Since the public opening of ChatGPTChatGPTnews in the field of artificial intelligence (AI) is multiplying at great vitessevitesse. For competitors, it is therefore crucial not to be left behind. GoogleGoogle had announced its own chatbotchatbot Bard in a rush. Last week’s launch of GPT-4 and the new version of ChatGPT certainly forced Google’s hand, which has now opened access to Bard au grand public.
Its operation is similar to ChatGPT: write a queryquery in natural language, and you will receive a written response. The AI is based on LaMDA, Google’s Large Language Model (LLM), the same model that sparked a bizarre case last year, when one of the engineers working on it said the chatbot had developed a consciousness. Unlike ChatGPT and its databasedatabase discontinued in September 2021, Bard is up-to-date and has web access. This time, the firm takes its precautions and highlights the risk of errors, and even goes so far as to provide a specific example where the AI is wrong about the name of a plant.
A launch in English and on the waiting list
However, Google Bard is more limited than its competitor. AI is not capable of producing code, while applicationsapplications and sites created with ChatGPT are multiplying (such as the game Sumplete, for example, inspired by sudoku). However, this is a feature planned for the future, as are “multimodal experiences”, in other words the possibility of interacting other than with text, for example with images, sound or videos.
Despite a public launch, Bard is not yet accessible to everyone, which shows how cautious Google is and would probably have liked to wait a little longer. Users must be of legal age, and the chatbot is only accessible in English and only from the UK or US (or using a VPN). Moreover, it will take register on the waiting list. However, Google wants to add more countries and languages in the future.
Google admits that its AI Bard was not ready for release at all
After a failed presentation of its chatbot Bard, Google’s stock market value lost 9%. The hasty launch was strongly criticized by the firm’s employees, and even the chairman of the parent company’s board of directors judged that the AI was not ready…
Article d’Edward BackEdward Backpublished on 02/14/2023
Tough week for Google, weighed down by artificial intelligence. While MicrosoftMicrosoft was about to announce the integration of OpenAI’s ChatGPT into its Bing search engine, Google couldn’t wait any longer and unveiled its own chatbot, Bard, early last week. However, the firm has rushed too much, and must face the consequences.
In a teaser video, the AI erred in assigning the James Webb Space Telescope the first photo of a exoplanetexoplanet (JWST is however the first to take a picture of an exoplanet in the middle infraredinfrared). A mistake that cost Google dearly since its stock market value fell 9% in one day, a loss of $100 billion.
A launch criticized at all levels
In a conference held on Monday, John Hennessy, chairman of the board of directors of Alphabet, the parent company of Google, indicated that the firm did not expect to launch Bard so quickly. ” I think Google was hesitant to bring it to market because they didn’t think it was market ready yet, but I think as a prototype it’s great technology “. The firm knew that Bard continued to give wrong answers, and would have accelerated the launch to avoid being left behind by Microsoft. This phenomenon, where the chatbot seems to be sure of itself despite a completely wrong answer, has been dubbed “ hallucinationhallucination “. This is common to all AIs of this genre, since ChatGPT is also frequently prone to hallucinations.
According to CNBC, employees also criticized the launch on the messagingmessaging internally, calling it “rushed and sloppy”. Some would not even have been aware that the launch was taking place. Despite the current craze, John Hennessy estimates that this kind of AI still needs a year or two before becoming a tool intended for the general public.
Google Bard crashes on a discovery of the James-Webb telescope
Google’s new artificial intelligence gave the wrong answer in a promotional clip released by the company. Something to worry about the veracity of the answers it will give once integrated into the search engine…
Article by Edward Back, published on 02/09/2023
With ChatGPT from OpenAI and Microsoft, Internet users quickly discovered that the AI did not know how to differentiate its sources, and tended to repeat proven facts as well as errors, misinformation or propaganda, without distinction. Warnings began to multiply about his tendency to be ” confidently incorrect », in other words to release erroneous information with complete assurance.
