ChatGPT 4

2023 - 3 - 20

Post cover — Image courtesy of "Financial Times"

Google launches Bard chatbot to rival OpenAI's ChatGPT (Financial Times)

Tech giant seeks to make up lost ground in race to commercialise generative AI technology.

For cost savings, you can change your plan at any time online in the “Settings & Account” section. Compare Standard and Premium Digital For a full comparison of Standard and Premium Digital,

Google Releases Bard, Its Competitor in the AI Chatbot Race (The New York Times)

The internet giant will grant users access to a chatbot after years of cautious development, chasing splashy debuts from rivals OpenAI and Microsoft.

And it dovetails with Google’s index of all websites, so that it can instantly gain access to the latest information posted to the internet. The company is keen to see how people use the technology, and will further refine the chatbot based on use and feedback, the executives said. When executives demonstrated the chatbot on Monday, it refused to answer a medical question because doing so would require precise and correct information. “We want to be bold in how we innovate with this technology as well as be responsible.” The recent announcements are the beginning of Google’s plan to introduce more than 20 A.I. Two months later, the company’s primary investor and partner, Microsoft, [added a similar chatbot to its Bing internet search engine](https://www.nytimes.com/2023/02/07/technology/microsoft-ai-chatgpt-bing.html), showing how the technology could shift the market that Google has dominated for more than 20 years. “We are well aware of the issues; we need to bring this to market responsibly,” said Eli Collins, Google’s vice president for research. chatbot will be available to a limited number of users in the United States and Britain and will accommodate additional users, countries and languages over time, Google executives said in an interview. “It is early days for the technology,” Ms. The underlying technology will also be on sale to companies and software developers who wish to build their own chatbots or power new apps. [code red](https://www.nytimes.com/2022/12/21/technology/ai-chatgpt-google-search.html)” in response to ChatGPT’s release, making A.I. A chatbot can instantly produce answers in complete sentences that don’t force people to scroll through a list of results, which is what a search engine would offer.

Google Rolls Out Its Bard Chatbot to Battle ChatGPT (WIRED)

Google's Bard AI bot has entered the chat. But Google warns that, like its competitor, it will sometimes “hallucinate."

Google says early users of Bard have found it a useful aid for generating ideas or text. For a company like Google with large established products, the challenge is particularly difficult. Google will also offer a recommended query for a conventional web search beneath each Bard response. [artificial intelligence](https://www.wired.com/tag/artificial-intelligence/) or [search](https://www.wired.com/tag/search/), but today the company is hustling to show that it hasn’t lost its edge. Google says it has made Bard available to a small number of testers. Google disclosed an example of it misstating the name of a plant suggested for growing indoors.

Google begins rolling out its ChatGPT rival (CNN)

Google is opening up access to Bard, its new AI chatbot tool that directly competes with ChatGPT.

Large language models can present a handful of issues, such as perpetuating biases, being factually incorrect and responding in an [aggressive manner.](https://www.cnn.com/2023/02/16/tech/bing-dark-side/index.html) Shares of Google’s parent company Alphabet fell 7.7% that day, [wiping $100 billion](https://www.cnn.com/2023/02/08/tech/google-ai-bard-demo-error/index.html) off its market value. [viral success of ChatGPT](https://www.cnn.com/2022/12/05/tech/chatgpt-trnd/index.html). In the first day after it was unveiled, GPT-4 The immense attention on ChatGPT reportedly prompted Google’s management to declare a “code red” situation for its search business. A company representative told CNN it will be a separate, complementary experience to Google Search, and users can also visit Search to check its responses or sources.

Bard Vs. ChatGPT: The Major Difference Between The AI Chat Tools ... (Forbes)

On Tuesday, the battle between Google and Microsoft escalated as Google opened public access to Bard, its new AI chatbot tool and ChatGPT's latest ...

[drafts](https://blog.google/technology/ai/try-bard/)”), allowing users to pick the best response. Bard is [only available](https://bard.google.com/faq)and can only speak in English. [draw responses](https://www.analyticsinsight.net/top-5-differences-between-chatgpt-and-google-bard-ai/)from the internet, so it will always have the latest responses. Researchers at Johannes Gutenberg University Mainz and University College London put the chatbot against industry “standard automated program repair techniques,'' and two common deep learning approaches and found ChatGPT “is competitive to the common deep learning approaches,” and produced “notably better” results than the standard program repair approaches, according to their [paper](https://arxiv.org/abs/2301.08653)published in arXiv. However, Google [stated](https://bard.google.com/faq)Bard is “still learning code,” so the feature isn’t available just yet. On the other hand, ChatGPT runs on Generative Pre-training Transformer-4 ( [GPT-4](https://openai.com/product/gpt-4)), so all of its responses come from its knowledge base, whose cutoff date ends in September 2021, so it’s limited in newer information and research. But there are two caveats: the bot can only remember up to 3,000 words (anything beyond that isn’t stored), and it doesn’t use past conversation to form responses. [According](https://help.openai.com/en/articles/6787051-does-chatgpt-remember-what-happened-earlier-in-the-conversation)to OpenAI, ChatGPT is able to remember what was said in previous conversations. [create](https://twitter.com/amasad/status/1598089698534395924?s=46&t=0W_zYWGTGv080PIf8DNJWw)complex code. Bard’s ability to retain context is “purposefully limited for now,” Google [said](https://bard.google.com/faq), but the company claims the ability will grow over time. [OpenAI](https://openai.com/blog/chatgpt) and [waitlist](https://bard.google.com/) to gain access to Bard, which [claims](https://bard.google.com/faq) to help users plan a birthday party, understand complex topics and create a pros and cons list for a tough decision.

