[November 17th Big Model Daily] Information Sam Altman Speech at the Cambridge University Debate Club｜In the future, the most important skill is to learn how to learn and how to generate new ideas!

Sam Altman Speech at Cambridge University Debating Club｜In the future, the most important skill is to learn how to learn and how to generate new ideas!

https://news.miracleplus.com/share_link/11783

During the British AI Safety Summit earlier this month, Sam Altman accepted the 2023 Hawking Scholarship Award on behalf of the OpenAI team at Cambridge Union. Participants included classmates from Cambridge Union and the Hawking family. Sam shared:

– What differentiates OpenAI from other AI companies? Different from Musk

– Progress in AGI so far, what will OpenAI look like in 10 years?

– How do today’s graduates adapt to this change?

– What did the British government do wrong?

– What are the qualities of a great founder?

– How to avoid losing critical thinking?

– When will OpenAI launch hardware?

– How far are we from a model of consciousness?

– What are your thoughts on the open source model? Thoughts on smaller models?

ChatGPT was exposed to test new features: learn all historical chats, reset memory, and “burn after reading”

https://news.miracleplus.com/share_link/11784

ChatGPT may be getting a major new feature. That is to remember your previous chat content and continue to learn from it (not simply saving history), so as to understand your preferences and information, which can be used in future chats. For example, you can give it a command to “always write code in python”, and then all the code you ask it to write will be in this language. When you tell it “I’m going to xx in two weeks,” and maybe you ask it some food questions, it will directly recommend that destination. More than one netizen has discovered that his ChatGPT has been quietly testing this function.

Cambridge Dictionary’s word of the year is out: The advent of the AI era is unstoppable

https://news.miracleplus.com/share_link/11785

If you ask which industry will have strong development momentum in 2023, most people will first think of the AI industry. From the beginning of the year to the end of the year, the development of AI has never stopped. The large-scale language model GPT-4 has strong development momentum, and OpenAI once again launched the latest artificial intelligence model-GPT-4 Turbo and GPTs on the 6th local time. In addition, various artificial intelligence companies are rushing to launch chat robots. Microsoft’s Bing and Musk’s “Grok, the best chat robot” all indicate that the artificial intelligence war has reached a fierce stage. This phenomenon has even affected other fields. On the 15th, Cambridge Dictionary announced that the word of the year for 2023 is “Hallucinate”. The original definition of this word is “seeming to see, hear, feel or smell something that does not exist, usually because of a medical condition or because you have taken something.” “Hallucinate” causes hallucinations due to “drugs”, but in the era of AI, “Hallucinate” has an additional meaning, and a new note is “When AI creates hallucinations, it will generate error messages.” The Cambridge Dictionary website explains that “Hallucinate” was chosen as word of the year because the new meaning “gets to the heart of what people are talking about about AI.”

Changing the future of music creation

https://news.miracleplus.com/share_link/11786

Google Deepmind announces state-of-the-art music generation model, as well as two new artificial intelligence experiments aimed at opening up a new playground for creativity. From jazz to heavy metal, electronica to opera, music is a beloved form of creative expression. With lyrics, melodies, rhythms, and vocals complex and dense, creating compelling music has been especially challenging for artificial intelligence (AI) systems—until now. Google Deepmind announced a partnership with YouTube to officially launch Lyria, the most advanced artificial intelligence music generation model to date, and two artificial intelligence experiments designed to open up new frontiers for creativity: 1. Dream Track – one of the YouTube short films An experiment designed to deepen the connection between artists, creators and fans through music creation. 2. Music AI Tools – We are working with artists, songwriters and producers to design a suite of tools to help them enhance their creative process.

Microsoft confirms it will bring Copilot chatbot to Windows 10

https://news.miracleplus.com/share_link/11787

Microsoft recently said that it will soon begin testing the Copilot chatbot in an update to the widely used Windows 10 operating system. It plans to provide Copilot in the preview channel to PCs running Windows 10 version 22H2, and will only be available for Home and Professional editions. version of the Win10 operating system.

Meta launches AI-based video editing tool

https://news.miracleplus.com/share_link/11788

On November 16, local time, Meta launched two new video editing features based on artificial intelligence, which can be used to post videos on Instagram or Facebook. The first feature, called “Emu Video,” can generate videos based on text-only, image-only, or combined text and image input. Another one, called “Emu Edit,” allows users to perform free-form editing through text commands.

China will become the leader in AI field! Google CEO: China and the United States should cooperate to regulate innovation

https://news.miracleplus.com/share_link/11789

Alphabet Inc. Chief Executive Sundar Pichai said he expects China to be at the “leading edge” of artificial intelligence and said the United States needs to work with China on regulation and innovation soon. “The scale of China’s AI research is staggering,” he said in an interview at the Asia-Pacific Economic Cooperation (APEC) conference in San Francisco. “China will be at the forefront of AI, that’s for sure.

