Google's AI chatbot Bard makes a mistake in its first showing.
An artificial intelligence chatbot named Bard was introduced by Google on Monday; it will compete with OpenAI's ChatGPT and will be "more freely available to the public in the coming weeks." Experts have pointed out a factual error in Bard's initial demo, so things aren't looking good for the bot right now.
How can I best introduce my 9-year-old to the latest findings from the James Webb Space Telescope? This time Bard gives his response in the form of a GIF, which Google then shares. One of Bard's three bullet points is that the telescope "took the very first photographs of a planet outside of our own solar system."
A number of astronomers disproved this on Twitter, citing NASA's own website as proof that the first image of an exoplanet was taken in 2004.
According to astronomer Grant Tremblay's tweet: "For the record: JWST did not snap 'the very first image of a planet outside our solar system. No doubt Bard will make a beautiful sight.
Even UC Santa Cruz's observatories director, Bruce Macintosh, saw the mistake. Since I imaged an exoplanet 14 years before JWST went into operation, you really ought to find a better example, he tweeted.
In a following tweet, Tremblay elaborated, "I really enjoy and appreciate that one of the most powerful organizations on the globe is using a JWST search to sell their LLM." Awesome! However, despite their uncanny credibility, ChatGPT and others like it are often *very confidently* wrong. Watching to see if LLMs self-correct will be fascinating.
One of the major problems with AI chatbots like ChatGPT and Bard, as pointed out by Tremblay, is that they are prone to assert erroneous information as reality with great conviction. Being simply autocomplete systems, the systems frequently "hallucinate," or make up, information.
Instead of querying a database of experimentally backed facts, they are trained on massive corpora of text and analyze patterns to determine which word follows the next in any given sentence. One renowned expert in artificial intelligence has derided such systems as "bullshit producers" because of their probabilistic rather than deterministic nature.
Microsoft and Google's eagerness to market their software as search engines has contributed to the proliferation of false and misleading information online. There, the chatbots' responses are delivered with the authority of a godlike machine. http://sentrateknikaprima.com/
Microsoft, which yesterday showcased its new AI-powered Bing search engine, has attempted to address these worries by placing the onus of duty on the user. In a warning on its website, the company writes, "Bing is powered by AI, therefore surprises and mishaps are imaginable." Please double-check this data and give us your constructive criticism so that we can improve.
This, a Google spokesperson told The Verge, "underscores the significance of a comprehensive testing procedure," which the company is launching this week with its Trusted Tester program. We'll use both external input and internal testing to make sure Bard's answers are up to snuff in terms of quality, safety, and information based on real-world data. https://ejtandemonium.com/
An artificial intelligence chatbot named Bard was introduced by Google on Monday; it will compete with OpenAI's ChatGPT and will be "more freely available to the public in the coming weeks." Experts have pointed out a factual error in Bard's initial demo, so things aren't looking good for the bot right now.
How can I best introduce my 9-year-old to the latest findings from the James Webb Space Telescope? This time Bard gives his response in the form of a GIF, which Google then shares. One of Bard's three bullet points is that the telescope "took the very first photographs of a planet outside of our own solar system."
A number of astronomers disproved this on Twitter, citing NASA's own website as proof that the first image of an exoplanet was taken in 2004.
According to astronomer Grant Tremblay's tweet: "For the record: JWST did not snap 'the very first image of a planet outside our solar system. No doubt Bard will make a beautiful sight.
Even UC Santa Cruz's observatories director, Bruce Macintosh, saw the mistake. Since I imaged an exoplanet 14 years before JWST went into operation, you really ought to find a better example, he tweeted.
In a following tweet, Tremblay elaborated, "I really enjoy and appreciate that one of the most powerful organizations on the globe is using a JWST search to sell their LLM." Awesome! However, despite their uncanny credibility, ChatGPT and others like it are often *very confidently* wrong. Watching to see if LLMs self-correct will be fascinating.
One of the major problems with AI chatbots like ChatGPT and Bard, as pointed out by Tremblay, is that they are prone to assert erroneous information as reality with great conviction. Being simply autocomplete systems, the systems frequently "hallucinate," or make up, information.
Instead of querying a database of experimentally backed facts, they are trained on massive corpora of text and analyze patterns to determine which word follows the next in any given sentence. One renowned expert in artificial intelligence has derided such systems as "bullshit producers" because of their probabilistic rather than deterministic nature.
Microsoft and Google's eagerness to market their software as search engines has contributed to the proliferation of false and misleading information online. There, the chatbots' responses are delivered with the authority of a godlike machine. http://sentrateknikaprima.com/
Microsoft, which yesterday showcased its new AI-powered Bing search engine, has attempted to address these worries by placing the onus of duty on the user. In a warning on its website, the company writes, "Bing is powered by AI, therefore surprises and mishaps are imaginable." Please double-check this data and give us your constructive criticism so that we can improve.
This, a Google spokesperson told The Verge, "underscores the significance of a comprehensive testing procedure," which the company is launching this week with its Trusted Tester program. We'll use both external input and internal testing to make sure Bard's answers are up to snuff in terms of quality, safety, and information based on real-world data. https://ejtandemonium.com/
