Select an episode
Not playing

2017–2024: Transformers to ChatGPT

A paper on “attention” unlocks generative AI. Text, images, and code appear on command; jobs and creativity are reimagined. Bias, deepfakes, and data scraping spark ethics battles. EU AI Act, U.S. orders, and safety summits draw new lines.

Episode Narrative

In the vibrant tapestry of technological advancement, the years from 2017 to 2024 stand out as a pivotal chapter, marked by innovations that not only transformed industries but also the very fabric of human interaction. It begins in 2017, when a paper titled “Attention Is All You Need,” penned by Vaswani and his colleagues, introduced the Transformer architecture to the world. This groundbreaking neural network design would soon become the foundation upon which almost all advanced generative AI systems were built. With an unparalleled ability to process and generate human-like text, images, and code, it opened the gates to a future where machine intelligence would echo human creativity in ways previously thought impossible. The dawn of this new era began quietly, but its impact would ripple across the globe.

Fast forward to 2018, when OpenAI released GPT-1, the inaugural model in a series of generative pre-trained transformers. It represented a significant step forward, showcasing that large language models could produce coherent blocks of text. Yet, the application was limited. While GPT-1 dazzled researchers, its capabilities were viewed through a lens of cautious optimism. What lay ahead was uncertain, but the foundations had been laid.

Then, in 2020, we witnessed a monumental leap with the introduction of GPT-3. With an astounding 175 billion parameters, this model astonished the tech world by generating essays, poetry, and even snippets of code, blurring the line between human and machine creativity. Conversations shifted; the discourse around authorship, bias, and misinformation took center stage. As GPT-3's capabilities became known, voices of caution emerged alongside the awe. How would humanity grapple with machines that could mimic our thoughts, our emotions, even our artistic expressions?

In 2021, another landmark was reached. DALL-E, a creation by OpenAI, enabled users to generate detailed images from mere text prompts. This democratization of visual art unleashed a wave of creativity that invited almost everyone to become an artist, albeit with a machine's help. Questions regarding copyright and originality intensified, signaling the complexities of a new world where machines could conjure imagery from imagination alone.

Then came November 2022, a time that forever altered our daily lives. ChatGPT launched and within two short months, it attracted 100 million users, making it the fastest adoption of a consumer application in history. As it became an everyday tool for writing, coding, tutoring, and entertaining, people found themselves engaging with AI in ways they had never envisioned. The lines blurred further between human decision-making and machine-generated suggestions, posing new challenges and opportunities for society.

Yet, the journey forward was not without its storms. In 2023, deepfake technology, bolstered by generative AI, emerged with a conviction that sent shivers down the spines of media outlets and governments alike. The capability to create convincingly altered reality raised alarms about potential disruptions to elections, the spread of disinformation, and a notable erosion of public trust in visual evidence. We stood on the precipice of a world where seeing could no longer mean believing.

Amidst these developments, the European Union finalized the AI Act — a framework that set strict regulations for high-risk applications within the realm of artificial intelligence. This marked a significant turning point in how society would engage with these technologies, emphasizing the necessity for transparency and accountability among tech firms. The stakes were high, for as innovation surged, so too did the need for governance.

By 2024, the U.S. government issued an executive order concerning AI safety. Developers of powerful AI systems were mandated to share safety test results, reinforcing standards for cybersecurity and consumer protection. It was a clear acknowledgment: the responsibility to safeguard humanity now rested heavily on the shoulders of those creating these transformative tools.

Then came the overwhelming saturation of AI-generated content on social media. Estimates indicated that over 60% of posts on platforms were now created or significantly augmented by AI. This phenomenon transformed online culture, altering the nature of discourse and the economics of attention. The digital landscape had been irrevocably changed, reshaping how we interacted with information and, in turn, with each other.

The effects were profound. Major labor markets began to witness tangible displacement of white-collar jobs, particularly in writing, translation, customer service, and graphic design. Companies rushed to adopt AI tools aimed at reducing costs and enhancing productivity. As opportunities vanished for some, a new profession emerged — the “prompt engineer.” These individuals became experts in communicating with AI systems, crafting precise inquiries to yield the desired outputs. Their value surged, reflecting a rapid evolution in the job market.

In the UK, the AI safety summit brought together global leaders, tech CEOs, and researchers to confront the existential risks posed by advanced AI. It became a defining moment — a crucial crossroads in international governance aimed at establishing a framework for this technology's future. The stakes were high, as misinformation campaigns fueled by AI technologies tested the resilience of democratic institutions in places like India, the U.S., and across the European Union. Fact-checkers and content platforms struggled to keep pace, highlighting a tension between the rapid advancement of technology and the mechanisms in place to govern it.

The culture wars around AI continued to grow. Audits revealed systemic racial, gender, and ideological biases inherent in major language models. Groups began calling for more inclusive training data and algorithmic transparency, striving to ensure these tools could serve everyone fairly. Alongside these ethical considerations arose a flurry of lawsuits. Artists, writers, and publishers sought justice against tech companies for using their works without consent to train AI models. The question of ownership in a digitally interconnected world had never seemed more urgent.

