COVID 19 and the emergence of Generative AI in 2020

The emergence of generative AI in 2020 and the COVID-19 pandemic were indirectly connected, marking a time of great change in society and in the use of advanced technologies. Here I detail how the two phenomena were related and how they impacted the development and adoption of these tools:

Pandemic as a catalyst for technological change.

• Accelerating digitization: During the confinement, the need for digital solutions skyrocketed. Many companies and users sought tools that facilitated telecommuting, distance learning and task automation, which created a favorable environment for technologies such as generative AI to begin to come to the fore.

• Demand for automated content: Digital content creation became crucial. Tools such as those based on GPT-3 (launched in June 2020) began to be used to generate articles, emails and automated content, optimizing processes at a time when teams were scattered and overloaded.

Advances in Generative AI in 2020

• Launch of GPT-3: OpenAI introduced GPT-3, one of the first widely known generative AI models, which demonstrated advanced capabilities in coherent and contextual text generation. This marked a turning point in how people and businesses perceived the potential of AI.

• Growing interest in automation tools: Companies in different sectors, such as marketing, education, and healthcare, began to explore the use of generative AI to handle workloads at a time when human resources were limited.

• GPT-3: The language revolution. Launched in June 2020, GPT-3 (Generative Pre-trained Transformer 3) represented a significant leap in text generation. This model, with 175 billion parameters, could perform complex tasks such as writing articles, answering questions and even generating programming code. During the pandemic, it was used in multiple applications:

• Customer service automation: Companies integrated GPT-3 into chatbots to handle large volumes of COVID-19-related queries.

• Educational content production: Institutions leveraged its ability to generate clear, personalized explanations for students in remote education.

The Role of Generative AI in the Face of COVID-19

During the pandemic, generative AI not only facilitated content creation, but also helped address critical issues related to health crisis management.

Scientific information processing

The volume of COVID-19 research grew exponentially in a short period. Generative AI models were used to:

• Analyze scientific data: they helped identify patterns in massive publications and generated abstracts accessible to clinicians and researchers.

• Propose hypotheses: By generating combinations of data, AI facilitated innovative approaches in the search for treatments and vaccines.

Communicating with the public

Governments and organizations implemented AI-powered virtual assistants to answer frequently asked questions, which helped alleviate pressure on healthcare lines. These tools, based on generative models, could provide quick and accurate answers about symptoms, preventive measures and local restrictions.

Handling disinformation

Although generative AI was also used to create fake news, its potential to combat disinformation stood out. Models such as GPT-3 were programmed to detect inconsistencies in content and generate reliable information.

Social changes and perception of AI.

• Increased exposure to technology: During confinement, dependence on technological tools increased. This caused AI, including its generative applications, to become a topic of general interest, moving from being a “niche” technology to an everyday occurrence.

• Creativity during confinement: Many people took advantage of time at home to experiment with generative AI tools to write stories, program applications or create art, as was the case with Jukebox (generative music) or DALL-E (launched in 2021, but developed during that period). Giving way to the other generative AIs.

Since 2020, generative artificial intelligences have evolved rapidly, especially in the areas of text, image, audio and video generation. Here's a summary of the highlights in each category:

1. Text Generation

• GPT-3 (2020): Developed by OpenAI, this AI revolutionized the generation of consistent and understandable text in multiple languages, being widely used for chatbots, content writing and more.

• ChatGPT (2022): Based on GPT-3.5 and GPT-4, it became popular as an accessible conversational tool for the general public.

• Claude (2023): Created by Anthropic, focused on security and generating text with greater accuracy and ethics.

• LLaMA (2023): Meta model designed for computationally efficient text generation and analysis tasks.

• Bard (2023): Developed by Google, combines generative models and advanced search tools.

2. Image Generation

• DALL-E (2021): From OpenAI, capable of generating realistic images from detailed textual descriptions.

• Stable Diffusion (2022): Open source model that democratized AI image creation by allowing custom modifications.

• MidJourney (2022): With an artistic approach, produces high-quality images with exceptional detail.

• Image (2022): Google model with emphasis on accuracy of visual representation and realism.

3. Audio Generation

• Jukebox (2020): From OpenAI, generates music in various styles from textual descriptions.

• VALL-E (2023): Created by Microsoft, synthesizes high-quality vocals with only a few seconds of sample time.

• Riffusion (2022): Generates music in real time using spectral imaging techniques.

4. Video Generation

• Runway Gen-1 and Gen-2 (2023): Models that allow the creation and editing of generative videos from text or images.

• Make-A-Video (2023): Meta project that creates short, realistic videos based on textual descriptions.

5. Code Generation

• Codex (2021): From OpenAI, the engine behind GitHub Copilot, facilitates automatic code generation for multiple programming languages.

• StarCoder (2023): Focused on developers, generates and explains complex code snippets.

6. Multimodal Generation (Text + Images/Video)

• DeepMind Gato (2022): Model that can generate text, images and even control robots on the same platform.

• Gemini (2023): From Google DeepMind, combines language capabilities with visual generation in an advanced model.

Key lessons from 2020

• Pandemic acted as a catalyst for the adoption of generative AI by accelerating digital transformation and increasing demand for automated solutions.

• The success of technologies such as GPT-3 and other generative AI showed the versatility of these tools in both practical and creative applications.

• Ethical challenges also emerged, such as the misuse of AI-generated content, highlighting the need for regulation and responsible development.