Silicon Valley thinks DeepSeek's success is shady

Hello, AI lovers! Here’s what you need to know about AI today:

👉 DeepSeek has spent much more than $6 million to train R1

👉 Scientist can now detect breast cancer years before its onset

👉 Microsoft has added DeepSeek models to its Azure cloud service

and many more!

📧 Did someone forward you this email? Subscribe here for free to get the latest AI news everyday!

Read time: 5 minutes

What’s going on: Dario Amodei, CEO of Anthropic (the company behind Claude chatbot), has challenged the narrative around DeepSeek's claim that they developed an advanced AI model, R1, with just $6 million. Amodei suggests this achievement was significantly more costly than similar efforts by US companies, which have spent billions on AI development.

What does it mean: While DeepSeek's reported cost was low, the actual investment in resources, particularly computing hardware, aligns more closely with US standards. Reportedly, DeepSeek had access to around 50,000 H100 Hopper generation Nvidia chips, costing around $1 billion. Seems like folks at DeepSeek are not just good at building cutting-edge AI but they are genius marketing experts as well! The “50,000 GPUs estimation” was first proposed by the CEO of Scale AI in an interview with CNBC. His claim was then supported by Elon Musk, and now Dario Amodei.

More details:

• Amodei advocates for strong export controls on GPUs to China, to prevent them from gaining a significant edge in AI.

• Anthropic's Claude 3.5 Sonnet, which outperforms DeepSeek's model in various benchmarks, was developed for a few tens of millions of dollars.

• Claude 3.5 Sonnet has enhanced reasoning, state-of-the-art coding skills, computer use, and 200K context window.

• Amodei believes US companies will soon match or exceed DeepSeek’s efficiency.

What’s going on: A groundbreaking study from the Norwegian Institute of Public Health (FHI) has demonstrated that AI has the potential to predict breast cancer risk in women years before it becomes clinically apparent. The research focused on mammograms collected from a large-scale Norwegian breast cancer detection program which ran from 2004 to 2018. Researchers were able to analyze these images to forecast which women were more likely to develop breast cancer.

What does it mean: Remember, the best treatment for cancer is to prevent it. Later-stage treatments (surgery, chemotherapy, radiotherapy, and immuno-therapy) are always painful, traumatizing and expensive. We are living in very interesting times now that we have the technology to detect diseases like Alzheimer’s and cancer, years before their onset. With the ability to predict cancer risk well in advance, healthcare providers could provide screening and preventive measures more effectively to those at highest risk.

More details:

• The study's methodology involved feeding the AI algorithm with mammograms from 116,495 women, out of which 1,607 were later diagnosed with breast cancer.

• The AI not only predicted the risk of cancer but astonishingly, it also indicated which breast would be affected, providing this foresight 4 to 6 years before the actual diagnosis was made.

• Their findings are detailed in a publication in the "Journal of the American Medical Association Network".

👋 Do you enjoy reading our newsletter? Let’s spread the joy by sharing the link below with your friends. You can also forward this email to them! Thanks!
the79newsletter.com/subscribe

☁ Microsoft has integrated DeepSeek's R1 model into its Azure AI Foundry service and GitHub. They have performed automated assessments of model behavior and extensive security reviews to mitigate potential risks of R1.

💰 SoftBank is in talks to invest up to $25 billion in OpenAI as part of a broader partnership that could see the Japanese holding spend more than $40 billion on AI initiatives.

📱 DeepSeek has reached the No. 1 spot on the US Google Play Store with its Android app, just days after topping the Apple App Store.

💲 Mark Zuckerberg downplays DeepSeek advancements and promised Meta will continue to invest 100s of billions of dollars in AI over the long term.

🚀 Ex-Google and Apple engineers have launched Oumi (open universal machine intelligence), an unconditionally open-source AI platform that could help to build the next DeepSeek. If you are interested to contribute to the development of AI, visit their website.

📃 A report by an app analytics firm, Appfigures, indicates that 85% of ChatGPT's mobile users are male, with over half being under 25 years old.

👩🏼‍💻 Clutch Security has raised $20 million to expand its non-human identity security platform, focusing on protecting machine identities like API keys and service accounts in an increasingly AI-driven enterprise environment.

✋🏼 DeepSeek's AI avoids answering 85% of sensitive China-related prompts, showing nationalistic bias and vulnerability to jailbreaking.

AI + learning roadmap

Amazon - ML Data Linguist III - Bilingual, Bedrock

Mixpanel - Data Scientist

Phantom AI - Data Labeling Specialist

NIQ - Data Scientist

Thank you for staying with us like always! If you are not subscribed, subscribe here for free to get more of these emails in your inbox! Cheers!