2024 will be remembered as the year artificial intelligence (AI) transcended its hype to showcase its multiple practical applications. From Apple’s groundbreaking entry to Nvidia's revolutionizing chips, this year marked the transition from theoretical possibilities to practical realities, changing how people work, create, and solve problems.
Interestingly, one of the most underappreciated aspects of AI advancements has been the phenomenon known as AI "hallucinations"—the generation of plausible yet false information. While the general consensus considers these outputs troubling, numerous scientists are beginning to see their potential for driving innovation and discovery. According to The New York Times, these unexpected outputs from AI models are proving surprisingly useful across various fields including medicine and climate science.
Scientists assert this 'creative aspect' of AI is invaluable particularly during the early stages of scientific discovery, where ideas often come from hunches and guesswork. Amy McGovern, computer scientist and director of a federal AI institute, stated, "The public thinks it's all bad. But it's actually giving scientists new ideas. It's giving them the chance to explore ideas they might not have thought about otherwise." This perspective has prompted researchers to reevaluate the role of AI outputs amid scientific advancements.
AI hallucinations not only spark creative exploration but also expedite the hypothesis generation process, leading to accelerated breakthroughs. MIT professor James J. Collins noted how these outputs have advanced his research on novel antibiotics, highlighting how "We're exploring. We're asking the models to come up with completely new molecules." Such insights point toward the transformative impact of AI hallucinations within the research environment.
Meanwhile, in rehabilitation, Dr. Michelle Ploughman, Canada Research Chair at Memorial University, focused on creating more effective treatments for individuals suffering from multiple sclerosis (MS) and strokes. Her collaboration with computer science expert Dr. Xianta Jiang has leveraged AI to analyze data from patients with MS through the application of 100,000 sensors embedded within specialized walking pads.
This technological advancement has revealed subtle gait differences between healthy individuals and those with MS, changes not easily detectable by the human eye. Ploughman articulated, "If you cannot measure it, you cannot improve it," emphasizing how precise measurements open opportunities for preemptive care and treatment strategies.
Ploughman initially viewed AI with apprehension but now considers it a powerful tool for research, stating, "I now see AI as another tool in my toolbox. Humans are still needed to ask the questions and find the right data, but we need computing far beyond what human capability allows." This shift signifies the increasing integration of interdisciplinary collaboration fueled by AI, piquing interest among the medical community.
The significance of academic collaboration has garnered national attention. Research Infosource recently reported Memorial University ranked first among Canadian universities employing AI to advance medical science. This finding reinforces AI's growing role within research frameworks, as institutions increasingly embrace the technological capabilities of AI systems.
The momentum continued as AI entered the consumer sector. Apple’s eagerly awaited launch of Apple Intelligence introduced several generative AI features, integrated with enhanced user experience protocols—a process underscoring the brand's commitment to privacy. Nevertheless, the introduction was complicated by several high-profile misinformation incidents, with false attributions causing distress among major news organizations. Reporters Without Borders underscored this concern when stating, "This accident demonstrates generative AI services are still too immature to produce reliable information for the public and should not be allowed on the market for such uses."
Despite this setback, Apple showcased potential AI applications through initiatives to improve user engagement and data processing. The field of AI applications seemed limitless, with Nvidia’s Blackwell chip representing another industry breakthrough. Its capability to handle generative AI workflows remarkably with 25 times lower energy consumption compared to its predecessor illustrated the shift toward efficiency and green technology.
The rollout of Nvidia's innovations encountered some initial obstacles, described by CEO Jensen Huang as “100% Nvidia’s fault,” leading to delays. Nonetheless, the impact was felt significantly across the industry, prompting companies like Microsoft and Meta to scramble for acquisitions, affirming how hardware advancements can be just as consequential as software innovations.
Then came Claude 3.5 Sonnet from Anthropic, which outperformed its predecessors through enhanced reasoning capabilities. The integration of features aimed at collaborative AI workflows provides glimpses of future interactions between humans and AI systems—one where AI can handle complex tasks autonomously.
The implementation of the EU AI Act also captured attention. Regarded as the world’s first comprehensive framework for regulating AI use and development, the Act necessitated companies to rethink their approach, adapting to new classifications and compliance requirements. James White, CTO at Calypso AI, noted, "The impact of organizations operating within the EU will likely depend on how the company is using the AI model and what risk category the use case falls under," demonstrating how regulatory shifts can influence the industry's direction.
Although the Act came with controversy due to potential loopholes, it solidified the necessity for accountability and transparency within the burgeoning AI sphere. OpenAI followed suit with its launch of the o1 model, revolutionizing reasoning processes through new APIs—offering users unprecedented control within scientific and programming capabilities.
With the rise of AI, societal concerns about data rights and ethical implementation took stage, especially highlighted by the rapid growth of Perplexity AI. Transitioning from startup to major player, its success was accompanied by significant debates over content ownership, prompting major legal battles.
This year remains pivotal, delineated by the movement toward specialized, thoughtful AI applications. Breaking barriers within traditional boundaries, fostering interdisciplinary collaboration, and introducing regulations reshaping the industry emerged as key themes underscoring the work required for innovation and responsibility.