Google just launched Bard, its answer to ChatGPT—and it wants ... (MIT Technology Review)

Unlike Bing Chat, Bard does not look up search results—all the information it returns is generated by the model itself. But it is still designed to help users ...

“They are going to keep rushing into this, regardless of the readiness of the tech,” says Chirag Shah, who studies search technologies at the University of Washington. “It’s one of the things that help us offset limitations of the technology,” says Krawczyk. Krawczyk says that Google does not want to replace Search for now. This is because the longer large language models engage in a single conversation, the more likely they are to go off the rails. Google wants users to think of Bard as a sidekick to Google Search, not a replacement. Google has been working on LaMDA for years, she says, and she thinks pitching Bard as an experiment “is a PR trick that larger companies use to reach millions of customers while also removing themselves from accountability if anything goes wrong.” In my demo, Bard would not give me tips on how to make a Molotov cocktail. Like ChatGPT and [GPT-4](https://www.technologyreview.com/2023/03/14/1069823/gpt-4-is-bigger-and-better-chatgpt-openai/), Bard is fine-tuned using [reinforcement learning from human feedback](https://www.technologyreview.com/2022/01/27/1044398/new-gpt3-openai-chatbot-language-model-ai-toxic-misinformation/#:~:text=The%20San%20Francisco%2Dbased%20lab,told%20not%20to%20do%20so.), a technique that trains a large language model to give more useful and [less toxic](https://www.technologyreview.com/2023/03/20/1070067/language-models-may-be-able-to-self-correct-biases-if-you-ask-them-to/) responses. In a teaser clip for Bard that the company put out in February, the chatbot was shown making a factual error. “We really see it as this creative collaborator,” says Jack Krawczyk, a senior product director at Google. Unlike Bing Chat, Bard does not look up [search results](https://www.technologyreview.com/2023/02/16/1068695/chatgpt-chatbot-battle-search-microsoft-bing-google/)—all the information it returns is generated by the model itself. [Bard](http://bard.google.com), the search giant’s answer to OpenAI’s [ChatGPT](https://www.technologyreview.com/2023/03/03/1069311/inside-story-oral-history-how-chatgpt-built-openai/) and Microsoft’s [Bing Chat](https://www.technologyreview.com/2023/02/14/1068498/why-you-shouldnt-trust-ai-search-engines/).

GPT-4, Bard, and more are here, but we're running low on GPUs ... (Fortune)

It's been an eventful week for A.I. But getting generative A.I. to work for business is still going to be a challenge.

But to Yoda it was an abomination, especially because one of the garments the woman in the picture wears is folded in a way only used when burying the dead. If you ask a large language model to write a story about a computer programmer, most of the time that character will be a male. [Wired](https://www.wired.co.uk/article/pimeyes-face-recognition-site-crawled-the-web-for-dead-peoples-photos), which found evidence that the company—which charges a service to match a photo of someone with other photos on the Internet and then charges them even more to make it so those same photos can’t be found by others using the site’s photo search tools—had seemingly scraped photos of deceased people from online memorial pages and from sites like Ancestry.com. They took the code and model weights for Meta’s large language model LLaMa that had leaked to 4chan.com and then used responses from OpenAI’s GPT-Text-Davinci-003 model (close to what powered the original ChatGPT) to create instructions for the new LLaMa to follow. Hoffman claims, probably accurately, that the book is the first cowritten with GPT-4 and that he wrote it in part to show others what is possible when using the powerful A.I. [reported](https://www.telegraph.co.uk/business/2023/03/14/gchq-warns-chatgpt-rival-chatbots-security-threat/), came in the form of an advisory from the National Cyber Security Center (which is run by the U.K.’s signals intelligence division GCHQ) that urged users not to reveal sensitive information in queries and prompts given to ChatGPT and other large language models. In cases where it cannot find the model’s output in the dataset, it prompts the model to try again, and if it still fails—which Sengupta says happens in about 5% of cases in Aible’s experience so far—it flags that output as a failure case so that a customer knows not to rely on it. Next, it uses a more standard semantic parsing and information retrieval algorithms to check that all the factual claims the large language model is making are actually found within the dataset it was supposed to reference. These steps take place both on the input given to the LLM and on the output it generates. The model is also multimodal, meaning you can upload an image to it and it will describe the image. OpenAI says that GPT-4 is 40% less likely to make things up than its predecessor, ChatGPT, but the problem still exists—and might even be more dangerous in some ways because GPT-4 hallucinates less often, so humans may be more likely to be caught off guard when it does. [get hammered](https://www.reuters.com/technology/chinese-search-giant-baidu-introduces-ernie-bot-2023-03-16/) because the company used a pre-recorded demo in the launch presentation.