In the year 2024, the initial hype surrounding generative AI will start to fade away as the setbacks of this technology become more evident. While generative AI has shown promise in terms of improving productivity in various tasks, it has also been plagued by issues such as false information and hallucinations.
As evidence mounts, it becomes clear that generative AI and large language models are prone to providing inaccurate and made-up information. The hope that supervised learning could quickly solve the problem of hallucination proves to be optimistic at best. The architecture of these models, which focuses on predicting the next word or sequence of words, makes it extremely challenging to anchor the predictions to verified truths.
Expectations of exponential improvements in productivity and the realization of artificial general intelligence (AGI) also fall short. Instead, the blame for productivity failures shifts towards businesses’ improper implementation of generative AI. The focus then shifts to identifying the specific human tasks that can be enhanced by these models and determining the additional training required for workers to make this a reality.
It becomes increasingly clear that achieving complex human cognition solely based on predicting words was nothing more than a pipe dream. However, there will still be believers who claim that achieving true intelligence is just around the corner. Unfortunately, many will overlook the issues at hand and continue to fear the “existential risks” of AI, while ignoring the more significant risks of uncontrolled AI deployment, such as job displacement, inequality, and threats to democracy.
In 2024, the adoption of generative AI will spread across numerous companies but will prove to be mediocre automation at best. While it may lead to job displacements, it fails to deliver the anticipated significant productivity improvements. The primary applications of ChatGPT and other large language models will be in social media and online search, where platforms will capitalize on personalized digital ads and intensify competition for user attention. Consequently, the prevalence of manipulation and misinformation online will continue to grow, leading to increased screen time and associated mental health issues.
Although there will be more AI startups, the industry will ultimately be dominated by a duopoly, with Google and Microsoft/OpenAI leading the field with their massive models. Other companies will be compelled to build their own apps on these foundation models, but the prevalence of false information and hallucinations will result in many disappointing outcomes.
Calls for antitrust actions and regulations will become louder, but antitrust efforts will likely fall short due to a lack of courage among courts and policymakers. However, there will be growing discussions around new laws and regulations, indicating a shift towards more bipartisan consideration of meaningful regulation. The US government’s failure to keep up with the technology will be more apparent in 2024, driving the need for catching up and addressing the challenges posed by generative AI.