The AI Rundown by Lightscape Partners - 11/04/24

Good morning, happy November, and welcome to another edition of The AI Rundown by Lightscape Partners.

• OpenAI has officially launched ChatGPT search, offering real-time answers on sports, stocks, news, and more. This new feature, which builds on the beta-tested SearchGPT, positions OpenAI in direct competition with giants like Google and Microsoft, adding another layer of rivalry despite Microsoft's substantial investment in OpenAI. ChatGPT Plus and Team users already have access, with broader availability on the way.

• Bret Taylor’s AI startup, Sierra, has attracted $175 million in new funding, giving it a valuation of $4.5 billion. Sierra's AI chatbots are designed to integrate with enterprise systems while minimizing hallucinations, providing brands with a more reliable tool for autonomous customer service. Investors such as ICONIQ and Thrive Capital are betting on its potential to transform customer interactions.

• Google is expanding AI capabilities for Maps and Earth with new generative features aimed at developers and urban planners. The updates allow for real-time insights using place data, and Rivian vehicles are set to incorporate these capabilities for delivering local information. Google's push into geospatial AI underscores its commitment to enhancing interactive, location-specific user experiences.

Stay with us as we unpack these top stories and explore their implications for the evolving AI landscape.

Our portfolio companies are hiring! Check out their job postings here.

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

 Hardware + Software

OpenAI will use AMD chips and explore custom AI hardware by 2026. Link.

• OpenAI plans to use AMD chips in its Azure-based infrastructure alongside Nvidia to support AI workloads.

• The company is also developing custom silicon with Broadcom, aiming for production by 2026.

• This move follows similar hardware strategies by other tech giants like Google, Microsoft, and Amazon.

• Experts suggest OpenAI may require substantial funding to compete with established AI hardware leaders.

AI-Generated Minecraft Demo Shows Potential for Real-Time Video Generation. Link.

• Decart and Lightscape portfolio company, Etched, created an AI-generated version of Minecraft as a demo for real-time interactive video generation, using no code and relying on AI-trained models. Try it yourself here!

• The AI model, named Oasis, was trained on millions of hours of Minecraft gameplay to generate next-frame predictions in real-time.

• The demo is limited by low resolution, short playtime, and hallucinations, but further innovations and a new custom chip, Sohu, could enhance performance.

• Decart and Etched envision using these advancements for other real-time applications, such as virtual doctors or tutors.

Models

Cohere releases new models to boost multilingual AI capabilities. Link.

• Cohere launched Aya Expanse 8B and 35B, enhancing AI performance across 23 languages.

• The Aya project aims to bridge language divides in foundation models, extending AI accessibility beyond English.

• Aya Expanse models outperformed similar-sized models from Google, Mistral, and Meta in multilingual benchmarks.

• Cohere used data arbitrage and preference training to ensure the models cater to global preferences, tackling safety challenges in multilingual contexts.

• The company also expanded capabilities in its Embed 3 and Command R models recently.

Product Launches

OpenAI launches ChatGPT search, expanding its competition with Google and Microsoft. Link.

• OpenAI officially launched ChatGPT search, offering real-time results for sports, stocks, news, and more, backed by partnerships with news and data providers.

• This new feature is a rebranded version of SearchGPT, which was beta-tested in July.

• The launch further positions OpenAI as a competitor to Microsoft, despite the latter’s $14 billion investment in OpenAI.

• All ChatGPT Plus and Team users can access the feature immediately, with a broader rollout expected over the coming months.

• OpenAI’s valuation recently hit $157 billion, despite projected losses of $5 billion on $3.7 billion in revenue.

Oracle introduces AI-powered electronic health record. Link.

• Oracle launched a new AI-enhanced electronic health record (EHR) system, marking its most significant update since acquiring Cerner in 2022.

• The new EHR features AI and voice activation, allowing doctors to access patient data via simple questions, reducing time spent navigating menus.

• Oracle aims to regain market position after recent struggles, hoping the new system's cloud and AI integration will attract hospitals.

• The new EHR, separate from Cerner's infrastructure, is set for early adoption next year while existing features like Clinical AI Agent are already embedded.

Enterprise + Consumer Applications

AWS and Box expand AI capabilities for enterprise content. Link.

• Box customers can now use Anthropic’s Claude and Amazon Titan foundation models through Box AI, thanks to a new AWS partnership.

• The integration allows Box Enterprise Plus customers to leverage Amazon Bedrock to build AI applications, combining advanced AI models with data stored on Box.

• A new Box connector for Amazon Q Business enables AI-driven tasks like summarizing and generating content using Box-stored data.

• The collaboration aims to simplify and secure the use of generative AI, enhancing productivity and innovation for enterprises globally.

Google Maps and Google Earth gain generative AI features for developers. Link.

• Google introduced new generative AI capabilities for Maps Platform and Earth, aimed at providing developers and urban planners with real-world insights.

• The "grounding with Google Maps" feature allows developers to use up-to-date place data to enhance AI-powered experiences, ensuring more factual and current answers.

• Rivian vehicles will soon use generative AI summaries to provide drivers with information on nearby restaurants, shops, and more, starting next month.

• These updates are part of Google's ongoing efforts to support generative AI in geospatial applications.

Funding & Valuations

Bret Taylor’s AI startup Sierra raises $175M at a $4.5B valuation. Link.

• Sierra, co-founded by Bret Taylor and Clay Bavor, focuses on AI customer service chatbots that integrate with enterprise systems to perform tasks autonomously.

• The company claims its tech is less prone to AI "hallucinations" and allows customization of chatbot personalities to fit corporate brands.

• The platform uses models from OpenAI, Anthropic, and Meta for task completion and accuracy checks.

• Investors include ICONIQ, Thrive Capital, and Greenoaks Capital, bringing Sierra’s total funding to $285 million.

Read AI raises $50M to expand its AI copilot tools. Link.

• Read AI secured $50M in Series B funding led by Smash Capital, bringing its total funding to $81M just six months after its Series A.

• The startup’s AI summarization tools aim to improve productivity by unifying and analyzing communications across platforms like Google Meet, Zoom, Slack, and Gmail.

• A new tool, Read AI for Gmail, integrates with Gmail to offer contextual insights, summaries, and AI-generated responses, part of its "copilots everywhere" initiative.

• CEO David Shim emphasized enhancing workflows by connecting messages and meetings, rather than changing existing processes.

Data Centers and Energy

AI Boom Puts Europe's Environmental Goals at Risk. Link.

• The demand for AI is increasing energy use in European data centers, threatening decarbonization goals, according to Goldman Sachs research.

• Cooling demands for AI chips are driving developers to lower water temperatures, which may lead to unsustainable practices, warns Michael Winterson, chair of the EUDCA.

• Nvidia's AI chips require extremely dense computing power, which produces significant heat, making efficient cooling critical.

• Companies are turning to liquid cooling and working with European regulators to address these energy challenges, balancing power efficiency with sustainability goals.

Microsoft struggles to meet AI data center demand. Link.

• Microsoft reported $65.6 billion in revenue for Q1 FY2025, surpassing expectations, but faced challenges due to data center constraints.

• Azure growth of 34% beat estimates, but Microsoft's inability to build data centers fast enough limits near-term growth.

• High AI demand has driven these constraints, with CEO Satya Nadella highlighting external third-party delays impacting capacity.

• Despite current challenges, analysts and Microsoft remain optimistic about long-term AI and cloud growth prospects.

Portfolio Job Opportunities

Articul8

Colleva

Etched

HealthBird

Private AI

Shield AI

Standard Bots

Toggle AI

Vizit

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