Abstract: According to news on August 23, MediaTek has confirmed that it will provide device-side generative artificial intelligence functions. MediaTek revealed that it is working with Meta to better support Llama 2 LLM on phones equipped with the company's next-generation flagship processor (the Dimensity 9300, if nothing else).
MediaTek said: "MediaTek's next-generation flagship chipset, due to launch later this year, will feature a software stack optimized to run Llama 2, and an upgraded APU with Transformer backbone acceleration, reduced footprint access and DRAM bandwidth usage , further enhancing the performance of LLM and AIGC.” The company added that the phones are scheduled to be available by the end of 2023.
Image source: Generated by Unbounded AI
Hot info
NetEase Youdao Adds Multiple AI Services
Recently, Netease Youdao Information Technology (Beijing) Co., Ltd. has undergone industrial and commercial changes, and its business scope has added sales of teaching models and teaching aids, sales of artificial intelligence hardware, general application systems for artificial intelligence, public data platforms for artificial intelligence, development of basic software for artificial intelligence, Wearable smart device manufacturing, building intelligent system design, residential interior decoration, etc.
Nvidia will launch a new AI game graphics technology DLSS 3.5 to improve the game screen
It was reported on August 23 that Nvidia has launched multiple AI technologies to improve the game screen and production process, including the launch of DLSS 3.5 based on AI-enhanced light reconstruction technology this fall. AI trained by NVIDIA supercomputers can create higher-quality ray-traced images for intensive ray-traced games and applications, further improving the performance of games under ray tracing.
Gartner: AI chip market size will reach $67 billion in 2024
According to a report on August 23, Gartner, a market research agency, said that the market for chips used to perform artificial intelligence (AI) workloads is growing at a rate of more than 20% per year. Market analysts predict that the AI chip market will reach US$53.4 billion in 2023, an increase of 20.9% over 2022, and 25.6% in 2024, reaching US$67.1 billion. By 2027, AI chip revenue is expected to more than double the size of the market in 2023, reaching $119.4 billion.
Mercedes-Benz to Launch New Generative Art NFT Collection Mercedes-Benz NXT Icons
On August 23, Mercedes-Benz announced that it will launch a new generative art NFT series Mercedes-Benz NXT Icons. It is reported that the story behind the series will span seven eras, dating back to the beginning of automobile history. Each epoch will contain approximately 2,700 NFTs, 18,860 NFTs for all seven epochs, and NFTs for an epoch will be released every few months.
MediaTek confirms: With Dimensity 9300 processor, generative AI will run on mobile phones
On August 23, MediaTek has confirmed that it will provide device-side generative artificial intelligence functions. MediaTek revealed that it is working with Meta to better support Llama 2 LLM on phones equipped with the company's next-generation flagship processor (the Dimensity 9300, if nothing else).
MediaTek said: "MediaTek's next-generation flagship chipset, due to launch later this year, will feature a software stack optimized to run Llama 2, and an upgraded APU with Transformer backbone acceleration, reduced footprint access and DRAM bandwidth usage , further enhancing the performance of LLM and AIGC.” The company added that the phones are scheduled to be available by the end of 2023.
Kunlun Wanwei launched the first search engine in China that incorporates a large language model - Tiangong AI Search, and has opened internal testing applications
On August 23, Kunlun Wanwei announced the official launch of Tiangong AI Search, the first domestic search engine integrated with a large language model, and opened the application for internal testing.
According to the official introduction of Kunlun Wanwei, traditional search engines are based on keyword matching, that is, to identify the search scope according to the keywords entered by users, and match massive information that may meet the user's intentions. The AI search based on large model capabilities is a generative search. Users can clearly express their intentions through natural language and obtain effectively organized and refined answers. It is reported that. Tiangong AI Search's "Question" function allows users to conduct more than 20 rounds of interactions on a question, so as to conduct in-depth exploration. In addition, Tiangong AI search will soon have multi-modal search capabilities such as image and voice, which will release productivity to a greater extent and improve user efficiency.
Alibaba Cloud releases the Zhihai-Sanle Education Model based on the 7B model of Tongyi Qianwen
On August 22, Alibaba Cloud announced that it will cooperate with Zhejiang University and Higher Education Press to release the Zhihai-Sanle Education vertical large model based on Tongyi Qianwen 7B model training. This model has been launched on the Alibaba Cloud Lingji platform (DashScope) Service, developers only need one line of code to use it.
Microsoft: AI design tool Designer has been integrated into the Edge browser
According to news on August 23, Microsoft announced a new design tool in October 2022 called Designer, which uses generative AI technology to provide powerful functions. In April 2023, Designer launched a public preview version, and today Microsoft announced that Designer has become a fully integrated feature of the Edge browser, which is currently only available in the US market. According to the Microsoft support page, while Designer is currently in public preview and free to use, when it is officially released, certain features will require an additional fee to use. There's no word yet on when Designer will be available in Edge outside of the US.
