Near-term AI R&D automation
“AI will automate AI research and reach recursive self-improvement within the next few years (by about 2027-2028)”
25 people · 15 agree · 5 disagree · 10 no clear side · 1 unclear · 54 views · 2026-11-09 to 2026-11-20
Who stands where on the statement above, from their own words. A person's side is shown only where two separate readings of the quote agree; otherwise they are under "Side unclear". This page does not say who is right.
Agree 15 people
- ✓Dwarkesh Patel
If I had to pick my median year for when automated researchers will be as good as the very best human AI researchers (and assuming the labs don't intentionally slow down), I'd guess 2028.
But I think AI research is verifiable enough that AIs will soon be as good at it as the best humans.
If pressed, I’d give a 50% chance to the “spark” scenario and a 50% chance to the “software-only singularity.”
now I think scaling up in-context learning, so that you have sessions that are effectively many months long, will probably be enough to let a single instance learn on the job effectively.
the deliberate effort that AI companies are putting into building a flexible and general intelligence that can ramp up rapidly in any new context should advance much further.
- Quote not reprinted: this page already repeats as much of this article as we allow.
- ✓METR
We think that current public evidence is insufficient to rule out drastic AI R&D acceleration in the next couple of years.
It may in fact be possible to drastically accelerate AI R&D via easy-to-verify tasks alone.
We find that current public frontier models have extremely strong capabilities on optimization tasks, often outperforming skilled humans.
we think it is prudent to consider the possibility that AI may be able to drastically accelerate AI R&D before we see good performance on many ‘realistic’ tasks
- ✓Zvi Mowshowitz
If your plan presumes that it has more time than this then it is not a good plan.
Those types of methods have predicted ~7 of the last zero times that AI has hit a wall and stopped advancing its capabilities.
- ✓Import AI
I believe we’re in the early stages of the singularity, defined as the period when the rate of technological and scientific progress exceeds the rate at which we humans can parse out and make sense of it.
- ✓Ryan Greenblatt
This report on AI R&D automation (RSI) is a big update towards ASI within 1-2 years.
- ✓Vals
Its RSI Index trend points to August 2027 for fully automated research, which may be conservative given capability limits on released models
- ✓AI Futures Project
It now forecasts AI matching top human AI researchers likely in 2027, about 5 months ahead of AI 2027.
- ✓Ray Schroeder
October 2026 marked the beginning of the predicted era of the emerging singularity, in which the agentic systems are capable of fully building their own successors
- ✓Theorem and Anthropic
AI systems are going to become much more capable in coming years
- ✓Ray Kurzweil
some futurists such as Ray Kurzweil have hypothesized that we will someday soon pass through a Singularity
- ✓Ajeya Cotra
this naively implies that we would reach extreme superintelligence within a year.
- ✓Anthropic and OpenAI
as AI becomes better at improving itself, competitive pressures and governance challenges will increase, and a key task will be to ensure humanity remains a part of the process
- ✓Researcher 8
I do think I have more taste in terms of the ideas. But I have no confidence that will continue to be true for any significant period of time.
- ✓Dwarkesh Patel
- Quote not reprinted: this page already repeats as much of this article as we allow.
- ✓Others in the AI community
others contend that models trained on easy-to-verify tasks will eventually be able to successfully complete open-ended tasks as well
Disagree 5 people
- ✗Anthropic and OpenAI
So long as model performance on AI R&D-related tasks is heterogeneous, lower-performing categories and tasks could become bottlenecks to high overall uplift.
the ability to develop and vet conceptual research ideas (research taste) may be a bottleneck for recursive self-improvement
- ✗Cal Newport
I believe that the idea of RSI leading to an intelligence explosion is highly speculative and based on many unproven assumptions.
We cannot simply assume that any science fiction scenario we conjure about powerful AI is possible if we give machines enough time to self-improve.
- ✗Dean W. Ball
Progress in "taste," on the other hand, while nonzero, has been small by comparison.
