DeepSeek caused $1 trillion+ market cap drop in USA

Hi everyone! Here’s what you need to know about AI today:

👉 DeepSeek beats OpenAI with its multi-modal model

👉 Free Chinese agentic AI model can easily replace OpenAI’s “Operator”

👉 US stock market $1+ trillion loss due to DeepSeek-R1

and many more!

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

Read time: 4.9 minutes

What’s going on: DeepSeek, the viral Chinese AI company, seems unstoppable at this point! They have released a new family of multi-modal AI models called Janus-Pro, which they claim surpass the performance of OpenAI's DALL-E 3. The models come in various sizes, with parameter counts ranging from 1 billion to 7 billion, and are designed to generate high-quality images from text prompts. (Check out the models on Hugging Face)

What does it mean: The release is significant as it not only shows the capabilities of DeepSeek but also highlights the competitive landscape in AI, particularly in image generation. DeepSeek's achievements are sparking a lot of discussions about the potential shift in AI dominance from the US to China. The release of DeepSeek-R1 already has caused the US stock market to drop $1+ trillion in value, with the biggest loser being Nvidia (its stock value just fell 17% yesterday and they have lost almost $500 billion overnight 🤯).

More details:

• The largest Janus-Pro model, Janus-Pro-7B, beats DALL-E 3 as well as other models such as PixArt-alpha, Emu3-Gen, and Stability AI‘s Stable Diffusion XL on two benchmarks: GenEval and DPG-Bench.

• Janus-Pro is open-sourced under MIT License and can both analyze and create new images.

What’s going on: It looks like China has officially stolen the AI spotlight! Yesterday, Alibaba’s Qwen team released a new family of AI models, Qwen2.5-VL, that can perform a number of text and image analysis tasks in the form of an AI agent. The models can parse files, understand videos, and count objects in images, as well as control a PC similar to the model powering OpenAI’s recently launched Operator.

What does it mean: Everyday that passes by, it is getting harder and harder for OpenAI to stay relevant in the market financially. Just a couple of days ago, OpenAI released Operator with a hefty price tag of $200/month; but multiple companies have already shipped free and open-source alternatives to the “Operator“. For instance, ByteDance has recently released UI-TARS and Abacus AI launched a new “Computer Agent” feature that you can run on ChatLLM.

More details:

• The best Qwen2.5-VL model beats OpenAI’s GPT-4o, Anthropic’s Claude 3.5 Sonnet, and Google’s Gemini 2.0 Flash on a range of video understanding, math, document analysis, and question-answering evaluations.

• Qwen2.5-VL is available to use in Alibaba’s Qwen Chat app and to download from Hugging Face.

• The model can analyze graphics and charts, extract data from scans of invoices and forms, and understand long (multiple-hour) videos.

• Watch two cool examples of Qwen2.5-VL model being used in real-time, here and here.

💬 Grok 3, xAI's next-generation AI model, briefly became accessible to some users on X (formerly Twitter), with capabilities in logical reasoning and coding before access was revoked.

👨🏼‍💻 DeepSeek, a trending Chinese AI startup, has temporarily paused new user signups due to a cyberattack, while existing users can still access their accounts.

📱 Meta AI now utilizes data from users' Facebook and Instagram profiles to tailor responses, enhancing personalization but raising privacy concerns.

🤖 1X, a robotic startup backed by OpenAI, has acquired Kind Humanoid, a Norwegian robotics startup.

📰 Quartz, international business news website, has been using AI to generate news articles which has raised questions about transparency and the role of AI in journalism.

🐋 Former Intel CEO, Pat Gelsinger, is using DeepSeek's AI models instead of OpenAI at his startup Gloo, which shows the strong influence of DeepSeek on American startups.

💊 Reid Hoffman-backed AI drug discovery startup, Manas AI, has raised $24.6 million which is significantly less than other AI drug discovery startups (some of them have raised $1+ billion initially). FYI, Reid Hoffman co-founded LinkedIn back in 2003 and sold it to Microsoft in 2016.

AI + SEO problems

BambooHR - AI Engineering/Data Science Intern

Twilio - Machine Learning Engineer

Hayden AI - Data Analyst

ServiceNow - Senior Software Engineer, Conversational AI

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!