Google will delay the release of Gemini Al, expected to launch in the first quarter of next year

https://news.miracleplus.com/share_link/11790

According to The lnformation, two people familiar with the matter revealed that Google will delay the release of its OpenAl competitor Gemini Al. Gemini won’t be available until the first quarter of next year, people familiar with the matter said. Earlier this year, Google representatives told some cloud customers and business partners that they would launch Gemini, the company’s new conversational AI, a large language model, by November.

Sequoia Partners: Artificial intelligence will bring one billion more developers to the world, comparable to the value of AGI

https://news.miracleplus.com/share_link/11791

David Cahn, a partner at Sequoia, a well-known venture capital institution, recently issued an article stating that he believes that the number of developers will continue to grow rapidly and that artificial intelligence will give birth to 1 billion developers. The introduction of these 1 billion developers into the software industry can rival AGI in terms of economic impact. . If we can make software engineering more widely available, aided and accelerated by AI, it could help automate many non-tech industries and create an economic surplus.

Unity launches beta version of game generation AI platform Muse, priced at US$30 per month

https://news.miracleplus.com/share_link/11792

Unity announced that the Muse AI platform has officially launched an early access version for public testing, with a paid version costing $30 per month. Originally launched in July, Muse is a set of AI-driven tools that allow users to obtain usable game code by interacting with a chatbot, aiming to streamline the game development process.

Stability AI’s head of audio team resigns from the company, questioning copyright justification for content training

https://news.miracleplus.com/share_link/11793

Ed Newton-Rex, head of Stability Al’s audio team, has announced his resignation from Stability AI, questioning the company’s assertion that using copyrighted content in artificial intelligence training is “fair use.”

There is a major bug in the ChatGPT web version. Modify the URL to “upgrade” GPT-3.5 to 4.0 with one click

https://news.miracleplus.com/share_link/11794

According to feedback from multiple users on the X platform, a major vulnerability has recently appeared in the ChatGPT web version. The GPT-3.5 version can be “upgraded” to 4.0 with just one click by modifying the URL. It is reported that users only need to enter the official chat page of ChatGPT and add “/?model=gpt-4-gizmo” after the URL to use GPT-4.0. Apparently, there is a serious “temporary worker” level bug in the authentication system. .

Honor will launch a self-developed large-scale device-side AI model with 7 billion parameters and new cloud services

https://news.miracleplus.com/share_link/11795

Honor CEO Zhao Ming released a speech on the third anniversary of the founding of the new Honor. He said that Honor’s current growth rate in overseas markets exceeds 200% and has achieved profitable growth within two years. In addition, Honor will soon launch a self-developed 7 billion parameter end-side AI large model and a new cloud service.

Baichuan Intelligent cooperates with Pengcheng Laboratory to break through the long window technology of domestic large-scale computing power model

https://news.miracleplus.com/share_link/11796

On November 16, Baichuan Intelligence and Pengcheng Laboratory announced that they would work together to explore large model training and applications, and jointly develop a 128K long window large model “Pengcheng-Baichuan Brain 33B” based on domestic computing power. This is a practice of innovation and implementation of large-scale domestic computing power models, and has a positive demonstration effect on the development of large-scale domestic computing power models.

TACO-LLM released! Helping large models achieve extreme acceleration

https://news.miracleplus.com/share_link/11797

TACO-LLM is a large language model inference acceleration engine launched based on Tencent Cloud’s heterogeneous computing products. It is used to improve the inference performance of language models. By making full use of the parallel computing capabilities of computing resources, TACO-LLM can handle more language model inference requests at the same time, and has successfully provided customers with an optimization solution that takes into account high throughput and low latency, with throughput performance increased by 78%. TACO-LLM can reduce the waiting time for generating results, improve the efficiency of the inference process, and help users optimize business costs.

More than 80% of college students surveyed have used AI tools – 77.51% of respondents are optimistic that AI tools can help improve efficiency, and 59.65% are worried that using AI tools has certain risks.

https://news.miracleplus.com/share_link/11798

With the continuous development of artificial intelligence technology, AI tools have begun to enter the daily lives of ordinary people. Recently, China Youth Daily and China Youth Daily launched a questionnaire survey on the use of AI tools among college students across the country, and a total of 7,055 valid questionnaires were collected. The survey results show that 84.88% of the respondents have used AI tools, of which 16.30% often use AI tools, 57.49% use them occasionally, and 19.43% rarely use them.