When the AI is wrong: the slips of ChatGpt by Microsoft and Bard by Google
The challenge between search engines is officially open. But can haste also be a bad advisor for artificial intelligence? Some errors made by ChatGpt and Bard
Fortunately, there is still no infallible being on this earth. They are heartened by the mistakes made in recent days by the two most rumored and commented artificial intelligences, ChatGpt, which Microsoft wanted to integrate into its search engine, Bing and Bard which Google has unsheathed in a hurry to remain in the most important ganglion of the Net.
WHAT IS GOING ON
In recent days, the world's leading search engine presented the anti ChatGPT with a demo and various tabs with examples to demonstrate to everyone that Mountain View's AI has nothing less than its competitors. But Bard was spectacularly wrong on the James Webb Space Telescope.
Not to be a ~well, actually~ jerk, and I'm sure Bard will be impressive, but for the record: JWST did not take “the very first image of a planet outside our solar system”.
the first image was instead done by Chauvin et al. (2004) with the VLT/NACO using adaptive optics. https://t.co/bSBb5TOeUW pic.twitter.com/KnrZ1SSz7h
— Grant Tremblay (@astrogrant) February 7, 2023
In one of the requests made to Bard, several astronomers and experts pointed out a significant mistake on Twitter. He was asked to tell a nine-year-old how the James Webb Space Telescope, sent into orbit in 2021, works. «The JWST has taken the first ever photos of a planet outside our solar system. These distant worlds are called “exoplanets”», explains pedantic Bard. Wrong. The first image ever taken of an exoplanet dates back to 2004 and the Very Large Telescope is the author. A crash on the stock market followed (also purely virtual, better to reiterate it) and some controversy over how Google's number 1 managed the dossier.
NOT ONLY BARD, ALSO CHATGPT “MISTAKES”
In short, Google is wrong and is overtaken by Microsoft? Not really, because even the advertisements that extol the qualities of the super Bing animated by the technology behind ChatGpt would contain errors and inaccuracies. The software engineer Dmitri Brereton is convinced of this and in his newsletter, DKB Blog , lists them one by one: «Bing's artificial intelligence got some answers completely wrong during the demonstration. But no one noticed," he writes. For example, completely incorrect information would be found – and based on non-existent reviews – regarding a vacuum cleaner, guilty according to the search engine of being too noisy, something not proven in any human review.
Brereton then cites another research, advertised on the Microsoft website, regarding some advice on nightlife for those planning a trip to Mexico City. The list of alternatives is rich and certainly more orderly than what we would find on Google, but the suggested restaurants do not pass the TripAdvisor tests. Finally, the AI has made some typos within a summary work of a financial document: the numbers of the original text do not correspond to those of the short text produced by the artificial intelligence. In short, even artificial intelligence is wrong, but human intelligence has certainly not been able to do better if these errors have been poured into the promotional material used to launch the new Bing…
This is a machine translation from Italian language of a post published on Start Magazine at the URL https://www.startmag.it/innovazione/quando-lia-sbaglia-gli-scivoloni-di-chatgpt-di-microsoft-e-di-bard-di-google/ on Wed, 15 Feb 2023 06:23:10 +0000.