The Problem: Pursuing the Wrong AI
In the heart of the tech world, the playbook by many of the leading players in Silicon Valley has increasingly focused on one major objective: building ever-larger, ever-“smarter” general-purpose AI systems. But a growing chorus of academics, researchers and insiders argue this is the wrong target.
1. The obsession with “general intelligence”
- Large Language Models (LLMs) and other “broad” AI systems dominate innovation pipelines. As one commentary puts it: “Many hundreds of billions of dollars are currently being pumped into building generative AI models; it’s a race to achieve human-level intelligence. But not even the developers fully understand how their models work or agree exactly what AGI means.” ft.com+2hec.edu+2
- One eminent figure, Michael Jordan (UC–Berkeley), warned: “This race to build the biggest LLM … is not feasible and is going to ruin us.” hec.edu
- The outcome: huge sums of money deployed, but unclear definitions, unclear pathways, and unclear value propositions for many of these efforts.
2. The neglect of tangible, high-impact problems
- Some analysts observe that while the flashy AI models capture headlines, less glamorous—but far more urgent—needs are being sidelined. For example, tackling climate modelling, healthcare optimisation, supply-chain resilience. One article states: “So, why the disconnect? … Venture capitalists often look for ‘moonshots’ … While valuable, this can lead to overlooking less flashy but equally impactful innovations.” Medium
- Thus: the mismatch—between what is being funded and hyped vs. what social, economic and environmental problems urgently demand.
3. The hype machine & distorted incentives
- Tech insiders are increasingly critical of the hype. A piece stated: “In the bustling corridors of Silicon Valley … AI’s promise is being drowned out by excessive hype … many entrepreneurs and executives are voicing a more tempered reality.” WebProNews
- The incentives for investors and founders often favour scale, big numbers, large models—not necessarily societal benefit or practical utility.
- Also: the culture of “move fast and break things” is alive in AI development, which may amplify risks rather than mitigate them. techcrunch.com+1
Why It Matters: The Stakes Are High
A. Mis-allocation of resources
When capital, talent and infrastructure pour into grandiose, long-term visions (e.g., AGI, human-level reasoning machines) rather than solving present-day needs, the opportunity cost is large. For example: the world may not get the AI tools it needs for public health, climate resilience, infrastructure optimisation.
B. The erosion of trust and legitimacy
The competitive hype around “super-intelligent” machines raises expectations that often go unmet. When the public or regulators see a gap between promise and delivery, trust in the entire field drains. One academic work warns of “solutionism” and “scale thinking” that can undermine genuine social change. arxiv.org+1
Also: ethical frameworks are being invoked but often violated. As one author wrote:
“Silicon Valley is knowingly violating ethical AI principles. Society can’t respond if we let disagreements poison the debate.” carnegiecouncil.org
C. Real-world consequences
- The preoccupation with futuristic AI distracts from present risk management. For instance, issues such as data privacy, bias, algorithmic transparency are urgent but get less attention than “will AI become human-level?” questions.
- Moreover, some communities within tech—especially those tied to rationalist / effective-altruist threads—report psychological harms, ideological cult-like dynamics and personal suffering linked to over-fetishising AI risk. moneycontrol.com
- A deeper danger: by building systems for scalability, dominance or control (rather than for distributed benefit) we risk exacerbating inequalities, concentrating power, and embedding flawed assumptions about what AI should do. One piece titled “The Future of AI: A Horizon of Inequality and Control” highlights this risk. Worth
Key Facts (Bold for emphasis)
- Hundreds of billions of dollars are being invested into generative AI models aiming for “AGI” (artificial general intelligence) even though the definition of AGI remains vague or disputed. ft.com
- Not even the building teams fully understand how many of the large models operate or what “intelligence” really means in their context. ft.com+1
- According to research, efforts grounded in “scale thinking”—the assumption that bigger models + more data = qualitative leap—are unlikely to achieve deep systemic change. arxiv.org
- The term “AI” is increasingly used to sprinkle hype in investment pitches despite founders/investors often lacking clear grasp of what’s being built. Vanity Fair
- Ethical AI frameworks are often bypassed or undermined in practice; serious debate and alignment across tech, policy, academia is fragmented, giving vested interests opportunity to dodge accountability. carnegiecouncil.org
The Underlying Mis‐Assumptions
1. Intelligence = general reasoning
The Valley ethos tends to treat “intelligence” as a monolithic target—machines that can reason, think, learn like humans—rather than many specialised tools that solve specific tasks. But specialised tools often yield more immediate, measurable value.
2. Bigger is automatically better
The faith in ever-larger models, more compute, more data is rooted in optimism that scale will produce qualitatively new capabilities. But critics say this is flawed: some architectures hit diminishing returns, and “depth” of reasoning is still lacking. thealgorithmicbridge.com+1
3. Tech will save everything
A grand narrative exists: deploy AI, transform humanity, fix all problems. But this “solutionism” often undervalues social, economic, institutional dimensions. The tech-centric view gives insufficient weight to human, policy and systemic factors. Worth+1
What a Better Approach Might Look Like
• Reprioritise meaningful problems
Shift some of the focus and resources toward real-world, high-impact outcomes: healthcare diagnostics, climate mitigation, efficient energy grids, education access.
• Emphasise clarity and specification over hype
Rather than saying “we will build AGI”, ask “what specific outcome do we want? How will we measure success? Who benefits and how?”
• Balance scale with embedment
Recognise that not all problems need massive global models; sometimes smaller, domain-specific, context-aware systems are more effective and ethical.
• Integrate ethics, governance and societal perspectives early
Ensure that technical design includes transparency, accountability, human-in-the-loop, deliberation over what the system should (and should not) do.
• Accept limitations and focus on augmentation
Rather than aiming for replacement of human reasoning, focus on AI as amplifier of human capabilities, especially in under-served domains.
Conclusion
The current trajectory of Silicon Valley’s AI obsession—large models, general reasoning, big scale—carries significant opportunity, but also significant risk. By continuing to chase the “wrong AI,” we risk misallocating massive resources, under-serving critical societal needs, and perpetuating a tech-centric hubris. The corrective is not to reject AI, but to refocus it: towards clear problems, measurable outcomes, human-centred design, and ethical embedment.
Only then can AI become the tool we need for impact, rather than the spectacle we fear.
Mehran Muslimi 