In the years to come, 2023 might be remembered as the year of unparalleled hype surrounding generative AI, with ChatGPT becoming the fastest-spreading new technology in history. However, 2024 will be the year of recalibrating expectations. While generative AI offers significant opportunities to enhance productivity in various tasks, the discrepancy between the hype and reality will lead to disappointing setbacks.

Generative AI and large language models, despite their impressive capabilities, will continue to provide false information and be prone to hallucination, where AI invents information and gets it wrong. Efforts to address this issue through supervised learning will prove challenging. The architecture of these models, which relies on predicting the next word or words in a sequence, makes it difficult to anchor predictions to established truths.

Expectations of exponential productivity improvements across the economy or the advent of artificial general intelligence (AGI) will be tempered. Instead, the conversation will shift towards understanding which human tasks can be enhanced by these models and the necessary training for workers to leverage them effectively.

The notion of achieving complex human cognition through predicting words will be increasingly questioned. Some will continue to believe that intelligence is just around the corner, while others will miss the real issues, focusing on the "existential risks" of AI instead of the more immediate risks it poses to jobs, inequality, and democracy.

In 2024, generative AI will be adopted by many companies, but it will likely result in "so-so automation" that displaces workers without delivering significant productivity gains. ChatGPT and other large language models will be widely used in social media and online search, leading to increased manipulation, misinformation, and screen time, accompanied by mental health issues.

AI startups will proliferate, and the open-source model will gain traction. However, this will not prevent the emergence of a duopoly, with Google and Microsoft/OpenAI dominating the field. Many companies will be forced to rely on these foundation models for app development, leading to disappointing results due to false information and hallucinations.

Calls for antitrust action and regulation will intensify, but neither courts nor policymakers will have the courage to break up the largest tech companies. Regulation efforts will be slow, as the US government struggles to catch up with the technology. The need for new laws and regulations will become increasingly apparent, leading to bipartisan discussions.

2024 will be a pivotal year for AI, as the hype surrounding its capabilities gives way to a more realistic understanding of its limitations and risks. It will be a time for reflection, recalibration, and the development of strategies to address the challenges that AI presents to society.