The dual nature of the AI revolution currently unfolding in Silicon Valley is becoming starkly apparent. On one hand, we're witnessing remarkable achievements like Microsoft's AI doctor delivering diagnoses four times more accurate than human physicians. On the other hand, we're seeing the shocking reality of employment rates for computer science graduates from prestigious universities like Stanford and Berkeley plummeting below 50%.
Microsoft's AI Doctor 'MAIDX': A Revolutionary Approach
Microsoft's MAIDX (AI Doctor), announced in June this year, demonstrates a completely different approach from existing AI medical systems. What's most fascinating is its adoption of an "orchestration agent" methodology.
This system operates much like a real hospital where multiple specialists collaborate on a case. One agent identifies the patient's problem, another formulates hypotheses, yet another selects necessary tests, and a final agent creates checklists. These agents engage in discussions, ask each other questions, and work together to reach a final diagnosis.
Technically, it employs a "Chain of Debate" technique. While the traditional "Chain of Thought" involved a single AI reasoning step by step, Chain of Debate involves multiple AIs engaging in expert panel-like discussions to reach conclusions.
The benchmark results were astounding. In tests involving 300 extremely difficult cases from medical literature, general physicians achieved about 20% accuracy, while MAIDX reached 80% accuracy—a four-fold improvement in performance.
Fierce Competition in the Medical AI Space
Microsoft isn't the only player entering this field. Google DeepMind already announced AIM, a medical AI system, in early 2024, showcasing a system that diagnoses through direct multi-turn conversations with patients. The accuracy improves as the conversation progresses.
The reason Silicon Valley's big tech companies are focusing so intensively on healthcare is clear. As AI technology evolves from general-purpose infrastructure to industry-specific applications, healthcare represents the largest market opportunity. Additionally, Silicon Valley moguls' desire for longevity plays a role—since they want to live longer themselves, they're pouring massive investments into this field.
Indeed, many investment firms are concentrating on biohealth sectors. Companies like Noom have successfully gone public with AI model-powered healthcare services.
The Shocking Reality for CS Graduates
However, this AI advancement isn't a blessing for everyone. The current situation in Silicon Valley is quite serious. Reports indicate that employment rates for computer science graduates from top-tier universities like Stanford and Berkeley have dropped below 50%.
In the past, companies would actively court these prestigious university graduates, saying "please come work for us." Offer rates reached 200-300%—meaning each graduate would typically receive 2-3 job offers. Now, less than half are finding employment.
Behind this phenomenon lies the rapid advancement of coding AI. Since AI handles basic coding tasks so well, companies no longer feel the need to hire junior developers. With productivity improvements of 10x, they simply don't need as many people.
Big tech companies now reportedly require hiring requests to include written justification for "why AI can't do this job." If a task can be replaced by AI, they won't approve the hire. They're even laying off existing employees, saying "we don't seem to need this many people."
Meta's Aggressive Talent Acquisition and Industry Upheaval
In this environment, Meta's moves are drawing particular attention. Meta has been aggressively recruiting AI talent with offers worth hundreds of millions of won annually, particularly poaching many talents from OpenAI.
There's a deep background to Meta's strategic shift. While Meta had been operating Facebook AI Research (FAIR) since 2013 and splitting AI research dominance with Google, they began falling behind after ChatGPT's emergence shifted the paradigm to large language models. They tried to catch up with their Llama model, but it fell short of expectations.
Ultimately, Meta decided to essentially disband their existing team and form a separate "super team" suited to the new paradigm. They're boldly rebuilding an organization built over 12 years. Existing employees suddenly found themselves either losing their jobs or having to find other roles.
This aggressive recruitment by Meta poses a major crisis for OpenAI. As a non-profit organization, OpenAI struggles to match Meta's salary offers and continues losing key talent. While Sam Altman has previewed GPT-5, other big tech companies are expected to simultaneously release similar-level models, making it difficult to maintain any special advantage.
The Path to AGI: The Era of Agent Collaboration
Amid these changes, a new perspective on AI development direction is gaining attention. Rather than trying to solve everything with a single massive model, an approach where multiple specialized agents collaborate might be more effective.
Looking at human civilization's development, progress hasn't come from individuals knowing everything, but from various experts communicating and collaborating. AI is likely to develop similarly, combining the time axis (deep reasoning) with the spatial axis (diverse expertise).
For example, when creating a stock investment AI, you might have specialists in stock analysis, market economics, industry expertise, investment strategy, and technical analysis, each playing their role. These would discuss and collaborate to make final decisions, with interactions that consider individual investment preferences (stability-focused, momentum trading, value investing, swing trading, etc.).
This approach resembles the structure of the human brain, where neurons form clusters, each region has unique functions, and they connect to create overall intelligence.
Conclusion: A New Paradigm for the AI Era
The changes we're witnessing aren't simply technological advancement—they represent a paradigm shift. In healthcare, AI is showing performance that surpasses human doctors, while traditional IT jobs are rapidly disappearing.
But this doesn't mean AI will replace everything. Rather, we're entering an era where collaboration between humans and AI, and among multiple AI agents, becomes increasingly important. AI will handle simple, repetitive tasks while humans focus on more creative and strategic roles.
The key is adapting to these changes. The old model of joining large corporations and growing through training programs no longer works. We've entered an era where only those with immediately applicable skills can survive.
As Silicon Valley's harsh reality demonstrates, the AI era is one where opportunities and crises coexist. Rather than fearing change, we need an attitude of actively seeking new possibilities and responding proactively. This is the time for such an approach.