Why IBM CEO Arvind Krishna is still hiring humans in the AI era

Key Takeaways

IBM is an enterprise-focused company, not consumer-facing, emphasizing B2B AI and hybrid cloud solutions.
The company evolved its AI strategy from early monolithic Watson to modular LLM-based Watson X for enterprise use.
IBM predicts significant AI cost reductions, potentially 30 times cheaper within five years, driven by hardware and software advances.
IBM is making substantial, long-term investments in quantum computing, projecting a multi-billion-dollar market.
CEO Arvind Krishna believes current LLMs are robust for enterprise but unlikely to achieve Artificial General Intelligence (AGI).
AI is expected to displace up to 10% of U.S. jobs due to productivity gains, while also enabling new roles.
IBM leverages an internal AI coding tool, boosting employee productivity by 45% in four months, leading to market share growth.

Nilay Patel questioned if IBM's early AI efforts with Watson, which won Jeopardy in 2011, caused them to miss the current generative AI boom.
IBM CEO Arvind Krishna admitted Watson's early application in healthcare was 'inappropriate' but affirmed the value of its foundational research.
Krishna expressed optimism that the current AI situation is not a bubble, comparing it to the dot-com era's eventual realization of online computing's promise.

The guest acknowledged current high AI costs for data centers and GPUs, noting undefined ROI for some expenditures.
IBM projects AI cost reduction will take time, with a potential 10x advantage in semiconductor capability over five years.
Factors driving down costs include new processor designs from companies like Groq and Cerebras, alongside software optimizations, potentially leading to a thousandfold decrease over time.
The guest specifically projects AI costs could become 30 times cheaper within five years due to chip designs and software efficiencies.

The conversation compared economic incentives for acquiring AI users to past trends seen in social media and fiber optic infrastructure development.
The guest drew a parallel between current AI infrastructure investment and the fiber optics boom of the early 2000s, where underlying infrastructure endured despite many associated business model failures.
Krishna noted that the higher failure rate of current AI compute hardware, such as GPUs, is considered an acceptable trade-off for increased performance.

Arvind Krishna discussed the early internet economy's excitement and eventual pop, drawing parallels to the current AI boom and its potential to transform the economy.
He emphasized the significant physical economy, involving frontline workers, warehouses, and logistics, is becoming more efficient with AI technologies.
The conversation anticipated a massive increase in application development due to AI, comparing its disruptive potential to the impact of the smartphone ecosystem.

IBM's strategy under CEO Arvind Krishna focuses on a hybrid cloud approach and AI deployment specifically for enterprise clients, avoiding direct competition in consumer AI.
The company is shifting away from commodity services to high-value areas like AI and quantum computing.
This strategic pivot involved significant organizational changes, including spin-offs and aligning sales incentives, to drive growth above GDP.

The guest explained their decision-making framework, emphasizing client benefit as a core driver for strategic choices.
They triangulate information from internal stakeholders, clients, and external partners to validate assumptions before making significant company decisions.
This process helps in forming convictions and informing execution, building on personal strengths as a technologist.

IBM's investment in quantum computing acknowledges that utility-scale computing is not yet a reality, with initial five years primarily an internal bet.
Since 2020, IBM has engaged 300 clients in research and commercial capacities, exploring quantum use cases in bond trading, materials science, and aerodynamics.
IBM made its quantum software open source, attracting 650,000 global users and validating demand for new problem-solving approaches.
The guest expects quantum processing units (QPUs) to complement, not displace, CPUs and GPUs, opening new markets, with a high probability of remarkable results in 4-5 years.

The guest contrasted the AI market's consumer-focused bets, like OpenAI's rumored $1 trillion IPO, with IBM's enterprise approach to AI and quantum computing.
They noted Microsoft's deal with OpenAI reportedly expires upon Artificial General Intelligence (AGI) declaration, questioning the return on investment if AGI is undefined or not achieved.
Krishna expressed skepticism about current LLM paths achieving AGI, estimating a 0-1% chance, believing AGI requires fusing knowledge with LLMs.
While LLMs are robust for enterprise use and promise trillions in productivity, they are seen as statistical approaches lacking deterministic or knowledge-based components.

The guest predicted up to 10% job displacement in the U.S. due to increased AI-driven productivity, concentrated in roles that can be automated.
They suggested that companies may hire more in different roles as overall productivity increases, reaching a stable equilibrium.
Arvind Krishna highlighted IBM's internal experience where a self-developed AI coding tool increased productivity by 45% for 6,000 employees in four months, enabling more product development rather than headcount reduction.