Weekly AI Highlights Review: December 17–25

As 2024 comes to a close, the lesson seems to be that it is never too late to raise more funding. This week, four companies announced they had closed their respective funding rounds right on time for Christmas. First, data, analytics, and AI company Databricks secured an impressive $10 billion Series J from notable investors including Thrive Capital and co-leads Andreessen Horowitz, DST Global, GIC, Insight Partners, and WCM Investment Management. In addition to sharing the news about its new funding, Databricks also shared that it expects to reach positive free cash flow and exceed $3 billion run-rate revenue by January 31, 2025, after reaching 60% growth year-over-year during Q3, 2024.

Then, three startups working on promising applications of generative AI announced their respective funding rounds. Decart, a relatively young startup that made enough money from AI infrastructure to pivot into developing 'open-world' models, closed its $32 million Series A led by Benchmark less than two months after raising $21 million in seed funding in a round led by Sequoia and Zeev Ventures. Decart gained notoriety after releasing a technical demo of Oasis AI, a model capable of generating "playable" Minecraft-like worlds. World models like Decart's are shaping to be the next frontier for generative AI in parallel with 'agentic' AI.

The remaining two startups are arguably leaders in their spaces, which also happen to be hot topics as some of the most profitable generative AI use cases of late: search and coding assistance. Anysphere, the parent company of the popular AI-powered coding assistant Cursor, recently made the news for its acquisition of rival coding assistant Supermaven and for unintentionally sparking a bidding war among several investors due to experiencing explosive growth that took it from $4 million annualized recurring revenue (ARR) in April to $4 million monthly in October. A Series B funding round led by Thrive Capital resulted from those unexpected negotiations, in which Anysphere raised $100 million at an estimated $2.6B valuation.

Lastly, reports emerged that Perplexity AI has finalized its fourth funding round this year, raising $500 million at a $9 billion post-money valuation. With this investment, Perplexity effectively increased its valuation 18x, from about $500 million to $9 billion, in just one year. The unarguable leader of AI-powered search, Perplexity will soon face increased competition, especially as OpenAI and Google strengthen their respective AI-powered search offerings. Perplexity also announced it has acquired Carbon, a startup specializing in data connectivity, intending to integrate Carbon's technology into its products so they can connect directly to enterprise apps like Notion, Google Docs, and Slack.

Other notable headlines this week

Google managed to cast a shadow that's difficult to ignore over OpenAI's recent press engagement. Just as the latter was rounding up its twelve days of announcements with the o3 preview, Google was relentless in its announcements, dropping a 'thinking mode' for its experimental model Gemini 2.0 Flash and Veo 2, a video generation model capable of producing consistent minutes-long video clips in resolutions up to 4K, theoretically blowing out of the water not only Sora but many other rivals. Veo 2 is exclusively available through VideoFX and is limited to creating 8-second samples at a 720p resolution. Google plans to expand access to Video FX and eventually bring Veo 2 to power other experiences like YouTube Shorts.

Less than three weeks left to apply to Mistralship—Mistral AI's startup program: Mistral AI recently started receiving applications for its startup program, Mistralship. Ten selected applicants will benefit from La Plateforme credits, one-on-one support, and early access to models and products in the first six-month cohort. Applications close on January 5, 2025.

YouTube's new setting lets creators opt their content into third-party AI model training: YouTube has launched a new setting that opts creators out of third-party model training and enables them to choose whether they want all, some, or no third parties to train models using their content.

Writer's Palmyra Creative aims to add depth, variety, and originality to its outputs: Writer has launched Palmyra Creative, a new AI model designed to produce more creative and engaging outputs. According to Writer, Palmyra Creative can be combined with the company's specialized models to bring creative writing to domains such as finance and healthcare.

AI search monitoring platform Otterly.AI emerges from stealth with 1000+ customers: Otterly.AI, the first AI search monitoring platform, has launched from stealth with over 1,000 users, helping brands track and optimize their visibility across major AI search engines like ChatGPT and Google AI Overviews as traditional search traffic faces disruption from AI adoption.

YouTube partners with CAA to launch AI likeness protection tools: YouTube has partnered with the Creative Artists Agency to develop and test new AI tools that will help celebrities and creators identify and manage AI-generated content featuring their likenesses on the platform.

Grammarly acquires Coda to transform its AI assistant into a productivity platform: Grammarly has acquired Coda to create an AI-powered productivity platform that combines Grammarly's writing assistance technology with Coda's document and workflow tools, with Coda's co-founder Shishir Mehrotra taking over as CEO of the combined company.

GitHub celebrates reaching 150M developers with a Copilot free tier: GitHub has launched a free version of its AI coding assistant Copilot with limited monthly completions and chat messages, aiming to expand access to AI programming tools as part of its mission to reach one billion developers worldwide.

Instagram plans to launch more AI-powered features for creators next year: AI-powered video editing features are coming to Instagram. These tools will be based on Meta's recently previewed Movie Gen model suite. According to the announcement, the upcoming features will enable creators to make edits like changing their outfits or modifying the background of the video.