📊 Full opportunity report: The Ghost Story Became a Forecast. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Clark’s latest analysis presents a 60% probability of automated AI research by 2028, but also highlights a 40% chance of fundamental limitations slowing progress. This shift in perspective influences AI policy and research planning.
Jack Clark’s latest essay reveals a bivalent forecast for AI development, assigning a 60% probability to automated AI research by the end of 2028 and highlighting a 40% chance that current technological paradigms will reveal fundamental limitations, requiring new approaches. This marks a significant shift in the discourse on AI timelines and risks.
Clark’s essay, part of his ongoing series on AI progress, explicitly states a 60% probability that automated AI research will be achieved by 2028, with a 30% chance by 2027 if certain corporate milestones are met. However, he emphasizes a 40% probability that progress will hit a fundamental ceiling before 2028, indicating that current paradigms may be insufficient for further breakthroughs. This latter scenario suggests that the field might discover intrinsic limitations within existing AI architectures, necessitating a paradigm shift.
Clark’s analysis hinges on the idea that the 40% probability is not merely a slower trajectory but a sign that the current approach to AI development is fundamentally flawed. If true, this would transform the research landscape, requiring new theories and possibly years of additional development before achieving similar capabilities. The essay also discusses the significance of the 30% chance of reaching automated AI R&D by 2027, based on corporate commitments and technological milestones, which Clark views as a separate but related probability.
The ghost story
became a forecast.
Reading Clark’s closing — the bivalent 60%/40% credence. The 30% by 2027 alternative. What it means when a frontier-lab co-founder publicly says “I’m persuaded.”
Jack Clark’s closing section — “Staring into the black hole” — contains the most important sentence in the essay for the public discourse. Not the 60%/2028 number — though that’s the technical claim that gets quoted. The discourse-crossing sentence is the personal credence statement: “I have written this essay in an attempt to coldly and analytically wrestle with something that for decades has seemed like a science fiction ghost story. Upon looking at the publicly available data, I’ve found myself persuaded that what can seem to many like a fanciful story may instead be a real trend.”
The standard discourse reads 40% as benign — “slower AI.” Clark’s actual claim is stronger. The 40% reveals a fundamental deficiency within the current technological paradigm. Both outcomes are major findings. The franchise has read the 60% side. The coda reads the 40% side and the bivalence itself.
“For decades, it has seemed like a science fiction ghost story.“
The most important sentence in the essay is not the 60% number. The discourse-crossing sentence is the personal credence statement. When a frontier-lab co-founder publicly says “I am persuaded by the data that this is no longer science fiction,” the discourse changes.
“I have written this essay in an attempt to coldly and analytically wrestle with something that for decades has seemed like a science fiction ghost story. Upon looking at the publicly available data, I’ve found myself persuaded that what can seem to many like a fanciful story may instead be a real trend.”
AI Engineering: Building Applications with Foundation Models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Nine pieces. One structural finding.
Six different forms of evidence aggregating to one structural finding: the labs are building what they say they’re building; the forecast is the plan; the institutional response window is the only variable that remains unfixed.
Six different forms of evidence. One structural finding. The labs are building what they say they’re building. The institutional response window is the only variable that remains unfixed.
AI UNLEASHED: THE PARADIGM SHIFT IN INTERNATIONAL TRADE
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Three paths. All major. All need capacity.
Three structural possibilities for what the next 32 months produce. Asymmetric cost-of-being-wrong points toward building response capacity now. There is no scenario where the capacity goes unused.
~20 months
~32 months
field correction
Capacity built for 30%/60% paths is useful. Capacity built for 40% path is also useful (for field correction). There is no scenario where building response capacity now is wasted.
Clark stares into the black hole and says he’s persuaded. The franchise has been about reading that statement seriously. The reading: he should be. The implication: so should we.
