Fall 2024: Friday 11:00AM-11:50 AM KST (Thursday 7PM PT) @Zoom
In this course, you will learn from leading AI experts about their insightful ideas. The course will cover their thoughts, beliefs, and concepts, including their risky but impactful ideas and what they believe but cannot yet prove. It will also discuss what they think will change the world and the biggest challenges they face.
- Participations: 30%
- Report 40%
- Idea Proposal: 30%
Students should submit a 1-page talk report after each talk by Friday at 5 PM. The report should include but not limited:
- Summary: A quick summary of the talk.
- Learned: What did you learn from the talk?
- Surprises: Any unexpected things you learned from the talk.
- Questions: Any questions you asked during or after the talk (if any).
- Additional Questions: 2-3 additional questions you would like to ask.
After all talks, students should come up with project ideas and submit a proposal. Students can form a team with up to 5 members and are free to choose their teammates. This should include, but is not limited to:
To ensure a structured and impactful experience, please adhere to the following guidelines:
- Problem Definition
Identify a Specific Issue: Clearly articulate a real-world problem that you aim to address using generative AI. This could range from enhancing educational tools to improving healthcare diagnostics. For example, "Developing an AI-driven platform to assist students with personalized study plans."
Assess the Scope and Impact: Evaluate how widespread the problem is and the potential benefits of solving it. Consider factors such as the number of individuals affected, economic implications, and societal benefits.
- Solution Impact and Expected Outcomes
Define Positive Changes: Describe the tangible benefits your solution will bring. For instance, "Reducing study time by 30% through personalized learning paths."
Consider Long-Term Effects: Reflect on how your solution could evolve over time and its potential to inspire further innovations or improvements in related areas.
- Utilized Generative AI Technologies and Models
Specify AI Tools and Models: Clearly state which generative AI models or tools you plan to use, such as GPT-4 for natural language processing or DALL·E for image generation.
Outline Application Strategies: Detail how you intend to integrate these technologies into your solution. For example, "Leveraging GPT-4 to generate personalized study materials based on individual learning styles."
- Prototype Design and Visualization
Create a Visual Representation: Develop a prototype sketch on paper that illustrates the core features and user interface of your solution. This will help convey your concept effectively.
Include Key Elements: Ensure your sketch includes aspects like user interactions, data flow, and any AI components involved in the process.
Week 1 Sep 13: Course Intro & (Solar Pro) LLM Trend by Sung Kim (Prof at HKUST/CEO at Upstage)
Week 3 - Oct 11: "Phi-3.5: The Power of MoE in Advanced Training" by Dr. Young-Jin Kim (Principal Researcher at Microsoft)
Week 4 - Oct 18 (Time change 2PM KST): "Code & Synthetic data: a match made in LLM heaven" by Matthias Gallé (Head Research Scientist at Cohere)
Read and write 1 page report: https://www.khoslaventures.com/ai-dystopia-or-utopia/
Solar Pro: towards workspace LLMs
This talk will describe the requirements of AI enterprise solutions and how Solar, Upstage's LLM, plays to meet the requirements. In addition, the presentation will introduce Solar Pro, which is scheduled to be released in November 2024, including its key technologies such as depth-up-scaling and LP data pipeline.
- Director of Solar Intelligence Team at Upstage
- (Former) Lead of Document AI Model at Upstage
- (Former) Tech Lead of OCR at Naver
Week 11 - Dec 13 (Friday , 11-1PM, @서울대 302동 105호 (대강당)): Proposal Post-Presentation Lunch (Sponsored by Oracle)
- 윤진수, 국민은행 KB국민은행 테크그룹 경영자문역 (전 부행장님)
- 조현철 상무님, LGU+ Data Science Lab장
- Cindy Jin, Partner at SBVA
- 김태완 상무, Oracle, Master Principal Cloud Engineer
- NamKyu Choi, EICN, Solutions Development Division, Development of RAG, search engines, chatbot/callbot conversational engines, and text analysis solutions