Anthropic CEO explains DeepSeek AI
OpenAI eyes $25B investment from SoftBank
Microsoft brings DeepSeek AI to its cloud
See your finances more clearly with BILL & Ray-Bans
Over 60% of finance leaders that we surveyed this year said that financial automation helps them see a clearer picture of their financial operations.
Choose BILL Spend & Expense and get:
Real-time visibility and great rewards
Fraud protections
Automated expense reporting for your business
Plus, get a pair of Ray-Ban Smart Glasses when you demo BILL.1 Ready to see clearly for 2025?
1 Terms & Conditions Apply. See offer page for more details. The BILL Divvy Card is issued by Cross River Bank, Member FDIC, and is not a deposit product.
***
Anthropic CEO explains DeepSeek AI
Anthropic CEO Dario Amodei breaks down the engineering advancements in DeepSeek AI. Anthropic sells Claude AI. While markets and media focused intensely on DeepSeek’s R1 model, Amodei points out that the company’s more significant innovation came earlier. “DeepSeek-V3 was actually the real innovation and what should have made people take notice a month ago (we certainly did). As a pretrained model, it appears to come close to the performance of state of the art U.S. models on some important tasks, while costing substantially less to train.”
The distinction between V3 and R1 is crucial for understanding DeepSeek’s true technological advancement. V3 represented genuine engineering innovations, particularly in managing the model’s “Key-Value cache” and pushing the boundaries of the mixture of experts (MoE) method. This insight helps explain why the market’s dramatic reaction to R1 may have been misplaced. R1 essentially added reinforcement learning capabilities to V3’s foundation — a step that multiple companies are currently taking with their models.
Amodei “It’s been reported — we can’t be certain it is true — that DeepSeek actually had 50,000 Hopper generation chips, which I’d guess is within a factor ~2-3X of what the major U.S. AI companies have. Those 50,000 Hopper chips cost on the order of ~$1B. Thus, DeepSeek’s total spend as a company (as distinct from spend to train an individual model) is not vastly different from U.S. AI labs.”.
"All of this is to say that DeepSeek-V3 is not a unique breakthrough or something that fundamentally changes the economics of LLM’s; it’s an expected point on an ongoing cost reduction curve. What’s different this time is that the company that was first to demonstrate the expected cost reductions was Chinese"....
***
Kickstart Your Day with Huel
In a world where every second counts, Huel Black Edition is the smart meal solution for your busy lifestyle.
With 40g of protein and 27 essential vitamins in each serving, it’s designed to fuel you fast.
Plus, with flavors like Chocolate, Cookies & Cream, and Strawberry Shortcake, Huel Black Edition turns meal prep into a 30-second hack.
Ready to fuel up? Get 15% off your first order, plus a free t-shirt and shaker bottle with code BEHUEL15.
***
OpenAI eyes $25B investment from SoftBank
SoftBank is currently in negotiations to invest up to $25 billion in OpenAI, positioning itself as the largest financial supporter of the ChatGPT creator. "The talks are ongoing and the amount that SoftBank could invest in primary equity into OpenAI is a moving target," the source told Reuters, emphasizing the fluid nature of the negotiations.
An investment of $15 billion or more would establish the Japanese firm as OpenAI's largest individual backer. Currently, Microsoft holds the largest stake in the startup, having invested around $13 billion out of the $20 billion OpenAI has raised across various funding rounds, leading to a valuation of $157 billion in 2024. SoftBank itself acquired a $2 billion stake in OpenAI last year.
OpenAI's CEO, Sam Altman, has been seeking ways to reduce the company's reliance on Microsoft for computing resources. This includes initiatives like developing proprietary chips and collaborating with alternative cloud providers such as Oracle. Part of the Stargate AI agreement saw Microsoft agreeing to relinquish its exclusive cloud service provider status for OpenAI….
***
Microsoft brings DeepSeek AI to its cloud
Despite allegations from its partner OpenAI that DeepSeek may have infringed on intellectual property and violated terms of service, Microsoft remains keen on integrating DeepSeek's cutting-edge models into its cloud platform. In a recent announcement, Microsoft revealed that R1, DeepSeek's reasoning model, is now available on the Azure AI Foundry service.
The Azure AI Foundry service, a comprehensive platform for various AI enterprise services, now features R1, which Microsoft assures has undergone extensive evaluations for safety and security. In their blog post, Microsoft highlighted the rigorous red teaming and automated assessments conducted to ensure the model's reliability and mitigate potential risks.
Looking ahead, Microsoft plans to offer "distilled" versions of R1 for use on Copilot+ PCs, a line of Windows hardware designed for AI readiness. "As we continue expanding the model catalog in Azure AI Foundry, we’re excited to see how developers and enterprises leverage R1 to tackle real-world challenges and deliver transformative experiences," Microsoft stated in the post….