise. In that we approach ChatGPT with inquiries, it differs from Clippy in that respect.
However, the fact that it frequently hallucinates its responses—or, more frequently than you'd expect, simply tells users, "I can't answer that"—could turn ChatGPT-enabled Bing into a type of Clippy on LSD. The "New Bing" will be recognized for its distorted results if enough casual users experience them.
It doesn't matter whether a product gets better down the road; what matters is how people react when it first comes out. Since Microsoft created the Zune, it should also be aware of this. Launching ChatGPT before it is genuinely ready for prime time is similar.
Actually, "The New Bing" kind of begs to be the punchline. Or, if Bing and Edge were to prevail in the AI search race—whatever "winning" actually entails in this context—would you be willing to give up Google search and your Chrome browser in favor of Bing and Edge? Did not believe so. Tech inertia is a force that is vastly underappreciated.
In some situations, ChatGPT is impressive—real estate agents in particular love it(Opens in a new tab) for creating listings—but it makes people nervous in others. But if you get past the headline, every report about its disruptions appears somehow less important. There will be an uptick in student plagiarism as a result. However, it also has a feature that lets you know when a paper was written by ChatGPT(Opens in a new tab), which eliminates its own threat. It passed a test in law school! However, it only managed to earn a C-plus (Opens in a new tab). http://sentrateknikaprima.com/
The difficulty of creating "universal AI," the digital equivalent of a human brain, is well known in the field of artificial intelligence. Another long-held AI objective, the development of insect intelligence, is just just beginning(Opens in a new tab). If you didn't click on the links yourself, would you really trust ChatGPT to deliver your search results?
Depending on how much you, dear reader, are experiencing hallucinations, the answer may or may not apply to you. https://ejtandemonium.com/
Then you find out that the executive has a "hallucination problem" (Opens in a new tab). There is a 15 to 20% probability that they may simply make up sentences every time they speak (Opens in a new tab). A Princeton professor refers to the individual as a bullshit producer (Opens in a new tab). They are unable to distinguish between fact and fiction (Opens in a new tab). In five minutes, they'll take the stage to introduce a brand-new item. Do you still make them the center of attention?
This week, Microsoft and Google's responses were in the affirmative. After two months of ChatGPT's popularity, which saw 100 million monthly active users, Microsoft organized a last-minute surprise event to reveal that OpenAI will be bringing ChatGPT-style search to the Bing search engine and Edge browser. The day before and the day after it was launched in Paris, Google announced Bard, an AI search tool of its own. However, Bard encountered its own hallucination issue.
Microsoft CEO Satya Nadella told media gathered at the Redmond, Washington headquarters on Tuesday, "A new race starts today." Yes, it is lovely to believe that (Opens in a new tab). Microsoft, the chronically uncool kid on the IT block, would like you to believe that Google and Bing are competing for supremacy in all areas of search.
Condescension oozed from Google's pre-response announcing Bard(Opens in a new tab): Google CEO Sundar Pichai tweeted, "Six years ago, we re-oriented the firm around AI."
The "hallucination problem" and Google
It's a telling point, really. Even while Google, the market leader in search, has had years to add AI, ChatGPT's rival Bard is only just entering the beta stage with a small number of testers. The Bard unveiling had an unanticipated messiness to it despite Pichai's hipster act. By all the ChatGPT noise, Google also appears to have been caught off guard.
What other way is there to account for the embarrassing Bard error that was prominently displayed at launch—not at the actual event, where some demo blunders are anticipated, but in a pre-made GIF? A user is seen asking Bard for information about the James Webb Space Telescope that he may share with his 9-year-old.
It is incorrect that the JWST captured the first image of an exoplanet, which is one of those "facts." Bard was having delusions (Opens in a new tab). (UPDATE: A Financial Times reporter claims Bard's statements were technically accurate(Opens in a new tab), but that requires understanding terminology that no human would ever use, which is still another issue with AI search.))
It's understandable why parent firm Alphabet saw a loss of up to 8% in the value of its stock on the day of the Bard launch. Google highlighted the primary issue with AI search and further implied that the corporation cannot use its enormous data repository to fact-check itself.
Given that it already had a "hallucination problem" with its featured snippets(Opens in a new tab) at the top of search results back in 2017, Google ought to be aware of better practices. The snippets algorithm seemed to take special pleasure in making up lies about American presidents. What more could possibly go wrong?
Never rely just on one source: The updated guidelines for online learning
In other words, if you start using an AI search tool too soon, you run the risk of embarrassing yourself. Microsoft was fortunate in that there were no glaring mistakes at its launch ceremony. But why is ChatGPT-based search in such a hesitant beta stage if it weren't rife with errors? A sign-up sheet is available if you're interested in performing unpaid AI quality assurance work for Bing (Opens in a new tab).
When asked by Wired about ChatGPT's hallucination issue, Sarah Bird, Microsoft's Head of Responsible AI (a revealing title! ), responded, "There's still more to do there" (Opens in a new tab). Yes, seriously: the 15% hallucination rate was provided by a business that is competing to develop a ChatGPT fact-checker (Opens in a new tab). (UPDATE: A New York Times columnist's breathless investigation into New Bing(Opens in a new tab) discovered that it couldn't even produce a list of nearby kid-friendly activities or simple math queries correctly.))
Bird noted that although this functionality had been deactivated, earlier iterations of the software might have assisted users in plotting a school massacre. Nice to know! What possibly could go wrong next? There can't possibly be any other unexpected effect in this hallucinogenic beta search tool that might put a massive and legally weak internet behemoth in the hot seat.
Clippy. Zune. Fresh Bing.
Naturally, Microsoft is aware of this embarrassment because it produced Clippy, one of the worst software failures in history. The paperclip assistant had a reputation for giving out bad advise. In that we approach ChatGPT with inquiries, it differs from Clippy in that respect.
However, the fact that it frequently hallucinates its responses—or, more frequently than you'd expect, simply tells users, "I can't answer that"—could turn ChatGPT-enabled Bing into a type of Clippy on LSD. The "New Bing" will be recognized for its distorted results if enough casual users experience them.
It doesn't matter whether a product gets better down the road; what matters is how people react when it first comes out. Since Microsoft created the Zune, it should also be aware of this. Launching ChatGPT before it is genuinely ready for prime time is similar.
Actually, "The New Bing" kind of begs to be the punchline. Or, if Bing and Edge were to prevail in the AI search race—whatever "winning" actually entails in this context—would you be willing to give up Google search and your Chrome browser in favor of Bing and Edge? Did not believe so. Tech inertia is a force that is vastly underappreciated.
In some situations, ChatGPT is impressive—real estate agents in particular love it(Opens in a new tab) for creating listings—but it makes people nervous in others. But if you get past the headline, every report about its disruptions appears somehow less important. There will be an uptick in student plagiarism as a result. However, it also has a feature that lets you know when a paper was written by ChatGPT(Opens in a new tab), which eliminates its own threat. It passed a test in law school! However, it only managed to earn a C-plus (Opens in a new tab). http://sentrateknikaprima.com/
The difficulty of creating "universal AI," the digital equivalent of a human brain, is well known in the field of artificial intelligence. Another long-held AI objective, the development of insect intelligence, is just just beginning(Opens in a new tab). If you didn't click on the links yourself, would you really trust ChatGPT to deliver your search results?
Depending on how much you, dear reader, are experiencing hallucinations, the answer may or may not apply to you. https://ejtandemonium.com/
