Babbitt | Metaverse Daily Must-Read: AI has generated more pictures in the past 18 months than human photographers have taken in 150 years

Abstract: According to the statistical report published by Everypixel Journal, since 2022, in the past 18 months, artificial intelligence has generated 15 billion pictures, which is equivalent to the amount taken by human photographers in 150 years. The current CEO of Microsoft, Satya Nadella (Satya Nadella), believes that AI technology will detonate a new round of technological revolution, and its impact is equivalent to the meaning of the Internet wave mentioned in Bill's memo in 1995.

Image source: Generated by Unbounded AI

Hot info:

Report: AI has generated 15 billion images in the past 18 months, 80% from open source models

According to the news from IT House on August 18, according to the statistical report published by Everypixel Journal, since 2022, in the past 18 months, artificial intelligence has generated 15 billion pictures, equivalent to 150 years taken by human photographers. quantity. The report states that of the 15 billion images generated, 80% of them come from open source models, and that people use AI generation tools to create 34 million images every day.

According to the report, the most popular build tool is DALL-E 2. Adobe Firefly is the fastest growing, generating 1 billion images in just 3 months of launch. While many AI images come from open-source models, paid generative AI tools are also very popular. For example, Midjourney has the largest user community of over 15 million.

Microsoft plans to cooperate with Databricks to launch AI services, which may pose a threat to OpenAI

Microsoft plans to partner with Databricks to launch a new version of software that will help customers develop artificial intelligence (AI) applications for their businesses, The Information reported, citing The Information. This move may pose a threat to OpenAI, because OpenAI is also a company that provides AI services. Microsoft hopes to cooperate with Databricks to provide more powerful AI solutions and attract more customers.

Lenovo released two AI computing power servers, respectively for large model training and reasoning

According to the "Kechuangban Daily" report on August 18, at the 2023 China Computing Power Conference held today, Lenovo Group released its AI strategy and officially launched two AI computing power products: AI large model training server WA7780 G3 and AI training push All-in-one server WA5480 G3. It is understood that these two products now have orders and are ready to deliver applications.

ZTE has built a GPU computing power cluster at Nanjing Binjiang Base to conduct large-scale model training with tens of billions of parameters

According to a Shanghai Securities News report on August 18, ZTE is building an end-to-end AI comprehensive solution and independent large-scale model capabilities. At present, ZTE has built a GPU computing power cluster at the Nanjing Binjiang base, and is conducting large-scale model training with tens of billions of parameters. At the same time, ZTE is building a higher-performance GPU cluster, which will start large-scale model training with hundreds of billions of parameters after the construction is completed. The large model trained by the company will be used in conversational AI-assisted programming IDE, code generation, test case generation, and telecom intelligent operation and maintenance.

Former Google Brain top researcher founded artificial intelligence research company Sakana AI in Tokyo

According to a Reuters report on August 17, former Google Brain top researchers David Ha and Llion Jones announced the establishment of artificial intelligence research company Sakana AI in Tokyo. The company declined to disclose financing details. It is reported that Jones is the fifth author of Google's 2017 research paper "Attention Is All You Need", which introduced the "transformer" deep learning architecture, which later became the basis of ChatGPT, and is also the current development driven by generative artificial intelligence. basis of the product. Ha was a research director at Stability AI and a Google Brain researcher.

China Unicom plans to launch four types of industry models

According to Jinshi’s news on August 18, the reporter learned from the 2023 China Computing Power Industry Conference that China Unicom built an industry model for production and operation based on an open source general model. China Unicom intends to launch four large-scale industry models. The first is the communication anti-fraud model, the second is the network model, the third is the customer service and operation and maintenance model, and the fourth is the industry segmentation model.

ByteDance responded to the launch of the AI dialogue product "Doubao": it is still a small-scale invitation-based test, and there are many limitations

According to a report by Meijing.com on August 18, ByteDance has begun external testing of the AI dialogue product "Doubao". Currently, the relevant app can be downloaded through its official website, and it supports Douyin login. Doubao App is also an AI project previously code-named "Grace" in Byte. On August 18, a member of the "Doubao" project team responded to the reporter: "'Doubao' is a chat robot product, which is still in the early stage of development and verification. 'There are still many limitations, and the generated content may also be inaccurate. Test users are welcome to provide feedback to help us continue to improve the product."

NVIDIA H100 GPU sells for almost 10 times the cost

According to a report by "Kechuangban Daily" on August 18, Tae Kim, a senior writer at Barron's, quoted investment bank Raymond James data as saying that the cost of Nvidia to manufacture an H100 GPU is $3,320, while the price given to customers is $25,000- $30,000.

Microsoft CEO: The impact of artificial intelligence is as far-reaching as the Internet wave

According to the Golden Ten report, Microsoft's current CEO Satya Nadella (Satya Nadella) believes that AI technology will detonate a new round of technological revolution. Nadella said in the interview, "The far-reaching impact of artificial intelligence is equivalent to the significance of the Internet wave mentioned in the 1995 Bill memo."

Station B Chen Rui: Internally attaches great importance to AIGC, and has applied large-scale models to its business

On the evening of August 17, Bilibili announced its unaudited financial report for the second quarter of 2023. In this quarter’s conference call, Chen Rui, chairman and CEO of Bilibili, focused on AIGC’s empowerment of Bilibili. "Although we have not announced our large language model and have not participated in the 'Hundred Models War', our large language model has actually been useful in our business." Chen Rui said, for example, that station B has already begun to use AIGC technology is used to improve user experience and business capabilities. For example, in content security review, a large language model can replace a large amount of human investment in the past, and greatly improve the efficiency of review. Chen Rui said that the company attaches great importance to AIGC internally, "This is a major opportunity for a content platform like Station B. If we can effectively use this technology, it can improve the experience of users and UP owners, and improve the efficiency of business transformation. "

Metaverse recommended reading:

** "The hottest big language models all love "nonsense", whose "illusion" problem is the worst? "**

"AI hallucinations" (hallucinations) refer to chatbots completely fabricating information and appearing to spout facts in response to user prompt questions. OpenAI acknowledged in the report that "even the most advanced AI models are prone to lie generation, and they exhibit a tendency to fabricate facts in moments of uncertainty. These illusions are especially problematic in domains that require multi-step reasoning, where a logical error That's enough to undermine a larger solution."

40 million downloads in two weeks! There is Miaoya in the front and Remini in the back. Why are AI mapping applications so popular? "

"Practitioners who are really engaged in AI mapping look down on AI. The current AI image generation technology still has certain limitations and can only be used as an auxiliary tool. But this 'limitation' can already meet the needs of most players. Many experts In rejecting AI, in fact, laymen are more able to find some new business opportunities from it.”

** "Urgent Reminder! The overlord clause behind the AI is shocking!》**

No matter how technology develops, user data should not be used or sold openly. In the long-term speculation, instead of asking "are you willing to sacrifice some privacy for convenience", it is better to be vigilant from the source. A society that must choose between privacy and convenience should not be the society we pursue.

View Original
  • Reward
  • Comment
  • Share
Comment
No comments