In the face of all this, AI-generated music and voice cloning tools began to dominate the cultural landscape, allowing anyone to create songs in the styles of famous artists or impersonate public figures. Yet, this innovation came with a whirlwind of new copyright disputes and ethical dilemmas. The power of creation now came with the responsibility of consideration.

As 2024 progressed, the narrative around AI shifted once more. The once-looming “AI winter” — a narrative suggesting a slowdown in AI advancement — faded as investments in generative AI startups exceeded $50 billion. Tech giants and governments raced to harness this next phase of the digital revolution. Daily life increasingly became mediated by AI, with personalized news feeds and smart devices anticipating our needs. This new culture, characterized by algorithmic intimacy, also fostered a passive consumption of information.

As we approached 2025, projections hinted at a future where nearly all economically valuable tasks might be automated, challenging the very notion of work as we have come to know it. This speculation left many wondering: what becomes of human agency in a world increasingly defined by the actions of machines?

The years from 2017 to 2024 encapsulated a remarkable journey. From the innovation of Transformers to the global reach of ChatGPT, this era introduced technologies that profoundly changed our reality. Yet, each breakthrough also illuminated the delicate balance we must strike — between innovation and responsibility, creativity and ethics, free exploration and careful regulation.

As we look towards the horizon, one question echoes: what will the legacy of this time be? Will we embrace the possibilities, or will we navigate the complexities that lie ahead? In a world increasingly shaped by intelligent machines, the choice is ours.

Highlights

  • 2017: The paper “Attention Is All You Need” by Vaswani et al. introduces the Transformer architecture, a neural network design that becomes the foundation for nearly all advanced generative AI systems, enabling machines to process and generate human-like text, images, and code at unprecedented scale and quality. (Primary source: arXiv:1706.03762)
  • 2018: OpenAI releases GPT-1, the first in a series of generative pre-trained transformer models, demonstrating that large language models can produce coherent paragraphs of text, though with limited practical application at this stage. (Primary source: OpenAI blog)
  • 2020: GPT-3, with 175 billion parameters, stuns the tech world by generating essays, poetry, and even simple code snippets, blurring the line between human and machine creativity and raising urgent questions about authorship, bias, and misinformation. (Primary source: arXiv:2005.14165)
  • 2021: DALL-E, also from OpenAI, enables users to generate detailed images from text prompts, democratizing access to AI-powered visual art and sparking debates over copyright, originality, and the future of creative professions. (Primary source: OpenAI blog)
  • 2022: ChatGPT launches in November, reaching 100 million users in just two months — the fastest adoption of a consumer application in history — and becomes a daily tool for writing, coding, tutoring, and entertainment worldwide. (Primary source: Reuters, Statista)
  • 2023: Deepfake technology, powered by generative AI, becomes so convincing that major media outlets and governments issue warnings about its potential to disrupt elections, spread disinformation, and erode public trust in visual evidence. (Primary source: Brookings Institution, Wired)
  • 2023: The European Union finalizes the AI Act, the world’s first comprehensive legal framework for artificial intelligence, setting strict rules for high-risk applications, transparency, and accountability, with significant implications for global tech firms. (Primary source: European Commission)
  • 2024: The U.S. government issues an executive order on AI safety, mandating that developers of powerful AI systems share safety test results with the federal government and adhere to new standards for cybersecurity and consumer protection. (Primary source: White House)
  • 2024: AI-generated content saturates social media, with estimates suggesting over 60% of posts on some platforms are now created or augmented by AI, transforming online culture, discourse, and the economics of attention. (Primary source: Pew Research Center)
  • 2024: Major labor markets begin to see measurable displacement of white-collar jobs — especially in writing, translation, customer service, and graphic design — as companies adopt AI tools to reduce costs and increase productivity. (Primary source: McKinsey Global Institute)

Sources

  1. https://s-rsa.com/index.php/agi/article/view/15119
  2. https://managementworld.online/index.php/mw/article/view/1087
  3. https://www.nepjol.info/index.php/jtha/article/view/80915
  4. https://sijmds.com/index.php/pub/article/view/66
  5. https://medicaljournalssweden.se/actaoncologica/article/view/42567
  6. http://biorxiv.org/lookup/doi/10.1101/2022.06.28.497781
  7. https://reu.termedia.pl/Assessing-cardiovascular-risk-in-rheumatoid-arthritis-patients-on-Janus-kinase-inhibitors,207507,0,2.html
  8. https://medicalforummonthly.com/index.php/mfm/article/view/5359
  9. https://aacrjournals.org/cancerres/article/85/8_Supplement_1/3375/757252/Abstract-3375-Demographic-characteristics-and
  10. https://dl.acm.org/doi/10.1145/3701716.3719148