Google unveiled its own Bard chatbot earlier this week. The question of the accuracy of his answers was immediately raised by many Internet users. The answer was unwittingly given by Google in the form of a tweet from Sundar PichaiSundar Pichaithe CEO of Google.
The first photo of an exoplanet? Really ?
In a music video to promote Bard, he asks the following question: What new discoveries from the James Webb Space Telescope can I tell my 9-year-old about? “. The chatbot gives a three-part answer. If the first two points are correct, the third contained an error. The AI states that ” JSWT has taken the first-ever photos of a planet outside our own solar system “. Something to make the hair of astronomersastronomerswho did not hesitate to point out that the first photo of an exoplanet (dubbed 2M1207b) was taken by the Very Large Telescope of theTHATTHAT (Very Large Telescope, in English or VLT) in 2004, 18 years before the first images of JWST.
The error should be qualified, since the James-Webb space telescope was the first to take a photo of an exoplanet in the mid-infrared. Bard therefore truncated part of the information which changed its meaning. However, if the AI is wrong even in an official Google advertisement, this may cast doubt on the interest of its integration into the search engine. A user can more easily question the content of the sites found via a search, rather than a statement that seems to come directly from Google. After the era of fake newsare we going to face the age of AI misinformation?
Google draws Bard to counter ChatGPT
Faced with the ChatGPT craze, Google cracked down and, from Monday, is opening its conversational AI called Bard to the public.
Article of Louis NephewLouis Nephewpublished on 07/02/2023
At the end of December, Google had triggered a “red code”, a kind of alert to mobilize its troops responsible for the development of Artificial Intelligence (AI), in the face of the sudden rise in power of the chatbot ChatGPT. While the giantInternetInternet He also has several powerful AIs, he had never wanted to release them to the public due to ethical issues in particular.
Google is also very aware that a digital revolution like that of AI could knock it off its pedestal, as it happened to others when the firm launched its innovative search engine in the late 1990s. Not to leave too late, Google has just unveiled its answer to OpenAI, with a chatbot tool called “Bard”.
To announce the news, it was Sundar Pichai, the boss of Google himself, who split an article from blog. Initially, Bard will only be open to so-called “trusted” testers, only to be released to the public later in ” the weeks to come “. In principle, Bard works the same way as ChatGPT. It was fed by an extensive documentary database to generate compelling answers to user questions. The advantage of Bard, compared to ChatGPT whose knowledge base is limited and stops in 2021, is that its algorithm can fetch information for its answers from the Web. And the web is basically Google’s indexing base. That’s not all, as Sundar Pichai also explained that AI-powered features would be associated with the search tool.
An AI in Google’s search engine
This announcement is therefore a way to cut priority to ChatGPT when the AI was able to endanger the historical search engine. And then, it must be said that, on Microsoft’s side, the big maneuvers had also begun. The computer giant accidentally unveiled the integration of ChatGPT directly into the interface of its Bing search engine. The company must also organize an event today at its headquarters in Washington, which we suspect that the subject will be this novelty.
In the meantime, as for Bard, you should know that the machinery is nothing new. The conversational model is based on LaMDA, an algorithm that has been around for about two years. Just as GTP-3 powers ChatGPT, LaMDA dictates Bard’s responses. This AI LaMDA made headlines last summer, when one of its developers was fired after he published the script of a conversation with her.
The objective of the engineers was then to find out if the AI was aware of itself. The result was indeed disturbing. That’s why with Bard’s release to the mainstream, Google is crossing a line it was hesitant to cross until ChatGPT arrived. We can already see the shortcomings of this AI who is gifted in rhetoric, but who knows how to lie or misinform with conviction when she does not have the appropriate knowledge.
However, the integration of the same AI in a search engine with the colossal database of the Web could involve sacred risks in mattermatter of misinformation. As a reminder, last summer, Facebook’s conversational AI BlenderBot 3 had to be abandoned after a few weeks because of its propensity to support conspiratorial and anti-Semitic remarks. She was going to look for information on the Internet to feed her answers. It remains to be hoped that Google has succeeded in setting up guardrailsguardrails.
