#Fast Eight Talk##Make AI within reach##Google#"Sleepless night, the world is changing too fast", after browsing 24 hours of AI news, the media and readers were surprised.
On March 22, Google opened the chatbot Bard test, with practical actions to respond to Baidu's "Xia Mawei" a week ago, Microsoft updated Bing and Edge browser three major functions, outstanding graphics and text generation capabilities, like in the call board Baidu's multimodal generation. The second-generation post-production auxiliary tool Runway made a sentence generate blockbusters, design creative agency Adobe came up with a generative AI model set, and the AIGC field was also involving.
From text to pictures to creativity, generative AI allows the public to experience a dimensionality reduction blow to applications in the AI era, which seems to endorse the "AI's 'iPhone' moment", and also adds another firewood to the already white-hot AI melee at home and abroad.
Compete against each other
On March 16, Baidu CEO Robin Li "Amway" generative AI product Wen Xin said that Wen Xin Yiyan is the world's first technology factory officially released a benchmark ChatGPT product, Microsoft just called OpenAI, Google, Meta (formerly Facebook), Amazon, etc. have not made similar products.
OpenAI and Baidu showed off their muscles on those two days, and it wasn't until March 22, a week later, that Google belatedly opened Bard's access in the UK and the US, but only in English. Such caution may have something to do with Bard's mistake a month ago, which caused Google's market value to lose more than 700 billion yuan overnight. This time, Google did not open the closed beta on a large scale, but gradually released Bard's test quota, and the initial version only responded to text.
In line with Microsoft's new Bing usage patterns and functions, Bard also provides users with a separate chat interface where users can ask questions and search in text boxes, such as asking Bard to help you write a plan, list a recipe, search for a trivia, etc.
After Bard generated the content, Google placed four buttons: Like, Dislike, Regenerate, and Google it. This small detail allows users to seamlessly switch from Bard's chat interface to Google's search interface, figuring Google's positioning of Bard as "search complement", echoing Bard's summary of himself as "your creative collaborator".
The March 22 update of Microsoft's Bing and Edge browsers with three new features also makes search full of imagination. New features include Bing Image Creator, AI-powered stories, and Knowledge Cards 2.0, most notably Bing Image Creator, which can generate images. Bing Image Creator can generate images based on the user's description, which is similar to the image generation function provided by Wen Xin Yiyan.
By showing the latest technology, can you judge the superiority of Google and Microsoft in the field of AI? Wang Dongwei, director of Zhongtai Capital, believes that "at present, the two companies have focused on the research and application of AI technology in different fields. Microsoft excels in natural language processing and language models, and Google leads in computer vision and image recognition."
Dogfight at home and abroad
Large manufacturers have appeared in turn, and AI involution in vertical fields is also carried out at the same time.
Looking at overseas first, Runway's Gen 1 model, which can realize text-generated moving images, was upgraded to Gen 2 on March 22, which is smoother and higher-definition than Gen 1 and Gen 2 generated video. Design and creative agency Adobe showed off its content generation technology: Adobe Firefly, a set of creative generative AI models.
Coincidentally, NetEase Yaotai demonstrated the AIGC (artificial intelligence generated content) ability of a sentence change on the same day. "NetEase Yaotai can generate different styles of clothing with one click through text description, and users can freely customize the clothing of the virtual image by entering text, and the relevant graphic and text generation technology used behind it comes from the technology of NetEase Fuxi artificial intelligence research. NetEase Fuxi has built a multimodal understanding basic model that can better understand Chinese users based on self-built datasets, and uses this as a basis to realize open generative model training for metaverse scenarios," Mu Bai, head of NetEase Yaotai, told Beijing Business Daily.
More representative is the Wen Xin Yiyan, which launched the internal beta on March 16, this product has comprehensive capabilities in literary creation, commercial copywriting, mathematical calculation, Chinese understanding, multimodal generation, etc., such as summarizing the core content of "Three-Body Problem", suggesting continuation angles, naming the company, writing Slogan, editing press releases, and answering logical reasoning such as "chicken and rabbit in the same cage". Wen Xin also understands the meaning of "Luoyang paper expensive", can explain the economic theory corresponding to the event, and can compose Tibetan poems according to idioms. Generate Sichuan dialect version of speech, and produce videos to reflect its ability to generate in multimodality.
Eighty percent of jobs may be subverted
In the generative AI track, the display, upgrade, and release of various technologies and applications are almost measured in days, which has once again triggered people's thinking.
"AI's 'iPhone moment' has arrived," Nvidia CEO Jensen Huang repeated three times when releasing his latest graphics card on March 22. Another point he makes is more popular, to the effect that generative AI is a new type of computer that we can program in human language and anyone can command a computer to solve a problem.
Going back to the question that is asked every time a generative AI product goes live, whose profession will AI replace? OpenAI's recent answer is that GPTs are general-purpose technologies, affecting 80 percent of American jobs, affecting at least 10 percent of their tasks, and 50 percent of their jobs being automated by AI to some degree in about 19 percent of jobs, and that impact spans almost all industries. According to the paper published by OpenAI, there are 34 types of occupations that are not affected by AI, including dishwashers, plumbers, green chemicals and other blue-collar workers.
Is that really the case? Or can current AI products replace human work? A reporter from Beijing Business Daily has tested ChatGPT and Wen Xin's words.
In response to the reporter's demand to "write a live broadcast copy for an anti-aging cream for the workplace", ChatGPT wrote, "This cream uses the latest anti-aging technology, including a variety of natural plant ingredients and active peptides, which can quickly penetrate into the deeper layers of the skin, promote collagen synthesis, enhance skin elasticity and radiance, so as to achieve anti-aging effects", Wen Xin Yiyan provided the content as follows, "This cream uses the most advanced technology, By activating the regenerative mechanism of skin cells, it promotes the metabolism of the skin, making the skin firmer, smoother and more hydrated".
In this regard, Wang Tuo, the anchor of the live broadcast room, commented, "If we are in the live broadcast room, these three oral scripts need to be rewritten. From the perspective of compliance, creams belong to ordinary cosmetics, and this product used for health care cannot promote efficacy when live streaming. Secondly, there must not be a large number of adjectives and misleading descriptions in the oral broadcast script, which is easy to mislead consumers and harm consumers' rights, which is strictly prohibited in the live broadcast room."
There is no doubt that AI is in a high-speed development stage, how much impact will generative AI have on human life? Will global AI computing double, as OpenAI CEO Sam Altman predicts, every 18 months? Might as well let the bullets fly a little longer.
Beijing Business Daily reporter Wei Wei