Shulou(Shulou.com)11/24 Report--

First smile, Google's own development of artificial intelligence tools, unexpectedly advocated breaking up Google to stop monopoly? On the issue of antitrust, Google has just opened a public test of the generative AI Bard.

Bard is long overdue to support the spin-off of Google's Bard beta. On Tuesday, Google announced the release of Beta beta version of generative AI Bard, gradually opening testing to the registered public in batches. However, at present, public testing is only available to users in the United States and the United Kingdom, and only English is supported for the time being. Users in other regions and languages need to wait for subsequent updates to Bard.

Although Google has a long product line across many fields, Bard is undoubtedly the most concerned and important new product in recent years, and it is more directly related to Google's competitiveness and market prospects in the next few years. Although Google publicly unveiled Bard in early February, it did not open public testing, but only to Google employees and a small number of invited trusted testers.

In early February, Google made a factual error during the presentation of the Bard (misintroduction of the basics of the Weber Telescope). Although the glitch is so professional that only astronomers can detect it, it still dragged down Google's parent company's share price by 8%, wiping $100 billion off its market capitalization in one day.

Google CEO sent internal memos, perhaps because of this warning, Google executives are very cautious about the public test of Bard. "We have taken a responsible attitude towards research and development, including inviting 10,000 trusted people from different backgrounds and positions to conduct tests," Google CEO Pichai said in an internal memo to employees on Tuesday after the opening of the trial. "We also welcome all feedback, learn from it, iterate and improve constantly."

It seems that Pichai is vaccinating Bard against possible mistakes. "even after all these developments, we are still in the early stages of a long AI journey. As more and more people start to use Bard and test performance, they will surprise me and may cause problems."

At the same time, Google is playing down the impact of generative AI on its search business, emphasizing that Bard is not a substitute for search engines, but a supplementary aid. It is worth mentioning that Bard is a separate web dialog box, not as a Google search tool, but provides a Google search button after the answer.

ChatGPT exerts heavy pressure over the past three months, the hottest product in the technology industry is undoubtedly the ChatGPT. After the release of this generative AI product created by OpenAI, it has become popular all over the world, making users feel an unprecedented intelligent experience, directly driving the next technology tuyere, and even into the Chinese Internet venture capital circle.

On the other hand, the Internet giant Microsoft continues to increase its strategic investment in OpenAI, firmly tie it to its own Azure cloud platform, access its own office components and Bing search after the ChatGPT fire, make AI the core competitiveness of its products, and once again attack Google, which dominates the search market.

Behind this ChatGPT craze, nothing is more embarrassing than the industry giant Google. Although Google started developing generative AI as early as 2015 and launched the dialogue application language model LaMDA in 2021, it has been developing steadily internally.

In June last year, a Google AI engineer claimed that LaMDA had human emotions, although Google quickly transferred him from his job, but after the matter was exposed by the media, it still caused a lot of controversy and speculation, which made Google cautious about AI security and reluctant to launch it in a hurry.

After the launch of ChatGPT last year, Jeff Dean, head of the AI project, said at an internal meeting that although Google has similar AI tools, the technology is not yet mature and that launching the product too hastily could damage Google's reputation because users have great confidence in the information provided by Google.

However, Google now has to speed up its research and development to catch up with competitors' products. This is not the rhythm that Google is used to. In December last year, Google upgraded the product level of generative AI to the highest red, coordinated the company's internal AI research and development resources, and accelerated the launch of its own response products.

In fact, Google is in a dilemma over the pace of Bard. If R & D is slow, the stock will be dragged down by concerns about Google's competitiveness and industry prospects, but if it is pushed too hastily, outsiders will be even more disappointed with Bard's flaws, driving down the stock price. This may be the main reason why Pichai was frequently vaccinated during the Bard public test.

The biggest advantage of real-time networking, how does the actual use of Bard perform?

Compared with ChatGPT,Bard, the biggest advantage is the huge information resources, they can rely on Google's search tools to get the latest information, according to their own comb to give the answer. Previously, ChatGPT's training data are only up to 2021, and it is impossible to answer information inquiries after 2022. In addition, the response speed of Bard is significantly faster than that of ChatGPT.

Bard can search for information online, which means users can directly ask about today's sports scores, stock price movements, and a series of news developments. However, in the actual test, I found that there is still a time lag in Bard, which is not yet fully real-time.

But compared to ChatGPT's training database, Google Bard's networking ability still has a clear advantage, and they can give their own answers to the latest events. When I asked former US President Donald Trump about the possible recent indictment in New York and the collapse of Silicon Valley banks, Bard gave a simple but complete summary of the events.

It is Google's job to ask Bard to find restaurants to help users provide flight hotel reservation information. But there are also information biases in Bard. When I asked Bard on Tuesday to book himself the shortest flight from San Jose, Calif., to Charlotte, North Carolina, Bard gave me a completely wrong flight number. When I asked Bard again on Thursday, I found that Google had corrected the mistake and gave the correct flight number.

