Generative AI: Reshaping the World of Technology
Generative Artificial Intelligence(AI) is a new form of AI technology that breaking the barrier of technological revolution. It is a sub-field of AI that focuses on generating new data rather than analyzing and driving context out of the data. With the recent hype around ChatGPT, AI-powered artwork or the protein-folding breakthroughs, the popularity of generative AI tools has been significantly increased due to the simplicity of their user interfaces for creating high-quality text, graphics and videos in a matter of seconds.
The underlying concept of generative AI is not brand-new as the first chatbot, which demonstrated this technology, was introduced by MIT professor Joseph Weizenbaum in the 1960s when he developed Eliza. However, we were not able to make much advancement in this field the introduction of generative adversial networks (GAN) in 2014. GANs are an generative approach to create compelling texts, images or videos using deep learning methods, such as convolutional neural networks (CNN).
How does Generative AI work?
Generative AI starts with a prompt that could accept text as an input and then various AI algorithms return new content in response to the text prompt. The output content can include essays, answer to a user question, or realistic fake images/audios created from pictures or audio of a person. Recent version of generative tools provide better user experiences as that let you define a request in plain language.
Two recent AI advancement have played a huge role in promoting generative AI to mainstream: Tranformers and Large Language Models (LLMs). Transformer is a deep learning model that adopts the mechanism of self-attention which unlocks model potential to track the connections between words across pages, chapters and books rather than just in individual sentences. Therefore, it is used primarily in the fields of natural language processing (NLP) and computer vision (CV). It also enables to train ever larger models without having to label all of the data in advance. New models could thus be trained on billions of pages of text, resulting in answers with more depth. Large language models (LLMs) — i.e., models with billions or even trillions of parameters — have opened a new era in which generative AI models can write engaging text, paint photorealistic images, and even create entertaining videos on the fly. We have already noticed a huge jump in GPT model training parameters over the course of a few years. GPT3 parameter was 100 times bigger than it’s predecessor (175 billion vs 1.5 billion). The figure below illustrates the parameter growth of the popular transformers based LLMs that revolutionized the field of NLP.
Advantages and limitations of Generative AI
Generative AI showed potential to become a ground-breaking technology that can change the paradigm of industry practices. It can influence the implementation of automated workflow the can potentially benefit users in the following cases:
- automate the manual process of writing content or general powerpoint presentation decks
- reducing the effort of responding to emails by implementing chatbots
- improving the response to specific technical queries and minimize efforts of exploring technical information on internet
- summarizing complex information into a coherent narrative that can be easily comprehensible
- simplifying the process of creating digital content for businesses
The implementation of generative AI has not been flawless thus far. Even with it’s capability of generating realistic human-like texts or landscapes, the AI generated text summary or art work often raised eyebrows due to the following limitations:
- failure to identify the source of content in most cases
- generate misleading information as a result of hallucination impact
- challenging to assess the biasness of original sources and models.
- unlocking concerns about deepfakes, digitally forged images or videos, that can easily be manipulated by the cyber attackers to spread inaccurate information leading to social issues, prejudice and hatred.
The rise in popularity of tools like OpenAI’s ChatGPT and DALL-E has accelerated the utilization of generative AI and opened up a gateway to implement more business use cases aligning with industry standard. However, we need to be careful about the ethical usage of this new AI technology and alleviate pitfalls from mishandling by criminals. According to CB Insights , the total investment for the generative AI companies set a new landmark in 2022, with equity capital surpassing $2.6 billion over 110 separate agreements. Please keep an eye on six generative AI companies such as OpenAI, Hugging Face, Lightricks, Jasper, Glean, Stability AI.
