DeepSeek, a Chinese AI lab, has developed AI models that challenge the dominance of Silicon Valley's leading offerings. These models utilize a unique approach called test-time compute, breaking down complex tasks into smaller, manageable prompts. Business Insider tested one of DeepSeek's models, showcasing its ability to solve intricate math problems and demonstrate its thought process step by step. The model's transparency and adaptability highlight its potential in various AI applications.
DeepSeek, a Chinese AI lab, has recently released AI models that rival or surpass some of the leading offerings from Silicon Valley . DeepSeek employs a technique called test-time or inference-time compute, which breaks down queries into smaller tasks. Each task is then treated as a new prompt that the model tackles. This approach requires running a new request for each step, known as the inference stage in AI.
Business Insider recently evaluated one of DeepSeek's models using its DeepThink mode, which allows users to observe the model's entire thought process. They presented the model with a multi-step math problem suggested by Charlie Snell, an AI researcher at UC Berkeley. Snell explained that DeepSeek demonstrates strong performance on complex mathematical problems that necessitate extended and staged thinking. He selected problems from the American Invitational Mathematics Examination, a challenging test designed for high-achieving high school mathematics students.Snell shared, 'I put questions from that into the DeepSeek model. I read through the chain of thought. It was understandable.' The math problem Snell proposed for BI's DeepSeek demonstration was: 'Find a sequence of +, -, /, * which can be applied to the numbers 7, 3, 11, 5 to get to 24, where each of the given numbers is used exactly once.' BI input this prompt into DeepSeek's chat window on its website. The model initially acknowledged the challenge: 'Alright, so I've got this problem here: I need to use the numbers 7, 3, 11, and 5, and combine them with the operations of addition, subtraction, multiplication, and division, using each number exactly once, to get to 24. At first glance, this seems a bit tricky, but I think with some systematic thinking, I can figure it out.' It then embarked on a multi-step process spanning approximately 16 pages of discussion, encompassing mathematical calculations and equations. While the model occasionally made errors, it identified these mistakes and persevered. Instead of giving up, it promptly moved on to explore alternative solutions. DeepSeek's thought process was transparent, allowing users to witness its trial-and-error approach.Snell emphasized the significance of DeepSeek's ability to demonstrate its reasoning, stating, 'You can see it try different ideas and backtrack.' He particularly highlighted the following excerpt from DeepSeek's chain of thought: 'This is getting really time-consuming. Maybe I need to consider a different strategy. Instead of combining two numbers at a time, perhaps I should look for a way to group them differently or use operations in a nested manner.' This instance showcases DeepSeek's capacity to adapt its approach based on the complexity of the problem and its own progress
AI Deepseek Artificial Intelligence Silicon Valley Test-Time Compute Math Problems Chain Of Thought Business Insider
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Chinese AI Startup DeepSeek Rocks Silicon Valley with Open-Source ModelDeepSeek, a Chinese AI startup, has made a significant impact in Silicon Valley with its open-source large language model, R1, which outperforms leading American AI models despite using fewer and less powerful chips. The company's rapid progress has sparked debate about the future of AI development and the role of open-source technology.
Read more »
Chinese AI Startup DeepSeek Makes Waves in Silicon ValleyDeepSeek, a Chinese AI company, has rapidly gained prominence in Silicon Valley with its open-source large language model, R1, challenging the dominance of established American AI companies. R1's impressive performance and cost-effective development have sparked debate and highlighted the significance of open-source research in the AI landscape.
Read more »
Chinese AI Startup DeepSeek Stuns Silicon Valley with Rapid RiseDeepSeek, a Chinese AI company specializing in open-source large language models, has made a significant impact on the AI landscape with its latest model, R1. Despite utilizing fewer and less powerful chips compared to U.S.-based rivals, DeepSeek's R1 achieved impressive performance rankings, challenging the dominance of established AI companies.
Read more »
Silicon Valley praising Chinese AI startup DeepSeek: 'Profound gift to the world'A Chinese artificial-intelligence company has Silicon Valley raving, calling it 'amazing and impressive,'despite working with less-advanced chips.
Read more »
Chinese AI Firm DeepSeek Makes Waves in Silicon ValleyDeepSeek, a Chinese AI startup specializing in open-source large language models (LLMs), has surged to prominence with its release of R1, a powerful model designed for complex problem-solving. R1's impressive performance, achieved at a fraction of the cost and resources compared to US counterparts, has garnered attention and sparked debate about the future of AI.
Read more »
Chinese Startup DeepSeek Shocks Silicon Valley with Cost-Effective AIDeepSeek, a Chinese startup, has gained prominence by developing a large language model (LLM) that rivals the performance of models from OpenAI, Google, and Meta, but at a fraction of the cost. This achievement raises questions about the US's ability to maintain its AI dominance.
Read more »