AI video takes flight

AI Creators are making incredible short videos, Musk aims to release the Grok chatbot on X

Author

Tara Tan
November 30, 2023

Happy Thursday.

Between Stable Video Diffusion, Pika Labs, and Runway, AI generated videos are getting pretty darn good. Check out a great thread on recent AI videos.

In other AI news, there is a flurry of deals being closed before the end of the year, with avatar maker HeyGen closing a $5.6M raise (and hit $18M ARR within a year) and video generation startup (with just 4 employees) Pika Labs raising a whopping $55M.

As a recovering filmmaker, I’m excited about the democratization of storytelling tools. Right now, they’re mostly in the hands of early tech adopters, but I think its true impact will be seen when the filmmakers, storytellers, and artists get to wield it as a tool to create stories and montages that move us.

The Long Read

Bill Gates interviews AI professor Dr Yejun Choi on the difference between human and machine intelligence. But because it relies on patterns versus abstracting the concept the way you and I do, it can actually be a bottleneck for the technology and hurt the way it’s learning.

Episode 5: Yejin Choi

Listen to this episode from Unconfuse Me with Bill Gates on Spotify.

🔥 Latest news

  • Major world powers unite to sign agreement to advance AI safety and security. The United States, Britain, Australia, and over a dozen other countries unveiled an agreement on Nov. 26 containing guidelines for keeping AI safe and secure. The agreement lays out general, non-binding recommendations around securing AI systems, protecting data, and vetting suppliers. However, it does not address issues like appropriate AI use cases or data collection and privacy policies. https://tinyurl.com/3f858cur 

  • Google’s Bard AI chatbot can now answer questions about YouTube videos.  Google’s Bard AI chatbot has elevated its capabilities by enabling users to pose questions related to YouTube videos. While Bard initially gained the ability to analyze YouTube videos with the introduction of the YouTube Extension in September, it can now provide specific answers regarding the content of a video. https://tinyurl.com/46dymvv3 

  • Inflection-2 beats Google’s PaLM 2 across common benchmarks. Inflection, an AI startup aiming to create “personal AI for everyone”, has announced a new LLM dubbed Inflection-2 that beats Google’s PaLM 2. Inflection-2 was trained on over 5,000 NVIDIA GPUs to reach 1.025 quadrillion floating point operations (FLOPs), putting it in the same league as PaLM 2 Large. However, early benchmarks show Inflection-2 outperforming Google’s model on tests of reasoning ability, factual knowledge, and stylistic prowess. https://tinyurl.com/bddbtb5b 

  • AWS unveils new Trainium AI chip and Graviton 4, extends Nvidia partnership. At its annual AWS re:Invent developer conference in Las Vegas, Amazon announced a new version of Trainium 2, its dedicated chip for training neural networks. Trainium 2 is tuned specifically for training so-called LLMs and foundation models -- the kinds of generative AI programs such as OpenAI's GPT-4. The company also unveiled a new version of its custom microprocessor, Graviton 4, and said it is extending its partnership with Nvidia to run Nvidia's most advanced chips (GH200 Grace Hopper multi-chip) in its cloud computing service. https://tinyurl.com/32sktbuy 

  • Elon Musk’s xAI will launch ‘Grok’ chatbot this week: Here’s what to expect. xAI, the company Elon Musk founded to take on OpenAI and other startups racing in the elusive AI space, is set to launch its first offering ‘Grok’ this week. He announced that the AI chatbot will be integrated into the X platform (formerly Twitter) as part of an early access program. While many details remain under wraps, the X posts shared by Musk and his team at xAI indicate that Grok will be a ChatGPT competitor, which will be able to engage in back-and-forth conversations when prompted. It is expected to handle all sorts of queries from users, right from mathematical problems to code challenges. https://tinyurl.com/4bzsss4h 

🔧 Cool Tools and Experiments

  • Swiss startup vidby introduces Vidby Call Translator for Google Meet, offering real-time translation in 150+ languages https://tinyurl.com/2p8rwzwd 

  • Screenshot-to-Code: A tool that converts a screenshot to HTML/Tailwind CSS, uses GPT-4 Vision to generate the code and DALL-E 3 to generate similar-looking images https://tinyurl.com/thdvz8we 

  • Google Meet adds gesture detection to recognize when users raise their hand on a group video call and alerts the host https://tinyurl.com/upakx7w2 

💰 Latest Startup Fundings

  • Hololight, a software maker focused on AR and VR enterprise solutions, secures $12M in funding for the development of its XR streaming platform, Hololight Hub https://tinyurl.com/45xurpw8