Li Yanhong: It is expected that there will be a new version of Wenxin Yiyan at the end of the year
According to a report on August 22, Baidu founder Robin Li said in a earnings call that a new version of Wenxin Yiyan is expected to be launched by the end of the year. During the transition period before the launch of the new version, Baidu will promote the use of the current version and explore applications and scenarios for different industries. "Industry-specific models should be built on top of a strong foundational model, otherwise they are groundless. Over time, rapidly evolving foundational models will provide stronger support for industry-specific models. And those who can do foundational models There are only a few enterprises." Robin Li said.
OpenAI introduces GPT-3.5 Turbo fine-tuning and updates API
On August 23, OpenAI announced the launch of the GPT-3.5 Turbo fine-tuning function and updated API, enabling enterprises and developers to use their own data to build exclusive ChatGPT combined with business use cases. OpenAI said that from the early test data, the fine-tuned version of GPT-3.5 Turbo can match or even exceed the basic functions of GPT-4 on certain tasks. Additionally, fine-tuning capabilities for GPT-4 will be available this fall.
Investment and Financing
"Shuimu Molecule", a developer of large-scale biomedical basic models, has completed tens of millions of seed round financing
Recently, Shuimu Molecule completed tens of millions of seed round financing. Shuimu Molecule is a developer of large-scale biomedical basic models, dedicated to building basic large-scale models of the biomedical industry and a new generation of conversational biomedical research and development assistants. The company's products will serve all aspects of drug research and development, including early research projects, target discovery, molecular design optimization, clinical trial design, drug repositioning, etc.
Ideogram AI closes $16.5M seed round led by a16z and Index Ventures
On August 23, Ideogram AI, a generative AI tool, announced the completion of a $16.5 million seed round of financing, led by a16z and Index Ventures, with participation from Raquel Urtasun, Jeff Dean, Andrej Karpathy, AIX Ventures, and Golden Ventures.
Metaverse recommended reading
"Bain & Company: By 2030, the Global Metaverse Industry Will Reach $900 Billion"
According to a Bain & Company report, the metaverse industry will be worth $900 billion by 2030.
Though Mark Zuckerberg's fall 2021 "Meta" metaverse project -- complete with clumsy avatars and a half-baked ecosystem -- is pretty interesting. But Bane believes that the metaverse will not die.
https://www.8btc.com/article/6830594
"Focus on AIGC is not enough, you also need AI PC"
AI empowered PC? This prophecy may seem unremarkable, but it may be the closest to us and the fastest to see the timeline.
In addition to Nvidia, Intel also quickly announced that it will go all out to attack AI PCs. Intel CEO Pat Gelsinger said, "We believe that AI PCs are a key turning point in the PC market in the next few years, and we will integrate AI into every product we make."
Terminal manufacturers such as Lenovo, HP, and Dell will ultimately bring AI PCs to consumers. For more than ten years, Lenovo Group's status as the global PC hegemony has not been shaken, followed by HP and Dell. This seems to be a traditional and stable business.
https://www.8btc.com/article/6830568
"DingTalk AI, Break the Inner Scroll"
The bigger change will come in 2023. DingTalk announced that it will be connected to the Tongyi large model, and DingTalk will begin to have the ability to output AI. From the slash function in June, to the AI PaaS capability released on August 22, and today's concrete AI applications and functions such as digital employees, AI assistants, and happy spells have emerged.
The development path of DingTalk is gradually becoming clear. In the tide of the AI era, can DingTalk become the biggest variable?
https://www.8btc.com/article/6830562
"OpenAI launches GPT-3.5 Turbo fine-tuning version, users can have exclusive models"
Developers can now use their own data to tailor GPT-3.5 Turbo to their own use cases.
A fine-tuned version of GPT-3.5 Turbo is available now, and a fine-tuned version of GPT-4 will be available this fall. This update enables developers to customize models that better suit their use cases and run those custom models at scale. Early tests have shown that fine-tuned versions of GPT-3.5 Turbo can match or even exceed base GPT-4-level capabilities on certain vertical application tasks, which means that GPT-3.5-Turbo can now be as efficient as GPT-4, but at a higher cost. much lower.
https://www.8btc.com/article/6830554
Babbitt Park is open for cooperation!
Chinese Twitter: https://twitter.com/8BTC_OFFICIAL
English Twitter: https://twitter.com/btcinchina
Discord community: https://discord.gg/defidao
Telegram channel: https://t.me/Mute_8btc
Telegram community: https://t.me/news_8btc
