🕶️ AI takes a detour: scaling laws hit a speed bump

Six bullets of updates

1. ⚡ Enel enters the small modular reactor race alongside Ansaldo and Leonardo, combining deep pockets with rising AI power demand.

2. 🚀 Trump watches SpaceX's sixth Starship launch, marking a milestone as the upper stage reignites in space.

3. 🚫 The UK considers banning social media for under-16s due to online safety concerns and risks to children.

4. 🥧 Pie, a social app, raises $24M to combat loneliness by incentivizing event hosts in San Francisco.

5. 🎧 Chroma, backed by Twitter and Pinterest co-founders, is acquired by AI audio company Bronze after raising $7.19M.

6. 🌍 Yuka empowers 60M+ users to challenge brands over harmful additives in their products.

✦

✦

From scaling to smarts: AI’s midlife crisis

AI research is shifting as traditional scaling laws—boosting model performance with more compute and data—reach diminishing returns. This method, pivotal for advances like ChatGPT, no longer delivers rapid improvements. Instead, a new approach called “test-time compute” is gaining traction. This technique uses extra computational power after a prompt, allowing AI models to iteratively refine their responses, opening new possibilities for growth.

While this new approach holds promise, its scalability and practical applications remain uncertain, potentially requiring significant compute resources for complex tasks. Despite these challenges, industry experts remain optimistic. Innovations in user experience, tooling, and intelligent prompting could further enhance current models' capabilities without necessitating leaps in base model performance. As AI labs pivot to adapt, the next era of advancements will likely redefine how models are trained and utilized, even as users may only gradually notice these shifts.

✦

Startup Events and Deadlines

1. Today! Crash Course in Financial Modeling | Nov 21 | Webinar

2. Founders Social | Supermoon Ventures & Station3 | Nov 22 | NYC

3. From Pitch Deck to Funding | Nov 25 | Webinar

4. VC and Angel Investors NYC Meetup | Nov 26 | NYC

5. NYC B2B: Pre-Seed + Seed Founders & Investors Meetup | Dec 3 | NYC

✦

✦

These companies just raised money

1. 🚀 LogicMonitor raises $800M, hitting a $2.4B valuation as demand for AI and observability tools surges.

2. 📈 Odoo raises $527M in secondaries, pushing its valuation to $5.26B as the ERP market embraces AI disruption!

3. 🎯 Tekever lands $74M to supercharge dual-use drones, targeting U.S. expansion in booming defense markets.

4. ⚡ Federato grabs $40M to enhance AI-driven insurance underwriting, delivering quotes 90% faster.

5. 🧬 Converge Bio secures $5.5M seed funding to power biotech LLMs, striving to be the "everything store" for genAI in biotech.

✦

ChatGPT for classrooms: teaching aid or tech gimmick?

OpenAI has launched a free course aimed at helping K-12 teachers integrate ChatGPT into classrooms, developed in partnership with Common Sense Media. The program covers AI basics, ethical use, and pedagogical applications, with early deployments in U.S. schools showing positive feedback from participants. While OpenAI promotes the potential of AI tools for creating lesson plans and tutorials, some educators remain skeptical about its limitations and ethical implications. Concerns include unclear policies on data privacy, risks of over-reliance, and the lack of focus on non-OpenAI AI tools, potentially reinforcing monopolistic dynamics.

Despite mixed opinions, proponents argue that thoughtful integration of AI could enhance education. However, surveys show many teachers and administrators remain hesitant, citing fears of AI replacing critical teaching steps and challenges in managing its impact. With the AI in education market projected to grow significantly, OpenAI’s efforts highlight both the promise and controversy surrounding AI’s role in learning.

✦

✦

✦

How telehealth is poisoning America

✦