【作业】大学英语二.项目作业一
DeepSeek and Its Impacts
-- How will the newcomer in the fiercely competitive LLM market change us?
Li Siyuan, Zhao Zirui, Wang Zishi
Class 2413, CS, HUST
00.Introduction
Zhao Zirui
Today, I am thrilled to introduce DeeSeek and its impact, a groundbreaking AI-driven platform redefining the frontiers of data analysis and machine learning.
What is DeepSeek?
DeepSeek combines speed, accuracy, and scalability with proprietary algorithms to deliver industry-specific solutions. Its core mission? To enhance efficiency and decision-making across sectors—from healthcare to cybersecurity.
01.Impacts in the field of technology
Zhao Zirui
DeepSeek’s algorithmic efficiency sets new standards. Its Sparse Attention Mechanisms cut redundant computations in transformer models, slashing training costs by 50%. Meanwhile, Dynamic Neural Architecture Search automates neural network design—reducing model development time from months to days. Imagine a biotech startup accelerating drug discovery with this power.
Adaptability is another pillar. With Few-Shot and Zero-Shot Learning, DeepSeek trains models on minimal labeled data—a game-changer for niche fields like rare disease diagnosis. Even cross-domain transfer learning unlocks solutions: insights from gaming AI now optimize supply chains.
Scalability is achieved through distributed training frameworks, enabling trillion-parameter models across global GPU/TPU clusters. A multinational corporation recently trained a multilingual AI using 10,000+ GPUs in parallel. And with model compression, DeepSeek fits advanced AI into smartphones and IoT devices—democratizing access.
The impact on various technological breakthroughs
DeepSeek outperforms competitors. It’s 40% faster than GPT-4 in language translation and achieves 95% accuracy in autonomous vehicle object detection. In healthcare, it surpasses human radiologists in detecting early-stage lung cancer (AUC: 0.98 vs. 0.91). Cost-effectiveness shines too—SMEs save 35% on cloud costs via optimized resource algorithms.
Impact on related studies
DeepSeek drives a healthcare revolution: genome analysis time dropped from 2 weeks to 8 hours, accelerating pandemic responses. In education, DeepSeek Lite—a free version—empowers 50+ universities globally. What about Cybersecurity? Its adversarial training detected 99.7% of zero-day attacks in Pentagon trials. Even sustainability thrives: quantum-inspired algorithms cut energy use by 40%, while carbon credit partnerships offset 100% of training emissions.
02.Impacts in ecnomic and industry
Wang Zishi
[Opening]
Hello, everyone. Now I will show the topic of Economic and Industry Impacts. Let's examine how DeepSeek's technologies exemplify both the transformative power and complex consequences of AI adoption across industries.
[Manufacturing Disruption]
The numbers speak volumes. At Tesla's Shanghai factory, DeepSeek's vision systems achieved what once seemed impossible - a 300% reduction in defect detection time alongside 45% lower labor costs. But this efficiency comes at a human cost. The World Economic Forum warns of 12% global manufacturing job losses by 2030, even as AI-driven roles grow by 8%. The assembly line worker of tomorrow may need to become the AI supervisor of today.
[Service Sector Transformation]
In banking, chatbots powered by DeepSeek have enabled 30% reductions in call center staffing. Yet institutions now grapple with an unexpected challenge - 42% of customers report lower trust in fully automated services. Meanwhile, JD Logistics showcases AI's duality: While DeepSeek's route optimization boosted delivery speeds by 22%, automated sorting centers eliminated 15% of regional jobs. Efficiency gains and employment losses now walk hand-in-hand.
[The Great Debate]
OECD data reveals the core tension: Every 200% efficiency ROI from automation demands a $15,000 per worker retraining investment. Are we prepared to fund this transition at scale? The answer will define our economic future.
[AI-as-a-Service Revolution]
Now to the innovation frontier. DeepSeek's API handled 320 million daily calls in 2023 - 38% powering healthcare diagnostics that save lives, 29% driving financial models that protect assets. ClimateTech Inc. achieved a breakthrough, boosting extreme weather prediction accuracy from 78% to 91% using customized DeepSeek models. When AI meets domain expertise, it becomes a force multiplier for human capability.
[Long-Tail Innovations]
In sub-Saharan Africa, 12 million smallholder farmers now receive pest alerts through AgriTech systems combining DeepSeek's satellite analysis with local soil data.
Singapore's eldercare robots, enhanced by emotion AI, have reduced senior anxiety by 27% - proving technology can humanize care as populations age.
03. Societal and ethical impacts
Li Siyuan
Introduction
I'd like to focus on the society and ethical impacts of Deepseek. And in this part we will see what will this low-cost and open-source large language model bring to our ethical society.
Four features of DeepSeek
Firstly, before we really start this part. I'd like to make it clear for you that what features make DeepSeek different from other large language models. Those are low-cost, multilingual, open-source and transparency. DeepSeek-R1 operates at a fraction of the cost of its competitors. What's more, DeepSeek performs better in non-English languages such as Chinese, compared with its predecessor like ChatGPT. And what is the most surprising is the developing team of DeepSeek decide to make it open-source.That means you can download the releases on github for free and deploy it on your personal computer. For those who are already used to AI models, what can be new is its "deep seek" process in text format. That seems to make the thinking process of AI models more transparent.
What DeepSeek can help to change the conventional mode
In the past we face the problem that the popularity of large language models makes hardware more and more expensive and hardware makes the LLM more expensive because the AI models need a surprising number of those hardware to train. But after the appearance of DeepSeek we can see that it's okay to use less GPU/CPU to train a large language model and it can be at the similar level of those well-known LLMs like ChatGPT or Gemini. Maybe in the future the demand in hardware will be relatively loose and there will be less speculative froth. This will be beneficial to individuals and independent developers because they can buy hardware and the use large language models at a low prize. So you don't need to face the difficult choice between being left in isolated island in the digital age or be domained by big tech companies like Google and Microsoft.
How Multilingual facilitate fairness among different cultures
What's more DeepSeek seems to have a multilingual mastery. We can see in the graph on the left side -- it shows that DeepSeek's average performance in English contexts is not bad and it's performance in nonn-English contexts like Chinese has a definitely advantage over ChatGPT. I think multilingual ability is very important for a fair society because that people from different culture can share the benefits from the advance in technology. And you don't need to give up your culture, your tradition, your language to fit in those cut-edge technology.
The "Deep seek" process makes it more transparent and ethical
As what I already said, the "deep seek" process is what makes it new to users who are already used with other AI models. I think that it is thinking that creates the ethics of us human. By the "deep seek" process, we seem to be able to see the mideset of AI in text format. I'll say that it's a positive step towards an "ethical" AI. It makes the thinking process of AI transparent. I think that transparency will definitely make AI models more popular in some special place like hospital or transportation.
Summary
Li Siyuan
In summary, despite those powerful AI models that has already been used in the market DeepSeek will definitely bringdramatic El impacts on technology economic industry and over society.
That's all thank you.
浙公网安备 33010602011771号