- ✗Vernor Vinge
rapidity in cognition is not the same as intelligence. Speeding up the mind of a dog, he argued, wouldn’t enable the animal to play chess.
- ✗Sayash Kapoor
he thinks that research taste—the ability to choose promising questions, hypotheses, and experiments—will prove a major bottleneck in automating AI R&D
No clear side 10 people
- ~Zvi Mowshowitz
I don’t think full RSI and passing the relevant points of no return is a favorite to happen in 2027, but at this point I would no longer be surprised if that happened.
I also think there has been too much updating.
Do not make the mistake of treating such predictions as assurances.
Contra the AI Futures Project I believe an SAR in 2027 is under 50% to happen, but not so far under 50%.
- ~Dean W. Ball
allowing AI to control the development of its successors with no or little practical human oversight—essentially, handing over the keys of America’s largest AI companies to AI—is a bad idea.
But do the risks of weak recursion justify banning the practice of using AI as an input to make more and better AI? In my view, they do not.
- ~Tim Genewein
Whether RSI leads to ‘runaway’ progress is largely an empirical question.
RSI should advance fastest in domains that allow for automated verification without humans in the loop, such as mathematical proofs, while other domains might lag behind or taper off sooner.
- ~MIT Technology Review
That’s why AI systems have become so good at coding and math, two domains where it’s possible to determine automatically whether a solution does or doesn’t work.
it’s precisely this type of task that will prove essential if AI systems are to automate research and development.
- ~Dean W. Ball
allowing AI to control the development of its successors with no or little practical human oversight—essentially, handing over the keys of America's largest AI companies to AI—is a bad idea.
- ~Theorem and Anthropic
AI that can build itself would be a major development in the history of technology—one that could bring enormous good for the world in science, healthcare and beyond.
- ~Dwarkesh Patel
- Quote not reprinted: this page already repeats as much of this article as we allow.
- ~Former Anthropic employee
one company should not be able to “unilaterally” decide to pursue RSI
- ~David Sacks
Recursive self-improvement is the most dangerous thing anybody can be working on.
- ~METR
One possibility that we find plausible is that doubling times may tend to be shorter on less diverse sets of tasks.
Side unclear 4 people
- ·METR
The most straightforward explanation for this is that models have poor judgment about what makes a good task.
- Quote not reprinted: this page already repeats as much of this article as we allow.
we would consider weakness in ‘judgment’ and related capabilities to be reasonable candidates for what most delays drastic AI R&D acceleration.
- Quote not reprinted: this page already repeats as much of this article as we allow.
- ·Shelly Banjo
signs are emerging that the tech industry is — like its most famous product — developing the capacity for recursive self-improvement.
- ·Anthropic and OpenAI
the historical trends and metrics we report (which involve significant human-AI collaboration) may not be predictive of future AI-driven progress
- ·Tim Genewein
RSI loops can easily stall or degenerate if the feedback guiding them is poor.
Week by week
- · 15: Shelly Banjo ·, Anthropic and OpenAI ✗, Researcher 8 ✓, Zvi Mowshowitz ~, Dwarkesh Patel ✓, Dean W. Ball ~, Former Anthropic employee ~, David Sacks ~, Cal Newport ✗, Vernor Vinge ✗, METR ✓, Tim Genewein ~, Sayash Kapoor ✗, Others in the AI community ✓, MIT Technology Review ~
- · 11: Import AI ✓, Zvi Mowshowitz ~, Ryan Greenblatt ✓, Dean W. Ball ~, Vals ✓, AI Futures Project ✓, Ray Schroeder ✓, Theorem and Anthropic ~, Ray Kurzweil ✓, Dwarkesh Patel ✓, Ajeya Cotra ✓
Built 2026-11-21 06:50 UTC from the news knowledge
graph by code. Every quote was checked against the source's own words; which side a quote takes is a model's reading,
made twice.