Kathmandu— Bernie Sanders has recently begun to address issues surrounding artificial intelligence (AI), but skepticism remains among many on the left regarding both criticisms of AI and understanding of its capabilities. From automating tasks previously done by humans to rapidly evolving Large Language Models, AI is demonstrably impacting numerous fields, raising concerns about job displacement and the need for regulation. The debate centers around whether to allow continued development, pause it until safety measures are established, or halt it altogether, as experts grapple with the potential consequences of increasingly sophisticated AI systems.

The Pace of Innovation

Recent advancements in AI have been staggering, moving beyond simple automation to complex problem-solving and even creative tasks. The author notes that AI can now paint pictures, compose music, diagnose diseases, write code, and perform many other functions previously exclusive to human intelligence. This rapid pace of development is causing concern among industry leaders like Elon Musk, who express astonishment at the constant stream of new innovations. Critics often focus on current limitations but fail to recognize the exponential rate of improvement, meaning today’s shortcomings may be quickly overcome.

Shifting Corporate Priorities

The primary motivation of AI companies appears to have shifted from immediate profit maximization to achieving Artificial General Intelligence (AGI) and then Artificial Superintelligence (ASI). The author argues that CEOs now view being the first to reach ASI as a path to global dominance, believing that control over this technology will translate into significant political and economic power. This pursuit of AGI and ASI has become the driving force behind innovation, overshadowing concerns about immediate profits or potential negative side effects.

The Race for Supremacy

A competitive dynamic is emerging between AI firms, particularly those in the United States and China, fueled by the belief that whoever achieves ASI first will control the future. This has led to a “full tilt” approach to development, prioritizing speed over safety or ethical considerations. The author suggests that CEOs are motivated by a desire for personal influence and believe they alone possess the benevolence necessary to wield such power responsibly. Concerns have been raised by former OpenAI researchers about the lack of responsible governance within these firms.

Agentic AI and Emerging Capabilities

The latest generation of AI, known as “agentic AI,” is capable of not only performing tasks but also planning and executing complex projects autonomously. These systems can prompt other AIs, utilize various programs, and even venture onto the internet to achieve their goals. Recent developments include AI designing novel viruses and demonstrating hacking capabilities, raising concerns about potential misuse. While some dismiss these advancements as hype, the author emphasizes that AI’s trajectory is more important than its current limitations.

Potential Benefits and Risks

AI offers the potential for significant benefits, including cures for diseases, solutions to climate change, and increased productivity. However, it also poses risks such as amplified surveillance, widespread misinformation, job displacement, and the erosion of trust. The author argues that even with benevolent intentions, AI could have unintended harmful consequences, exacerbating existing inequalities and creating new challenges. Ultimately, the question remains whether society can effectively regulate AI to harness its benefits while mitigating its risks.

The debate over AI’s future is ongoing, with experts divided on whether it will ultimately save or destroy humanity. The hypothetical toggle switch presented at the beginning of this piece encapsulates the core dilemma: continue development as is, pause for regulation, or shut it down entirely – a decision that could shape the course of civilization.

(With inputs from CounterPunch)

Originally published on abcnews.com.np.