2.1. Text-Based AI models

Currently, five major applications have gained popularity in the AI market, offering a user-friendly interface. All of them function based on human conversation principles and are suitable for generating, editing, and translating texts. The differences lie primarily in the format and volume of input and output, as well as in their integration with other services (e.g., image generators, search engines, etc.). In addition to these, a few other useful text-based AI models are introduced here.

Microsoft Copilot

Website: https://copilot.microsoft.com/

Staff and students at the University of Tartu can use the enterprise version of Microsoft Copilot, an AI-based text assistant, ensuring that data used in conversations remains protected and does not leak outside the organisation.

Microsoft itself describes Copilot as “your everyday AI companion, offering AI-powered conversations on the web.” The enterprise version available to University of Tartu staff and students (ChatGPT-4 + DALL·E 3) ensures that conversation data remains secure within the organization. Through a convenient user interface, users can interact with AI and choose between different response styles. Microsoft Copilot is integrated with Edge and Chrome extensions. Users can provide input in text format or as a file (.doc, .xls, .pdf). Image files cannot be uploaded, but the output can be text, images, or tables. The maximum input length is 2,000 characters, and the output is limited to 8,000 characters.

As a more technical solution, Azure OpenAI API is available to university members, enabling access to OpenAI’s GPT-3.5 Turbo, GPT-4o, and GPT-4o Turbo models. These models are known for their ability to generate content, summarize large datasets, conduct semantic searches, translate natural language into code, and more. The Azure OpenAI API also allows interaction with AI, but through custom applications or interfaces. See the IT department’s Azure OpenAI API guide for more details.

Microsoft 365 Copilot is an integrated AI application that functions within various Microsoft Office applications, such as Word, Excel, and Teams, offering recommendations, content generation, and task automation. As of October 3, 2024, Estonian language support is not yet available (see Microsoft Copilot’s supported languages).

GitHub Copilot is an AI assistant for software developers, helping to generate code. The university does not currently provide centralized access to this service, but individuals can join the Campus Program, which allows the use of GitHub Copilot for free under an educational license.

ChatGPT (OpenAI)

Website: https://chatgpt.com/

ChatGPT offers advanced text generation, image generation, web browsing, and data analysis. However, the free version has certain limitations. Input can be provided as text or an uploaded file, and the output can be text, simple diagrams, tables, or images. The free version offers limited access to ChatGPT-4o (switching to the weaker ChatGPT-4o mini every five hours) and DALL·E 3 image generation (maximum of two images per day). The input and output limit is 4,096 characters, or approximately 3,000 words.

The paid version, ChatGPT Plus ($25/month), allows users to select between different models (ChatGPT-4o, o1-preview, o1-mini, GPT-4). It has a message limit of up to 80 messages every three hours, offers faster response times, and enables users to create custom GPT models.

Gemini (Google DeepMind)

Website: https://gemini.google.com/app

Gemini, formerly known as Bard, integrates with various Google services. Input can be provided as text, image files, or simple voice commands. The output includes text, basic tables, and images generated using the Imagen 3 image generator.

The Advanced version ($23.50/month) offers extensive Google integration, an image generator, and the ability to create custom models.

The free version supports up to 32,000 tokens, which corresponds to text components such as words or word segments.

Claude (Anthropic)

Website: https://claude.ai/

According to its developers, Claude prioritizes safety and reliability while offering large memory capacity, strong analytical abilities, and code-writing capabilities. Unlike some other AI applications, Claude does not conduct real-time internet searches. Input can be provided as text, .pdf, .doc, .txt, or image files, while the output is text-only.

The free version of Claude 3.5 Sonnet allows up to 30 messages per day and supports shorter prompts. However, it is often considered superior to other AI applications due to its ability to process extensive input and output (up to 75,000 words).

The Pro version ($20/month) offers access to Claude 3 models (Opus, Sonnet, Haiku) and allows around 45 messages every five hours.

Perplexity (Perplexity AI)

Website: https://www.perplexity.ai/

Perplexity functions as a search engine that provides answers with source references. It utilizes GPT-3.5, GPT-4o, Claude, and Gemini models while retrieving real-time information. Input can be provided as text or files, and the output consists of texts with cited sources.

The free version uses the GPT-3.5 model and allows up to five searches per day.

The Pro version ($20/month) operates on GPT-4o and Claude 3 models, offering more searches and file analysis capabilities.

DeepSeek (DeepSeek AI)

Website: https://deepseek.com

DeepSeek is a free Chinese AI application based on its own language model. It can be accessed via a Google account either through a web browser or as a downloadable desktop application. Due to high demand, the service is often overloaded, but it performs well on a variety of tasks. Input can be text or files, and output can be text, code, or mathematical solutions.

NotebookLM

Website: https://notebooklm.google/

NotebookLM is an AI application developed by Google that supports content analysis and creation. It helps users manage, analyze, and structure information from multiple sources while allowing them to take notes. A unique feature of NotebookLM is its ability to generate podcasts based on documents, making content acquisition easier and more versatile.

For more details, see: https://blog.google/technology/ai/notebooklm-goes-global-support-for-websites-slides-fact-check/

SciSpace

Website: https://scispace.com

SciSpace is designed for researchers and students to streamline the research process. Its main feature, SciSpace Copilot, acts as a personal research assistant, helping with literature reviews and summarizing relevant academic papers. Additionally, it supports semantic search, enabling users to quickly find relevant research findings and data.

Scopus AI

Website: https://www.scopus.com/

Starting in 2025, members of the University of Tartu network will have access to Scopus AI, a new feature of Elsevier’s citation database Scopus. Searches can be conducted using questions or hypotheses in any language. Scopus AI generates English-language responses based on article abstracts within the database.

For more information, see https://utlib.ut.ee/en/news/scopus-ai-now-available-ut-members

AI Tutor Pro

Website: https://www.aitutorpro.ca/

AI Tutor Pro is an AI-based learning assistant developed by Contact North, offering the flexibility to learn anytime, anywhere. The application supports multiple languages (Estonian is not yet available) and covers a wide range of topics, including business, history, psychology, science, and technology.