Since the rise of ChatGPT, many new AI models, which can be called the second generation of generative AI, have emerged. While many students still use ChatGPT, a significant number have switched to these new AI models. In this teaching tip, we explore how these new generative AI tools are advancing technologically. We will focus on generative AIs that students commonly use, such as Wrtn AI, Google Gemini, and Claude. We will also examine the differences between these AIs and ChatGPT, analyzing the strengths and weaknesses of each. We hope this will serve as a useful reference for educators looking to incorporate generative AI into their teaching practices.

1. Wrtn AI (https://wrtn.ai/)
 Wrtn AI is a generative AI service developed by a South Korea’s startup, which has been gaining attention recently due to its rapid growth. Wrtn AI offers large language models (LLMs) such as OpenAI's ‘GPT-4’ and Google's ‘PaLM2’ for free. Therefore, as shown in the image below, it has the advantage of allowing users to utilize various language models like GPT-3.5, GPT-4, and Claude 2.1 within a single platform.

 Wrtn AI's search analyzes the user's query intent and explores highly relevant and useful real-time web information. As a result, it provides users with the latest information. To overcome the limitations of traditional Large Language Models (LLMs) that struggle with current information, Wrtn AI uses Retrieval-Augmented Generation (RAG) and internet browsing AI agent technologies. Additionally, it enhances reliability by providing information sources, related content, and follow-up questions. Moreover, there are no specific limits on the number of questions and answers. Users can ask and answer questions freely and unlimitedly across from all devices, including PCs, mobiles, web sites, and apps. This versatility has contributed to its popularity.

 When you access the Wrtn AI service, it also suggests prompts with messages like "How about these prompts?" It offers guidelines for creating PowerPoint presentations, drafting initial e-mails, interpreting Python code, and more. These diverse prompt recommendations make it user-friendly, as users can utilize them according to their needs. Additionally, Wrtn AI has recently added features for users to create their own AI character chatbots, revamped the AI chat window to cater to users seeking both information exploration and creative activities, and redesigned the initial interface to enhance accessibility for various AI services to be provided in the future. Wrtn also has a community space where users can interact with each other. Wrtn for Students(https://wrtn.circle.so/c/25bd0b/) serves as a space where students and teachers can share useful prompts or services and prompts created using Wrtn.

2. Google Gemini (https://gemini.google.com/app)

 Gemini is a conversational generative AI chatbot developed by Google. It is the rebranded version of Google's AI chatbot, previously known as "Bard". Launched in December of last year, Gemini showcases cutting-edge technology across various fields, including text, images, audio, and video. Google designed Gemini to be multimodal from the early stages of development by pre-training it on various modalities. Therefore, Gemini can understand and infer all types of inputs, including text, audio, images, and video. Additionally, Gemini 1.0 can understand, explain, and generate code written in programming languages such as Python and Java. Furthermore, as shown in the image below, it can provide multiple answers to a single prompt and allows users to listen to the responses in audio form.

 Google's Gemini offers a paid model similar to OpenAI's ChatGPT Plus service. The paid subscription service, Gemini Advanced, is available in 9 languages, including Korean, and features the powerful 'Gemini Ultra 1.0' AI model. This model achieved a score of 90.04% on the Massive Multitask Language Understanding (MMLU) test, which evaluates global knowledge and problem-solving abilities across 57 subjects, including mathematics, physics, history, law, medicine, and ethics. This score is higher than that of experts. Additionally, Gemini Ultra demonstrated excellent performance in image tests, showcasing its complex multimodal reasoning capabilities. Compared to ChatGPT Plus, offers the distinct advantage of providing reference materials. It also excels in searching for the latest information and producing results quickly.

3. Claude 3.0 (https://claude.ai/)  
 On March 8th of this year, IT media reported the results of an IQ test conducted on AI chatbots by Maxim Lott, an American TV producer and data analyst. According to the results, Claude 3, developed by Anthropic, scored above the human average IQ of 100, achieving the highest score of 101. Following Claude 3, ChatGPT-4 scored 85 points, Microsoft's Bing Copilot scored 79 points, and Google's Gemini scored 77.5 points.