AI & DATA-DRIVEN FASHION DESIGN: A Practical Guide to Trend Forecasting & Creative Innovation with AI: How AI is Helping Fashion Professionals Work … & Design for the Future (Fashion Metaverse)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Implications for AI Development and Policy
This analysis challenges the optimistic view that AI will rapidly reach human-level capabilities within a few years. Instead, it presents a scenario where progress could stall due to fundamental scientific limits, prompting a reassessment of research strategies, investment, and regulation. The recognition of a potential paradigm barrier could lead to increased focus on foundational research and a more cautious approach to deployment and governance.
Moreover, the framing of the 40% as a structural risk rather than a mere delay emphasizes the importance of preparing for a possible paradigm shift. Policymakers, researchers, and industry leaders must consider these probabilities when planning long-term strategies and resource allocations, as the consequences of hitting such a ceiling could be profound and long-lasting.
GEN AI – THE NEXT GENERATION: How Children Will Grow Up In A World Deeply Influenced By Artificial Intelligence
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Clark’s Probabilistic Framework for AI Timelines
In his May 2026 essay, Clark revisits his earlier forecasts, where he previously leaned towards a more optimistic timeline for AI progress. His recent analysis introduces a nuanced, probabilistic view, emphasizing a bivalent outlook: a 60% chance of achieving automated AI R&D by 2028, and a 40% chance of encountering fundamental limitations that could delay or fundamentally alter the trajectory. Clark’s framing builds on prior discussions about compute, data, and architectural bottlenecks, but now explicitly considers the possibility that these constraints are insurmountable within current paradigms.
This shift reflects ongoing debates within the AI community about the sustainability of exponential progress and the potential need for paradigm-changing breakthroughs. Clark’s framing as a ‘ghost story’ that has become a forecast underscores the narrative shift from optimism to cautious realism, emphasizing that the field must prepare for both outcomes.
“The 40% probability signals that we may have uncovered a fundamental ceiling within current AI paradigms, requiring a new approach to move forward.”
— Jack Clark
Unresolved Questions About AI Paradigm Limits
It remains unclear whether the 40% probability reflects an actual fundamental limitation of current AI architectures or if it is a temporary plateau before further breakthroughs. Clark explicitly states that this is a structural risk, but the evidence for such a limitation is still emerging. Additionally, the timeline for potential paradigm shifts—if they occur—is uncertain, with years potentially passing before new approaches materialize.
Further, the impact of external factors such as regulatory changes, technological breakthroughs outside current paradigms, or shifts in investment remains to be seen. The precise implications of hitting a fundamental ceiling are also still being debated within the AI research community.
Monitoring Developments and Preparing for Paradigm Shifts
Key next steps include tracking corporate milestones, such as OpenAI’s and Anthropic’s progress toward automated AI R&D, and observing whether current limitations slow progress as Clark predicts. Researchers and policymakers should prepare for the possibility of a paradigm shift by investing in foundational AI research and developing contingency plans.
Further analysis and discussion are expected as new data from ongoing AI developments emerge, especially around the feasibility of overcoming current limitations. Clark’s forecast encourages a cautious yet proactive approach, emphasizing the importance of readiness for both rapid advancement and potential stagnation or fundamental change.
Key Questions
What does Clark’s 40% probability mean for AI progress?
It indicates a significant chance that current AI development paradigms may hit fundamental limitations, requiring new approaches to make further progress. This could delay or alter the expected timeline for achieving advanced AI capabilities.
How does Clark’s forecast differ from previous predictions?
Earlier forecasts often emphasized rapid progress within a few years. Clark’s recent analysis introduces a nuanced view, emphasizing the structural risk of hitting a ceiling and the possibility that progress may not be purely a matter of scaling compute and data.
What are the implications if the 40% scenario occurs?
If true, it would mean that current AI architectures are fundamentally limited, prompting a shift in research focus, potential years of delay, and a need for new paradigms. It could also influence regulation and investment strategies.
Is there a way to confirm whether a fundamental limit exists?
Confirmation would require ongoing empirical evidence from AI research, including breakthroughs or persistent stagnation despite increased compute and data. The community continues to debate this, and no definitive proof has yet emerged.
Source: ThorstenMeyerAI.com
