• AI Rundown
  • Posts
  • The AI Rundown by Lightscape Partners - 06/03/24

The AI Rundown by Lightscape Partners - 06/03/24

A new report on AI investing trends, China adds billions to its semiconductor fund, and Klarna reports millions saved from generative AI.

BAILEY/AGENCE FRANCE-PRESSE/GETTY IMAGES

Good morning, happy June, and welcome back to The AI Rundown by Lightscape Partners.

Last week in AI:

  • A new report from PitchBook shows early-stage venture funding has dropped 76% from its peak in 2023. The shift in funding comes as many startups now focus on bringing products to the market and investors brace for the rubber to hit the road.

  • China invested $48 billion more in its semiconductor fund. The investment illustrates the growing chip war and follows tariffs implemented by the US recently.

  • Klarna claims generative AI has cut marketing costs by $10 million annually. Savings came from implementing generative image models, translation models, and more.

If you haven’t yet, please show your support for the newsletter by subscribing!

Early-stage funding for generative AI startups drops 76% from its peak. Link.

  • Funding for generative AI startups has declined for two consecutive quarters, down 76% from its peak in Q3 2023.

  • VC deal value for early stage (pre-seed and seed) hit $122.9 million in Q1 2024 from $517.7 million in Q2 2023.

  • The investor ecosystem is stepping back after flooding the space to see how things play out.

China adds $48 billion to its semiconductor fund to lift chip-making capabilities. Link $.

  • The funding follows US efforts to tighten China’s access to advanced semiconductors and chip-manufacturing equipment.

  • This is China’s third installment of a national semiconductor fund.

  • China’s Ministry of Finance is the largest investor with follow-on from the country’s biggest state-owned banks and investment vehicles associated with local governments.

Klarna claims generative AI has cut marketing costs by $10 million annually. Link.

  • The fintech company Klarna is said to be using generative AI to run marketing campaigns and generate images.

  • The company cut its sales and marketing budget by 11% while increasing the total number of campaigns.

  • Image production costs have been reduced by $6 million using tools like Midjourney, DALL-E, and Firefly.

  • $4 million in savings came from reducing spend on external marketing suppliers for translation, production, and social agencies.

  • The company previously implemented OpenAI’s AI assistant and claimed it was doing the equivalent work of 700 FTEs.

Venture

Musk’s xAI successfully raised $6 billion to “catch up” to OpenAI and others. Link.

  • The funding round valued the company at $18 billion.

  • Investors include a16z, Sequoia Capital, and Saudi Arabia’s Prince Alwaleed bin Talal.

  • Musk says his AI business “still has a lot of catching up to do” concerning the competitive AI landscape.

  • The funds will be used to deploy xAI’s first product, build advanced infrastructure, and accelerate R&D of future tech.

LLMs

Mistral debuts a code-focused model. Link.

  • The model nicknamed Codestral is Mistral’s first coding-specific model.

  • The 22B parameter model is fluent in 80+ programming languages.

  • The release highlights the growing interest in leveraging LLMs to write code.

OpenAI begins work on the GPT-4 successor. Link $.

  • While the startup has likely been working on its next model for some time, the announcement comes just weeks after the groundbreaking release of GPT-4o.

  • The next model, which may be GPT-5, wouldn’t be available for several months, if not longer.

Ethics + AI Safety

Former OpenAI executive joins Anthropic. Link.

  • Former exec and AI safety lead Jan Leike announced on X that he will be joining OpenAI’s rival Anthropic to lead a new superalignment team.

  • He left OpenAI alongside Ilya Sutskever, resulting in the dissolution of OpenAI’s superalignment team.

  • Leike was vocal about OpenAI’s lack of respect for AI safety, claiming it was de-prioritized under ‘shiny products’.

Google’s AI Overview tool tells users to eat rocks, literally. Link $.

  • Due to hallucinations and misinformation, the tech giant will be scaling back the newly implemented AI search tool, AI Overviews.

  • A few hallucinations went viral on social media, including a response that told users to “eat at least one small rock per day” and to “add non-toxic glue to sauce” when making pizza.

  • The roll-back comes months after Google’s text-to-image generator blundered on its opening day announcement as well.

Upcoming Conferences

June 10 – 13

Moscone Center, San Francisco, CA

June 27

New York City, NY

Thank you for reading the AI Rundown by Lightscape Partners. Please send any questions, comments, or suggestions to [email protected].