 Users can automate complex tasks with Claude, such as planning and executing API, managing database, and interactive coding. They can also perform high-level analyses for brainstorming, generating hypotheses, reviewing research, creating charts and graphs, and analyzing financial and market trends. Anthropic, the company that created Claude, was founded by engineers who left OpenAI in opposition to OpenAI's commercial activities and creating a profit subsidiary. Anthropic offers three versions of Claude 3.0: Claude Hiku, Claude Sonnet, and Claude Opus. Users can access Claude Sonnet for free, while Claude Opus requires a $20 monthly subscription. Claude 3 Opus boasts benchmark scores that surpass those of GPT-4 and Gemini Ultra 1.0. Notably, in the Multilingual Math test, Opus's 0-shot performance surpassed the results of GPT-4's 8-shot performance. Although Claude 3 Sonnet has areas where it falls short compared to GPT-4, it outperforms other free products like GPT-3.5 and Gemini Pro in the Claude chat platform.

Another notable feature of Claude 3.0 is its enhanced focus on safety and ethical standards. Anthropic uses a proprietary training method called "Constitutional AI" to increase the safety of its AI models. This approach aims to embed clear ethical principles and behavioral norms within the AI, effectively preventing potential misuse. These efforts are more deeply integrated into Claude 3.0.

 We've explored the newly emerged generative AIs and their utilization in various universities. Wrtn AI, developed by a South Korean startup, offers major language models like OpenAI's GPT-4 and Google's PaLM2 for free. Gemini, Google's latest AI model, showcases new technologies across text, image, audio, and video, providing multimodal capabilities. Claude 3.0 is an AI-based platform capable of automating tasks such as API, database, and interactive coding, as well as performing high-level analyses.
 We hope this teaching tip helps our university students enhance their creativity and critical thinking skills by utilizing generative AI, enabling them to engage in more meaningful learning experiences with technology in the future.


Reference
Wrtn Homepage | Retrieved from https://wrtn.io/wrtn/

ZDNET Korea | Reporter Kim Mi-jeong | Wrtn AI Search Provides Answers with Real-Time Information | Retrieved from https://zdnet.co.kr/view/?no=20240326090257

Newspim | Reporter Song Eun-jeong | Wrtn Technologies, Launches Free Unlimited ‘AI Search’ Service | Retrieved from https://www.newspim.com/news/view/20240326000041

MT | Reporter Jeong Hye-in | ‘Bard’ disappeared… Google changes AI chatbot ‘Gemini’ and launches  paid service | Retrieved from https://news.mt.co.kr/mtview.php?no=2024020910325388222

Google Korea Blog | Sundar Pichai · Demis Hassabis | Introducing Gemini, the most capable and general-purpose AI model. | Retrieved from https://blog.google/intl/ko-kr/company-news/technology/gemini-kr/#performance

Google Korea Blog | Sundar Pichai | Opening a new chapter in the Gemini era | Retrieved from https://blog.google/intl/ko-kr/products/explore-get-answers/next-chapter-gemini-era-kr/

Gemini Release Update | Retrieved from https://gemini.google.com/updates

ZDNET Korea | Jeonghyeon Lee in Media Research Institute | ‘AI chatbot’ Claude 3 exceeds IQ score of 100 | Retrieved from https://zdnet.co.kr/view/?no=20240308100744

Modulabs | Youngbin Lee | Claude 3.0: Anthropic's AI Revolution, Opening New Horizons | Retrieved from
https://modulabs.co.kr/blog/anthropic-claude-3-0/

Anthropic | Introducing the next generation of Claude | Retrieved from https://www.anthropic.com/news/claude-3-family

Sang-eun Lee, Min-young Ku, Ye-jin Kim (2024). Trends in Generative AI Frequently Used by University Students (CTL Teaching Tips #38). Seoul: Sungkyunkwan University, Center for Teaching and Learning Innovation.