Revolution of Generative AI & Foundation Models: Beyond ChatGPT and OpenAI

Generative AI is a branch of artificial intelligence that focuses on creating new, hyper-realistic content rather than merely analysing or classifying existing data. The generated content can span different modalities, such as text, images, audio, video, code, or even synthetic datasets. Model parameters such as temperature, top_p, and top_n control the degree of randomness and, consequently, the creativity of the responses generated by these systems. In simple terms, generative AI is about teaching machines to imagine and create.

 

“Siri”, once a widely recognised household name, is now being complemented, and in some cases surpassed, by generative AI models such as ChatGPT, DALL·E, Gemini, and Stability AI, which have become synonymous with this new era of artificial intelligence.

 

Recently, Albania reportedly appointed its first AI minister, Diella, to oversee public procurement with the aim of reducing corruption in government tenders (source: Times of India).

 

We are witnessing this monumental technological shift largely because of the underlying architectural paradigm of generative AI: Foundation Models (FMs). These models are pre-trained on an unprecedented scale of diverse, largely unlabelled data collected from the internet. Through this process, they learn a vast range of patterns, relationships, and structures across multiple modalities (text, images, and more). Once pre-trained, these models can be fine-tuned for specific downstream applications with relatively small amounts of task-specific data. In essence, a foundation model can be viewed as a highly educated generalist that can quickly become an expert in many specialised domains.

 

Unfolding the Impact of Generative AI and Foundation Models: A SWOT Analysis

 

Strengths

 

The following are some of the key value-driving factors of this technology:

1. Personalisation
   Generative AI enables highly personalised experiences. For example, educational content can be adapted in real time to match a student’s learning style, while marketing campaigns can be tailored to millions of consumers simultaneously during a product launch.
2. Drug Discovery
   Scientists can use generative AI to hypothesise new drug compounds, design novel materials, and simulate complex biological processes. This capability significantly accelerates research cycles and innovation in the pharmaceutical and biotechnology sectors.
3. Creative Partnerships
   Foundation models can assist in writing novels, designing video games, generating art, and supporting other creative endeavours. Rather than replacing human creativity, they act as valuable creative collaborators that augment human potential.
4. Agentic AI
   We are now witnessing the rise of AI agents that can not only generate content but also plan and execute complex tasks, interact with software systems, and make autonomous decisions. These agents function as intelligent collaborators capable of achieving defined goals.

 

Weaknesses

 

High resource requirements

Training and operating these large models require significant computational resources, energy, and specialised expertise. As a result, access to such technologies remains limited for many organisations and stakeholders. This challenge also highlights the growing importance of sustainable AI practices.

 

Opportunities

 

Generative AI is transforming many professions rather than simply eliminating them. As a result, numerous new specialised roles have emerged, enabling organisations to effectively harness the potential of this technology.

 

Threats

 

While generative AI offers remarkable capabilities, it also introduces several challenges that the global ecosystem must address:

 

1. Biased Responses
   Foundation models can inherit biases from their training data. Continuous efforts are therefore required to implement robust content moderation tools to ensure fairness, prevent discrimination, and mitigate harmful outputs targeting particular social groups.
2. Hallucination Risks
   Foundation models may confidently generate factually incorrect or nonsensical information, commonly referred to as hallucinations. This creates a need for stronger mechanisms for verification, validation, and fact-checking of AI-generated content.
3. Decline in Human Creativity
   Excessive reliance on generative AI tools, particularly among younger users, may lead to a reduction in independent creative thinking and problem-solving skills.
4. Intellectual Property and Copyright
   The use of vast datasets for training raises complex legal and ethical questions related to fair use, creator attribution, and the ownership of AI-generated content.
5. Misinformation and Misuse
   Generative AI can be used to produce highly realistic deepfakes, propaganda, and phishing content, posing serious societal and security risks.

 

Conclusion

 

Looking ahead, progress in generative AI will require not only technological advancements but also thoughtful deliberation, ethical development, and responsible governance. Ensuring that these powerful tools benefit society in a fair and equitable manner is essential. Building a responsible, innovative, and accessible generative AI ecosystem is therefore a shared responsibility that extends across governments, industry, academia, and society as a whole.

 

Dr Alamelu Mangai Jothidurai
Professor
Presidency School of Computer Science & Engineering
Presidency University