ChatGPT supports third-party plug-ins to obtain real-time information, but today, OpenAI announced that it has approved a secure third-party plug-in to allow ChatGPT to access real-time information in a controlled environment, solving the biggest pain point that previously could not be connected to the Internet. The first batch of plug-ins are mainly from Expedia, Instacart, Shopity, Slack and other service platforms to help users get information about life and collaboration. This means that ChatGPT can also book meals and tickets, as well as retrieve company documents and notes.

The righteousness destroys the relatives and splits Google

Bard writes poems in vernacular to do math problems, tell a joke, and write a love poem. These dialogue operations are already a piece of cake for ChatGPT, and it is also easy for Bard here. Bard doesn't have much of a problem with these instructions and even offers different text options. Bard also allows users to copy text.

Bard answers how to comfort sick boys in terms of work assistance, Bard can also perform daily tasks, helping users write business emails and simple articles, as well as making forms, taking notes and doing analysis, but uploading files to Google Drive.

The problem of chicken and rabbit in the same cage is too simple for Bard. Although the Bard web end provides voice input, it cannot be used in practice. Although Bard claims to support multi-language translation, it indicates that it is unable to complete the instructions for the time being. All these details show that Bard is still a semi-finished product and needs further improvement.

Ask Bard about the film review of "the instantaneous Universe". Individuals don't know how to program and can't experience the difference in programming ability between ChatGPT and Bard. But from the feedback from programmers around me, ChatGPT seems to know more about their needs in this regard. "Bard does not give programmers the black background syntax highlighting they need. The quality of answers to questions in Python is also much worse than that in ChatGPT."

Bard has its own opinion output. A Google spokesman said after a public trial that large language models (LLM) such as Bard are trained based on public content, reflecting positive or negative views on specific politicians, celebrities and other public figures, and even giving views on certain aspects of controversial social or political issues in their answers.

However, Bard often does not show the source of information, so the accuracy needs to be judged by the user. On the other hand, users don't know how Bard came to a conclusion, and it's possible to be misled by Google's algorithm into believing its conclusion. Given the severe fragmentation of American society, this is likely to raise questions from conservatives.

Interestingly, when a tester asked Bard what he thought of the Justice Department's antitrust lawsuit against Google, Bard chose "righteousness to destroy relatives". Bard justifiably accused "Google of acquiring competitors and pressuring websites to use its tools to restrict its advertising transactions. These measures hurt consumers and stifled innovation in the digital advertising market. I hope the court will recognize the demand of the Justice Department and order Google to take measures to break the monopoly." It seems that Bard does not give special treatment to the parent company because of Google's products, and Google does not interfere with Bard on the most sensitive antitrust issues.

The escalation of the AI arms race perhaps, if Bard was a generative AI first released at the end of last year, then Bard's performance was enough to cause admiration and perhaps a wave of industry craze. But Bard was tested four months later than ChatGPT, and after many surprises from previous experiences with ChatGPT, it's really hard to feel fresh about Bard. Bard does not support other languages, pictures, programming, and users in the international market do not know when to use it.

Just before Google released the public beta version of Bard, OpenAI and Microsoft had already set off a new wave of AI arms race. Just a few days ago, OpenAI released a new generation of large-scale multimodal model GPT-4, which accepts image and text input, provides text output, and receives more data training (based on Microsoft Azure platform). It has more weight in the model file. In the words of OpenAI, GPT-4 has demonstrated human standards on a variety of professional and academic standards.

GPT-4 is also more creative and collaborative than previous versions, and its advanced reasoning capabilities surpass ChatGPT. According to OpenAI, GPT-4 has been arranged to take the American Bar qualification examination, the Law School entrance examination and the SAT Math and Writing in the National College entrance examination. In these tests, GPT-4 scored more than 88 per cent of candidates.

Not surprisingly, Microsoft launched a preview version of enterprise Azure OpenAI GPT-4 at the first time. Enterprises can build their own applications based on Azure OpenAI services. It only takes a few minutes for robot developers to build virtual assistants in natural language.

In this wave of generative AI arms race, OpenAI is undoubtedly the temporary leader, both in terms of research and development pace and intelligence, one step faster than Google, and let its strategic partner Microsoft see the dawn of catching up with Google. Since Bing joined ChatGPT in early February, Bing page visits have increased by 15.8%, app downloads have increased eightfold, and the number of daily active users has exceeded 100 million. At the same time, Google's search page visits fell by 1%, and search app downloads decreased by 2%.

The search market, which had been decided by the overall situation, has also been re-injected with competitive vitality because of the blessing of generative AI technology. However, this does not mean that Google's market position is about to shake. The number of daily active users of Google search is still more than 1 billion, more than 10 times that of Bing. If Google improves the user experience of Bard as soon as possible, opens comprehensive public testing to global users, and supports multiple languages, with a huge user base and data advantages, the future AI arms race is still uncertain.

The competitor in Bard's eyes finally I asked Bard who his main competitor was. Bard lists Microsoft's Xiaobing, OpenAI's GPT-3 and Amazon's Lex. " These models are trained based on a large amount of text and code, and can generate text, translate languages, write a variety of creative content, and answer users' questions. But I believe in my unique ability to get and process information from the real world through Google search, constantly keeping responses consistent with search results. This will enable you to answer your questions in a more comprehensive and informative manner. "

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.