Savlar4 sources healthy · latest episode 20 Oct 2026
Speaker

John Schulman

Guest on Dwarkesh Podcast. 121 claims across 1 episode, Sept 2026.

121claims, each timestamped
0 of 12signals on the brief
0verified changes of mind
27open predictions

Positions over time

Every pair of statements that was compared. Only a verified change counts as a change of mind; rewording, added nuance and different questions do not.

No two statements by this speaker have been compared yet.

Predictions

As stated. A window appears only where the speaker gave one.

ExpectsBySaidStatus
Each new AI model is expected to catch up with humans in some of the areas where humans currently have advantages over AI models. no date given11 Sept 2026pending
The cycle of initial excitement about new AI models followed by disappointment might continue; the number of repetitions is difficult to predict. no date given11 Sept 2026pending
The cycle of excitement about new AI models followed by disappointment may repeat more times than people expect. no date given11 Sept 2026pending
AI will eventually perform substantial research analysis and theory-building, using compute comparable to the compute spent on experiments. no date given11 Sept 2026pending
As AI capabilities advance, defining objectives and deciding what people want will remain the human role that lasts longest. no date given11 Sept 2026pending
Humans will continue to define desired AI behavior and objectives—including helpfulness, human-feedback objectives, constitutions, and model specifications—even if AI systems can perform all technical work. no date given11 Sept 2026pending
The need for humans to determine the right objective in AI alignment will persist for the foreseeable future. no date given11 Sept 2026pending
Companies may use company-specific models that learn from deployment and are continually improved by the company. no date given11 Sept 2026pending
Company-specific, deployment-learning models could be supplied either by the current oligopoly of model providers or by a currently smaller company. no date given11 Sept 2026pending
The speaker expects company-specific models that learn from deployment to change the competitive landscape somewhat. no date given11 Sept 2026pending
The first AI systems capable of automating AI research and development will probably combine human feedback that captures researchers' preferences with many practice environments for multi-step research projects. no date given11 Sept 2026pending
AI developers will likely combine human feedback and practice environments, addressing the most apparent weakness in each successive iteration. no date given11 Sept 2026pending
As researchers use AI systems, they will identify recurring weaknesses and address them through human feedback or purpose-built training environments. no date given11 Sept 2026pending
Sufficiently capable in-context learning could let models avoid finance-specific training by learning from books on the fly and adapting their work to the relevant jurisdiction. no date given11 Sept 2026pending
A superficial reward signal, such as whether a user accepts a code edit, might be reward-hacked. no date given11 Sept 2026pending
In some medium-length learning regimes, humans may update their weights more sample-efficiently than models. no date given11 Sept 2026pending
Models will probably continue to grow in size as available compute scales up and GPUs become larger. no date given11 Sept 2026pending
As high-quality pre-training data becomes scarce, data efficiency may matter more than compute efficiency in determining the extent of model growth, which depends in non-obvious ways on scaling laws. no date given11 Sept 2026pending
Whether model sparsity will keep increasing without limit; it is uncertain whether this will happen. no date given11 Sept 2026pending
There may be an optimal level of model sparsity, rather than model sparsity increasing indefinitely. no date given11 Sept 2026pending
AI will be extremely good at solving hard search problems if trained for those tasks. no date given11 Sept 2026pending
A version of the described AI remote-worker capability will exist in about a year, perform some tasks very well and others less well, and improve thereafter. 11 Sept 2027 – 11 Sept 202711 Sept 2026pending
AI may already have achieved the proposed productivity uplift in some types of work. no date given11 Sept 2026pending
AI researchers may achieve a 10x productivity uplift in about two years. 11 Sept 2026 – 11 Sept 202811 Sept 2026pending
The speaker estimates that AI will reach the level of capability being discussed in three to four years. 11 Sept 2029 – 11 Sept 203011 Sept 2026pending
AI will take longer to become capable at tasks involving 3D, spatial, or physical work. no date given11 Sept 2026pending
AI may take longer to master fields such as mechanical engineering when those fields are not receiving much attention. no date given11 Sept 2026pending

Appearances

Claims per episode