Savlar4 sources healthy · latest episode 15 Oct 2026
Speaker

Tom McGrath

Guest on Machine Learning Street Talk (MLST). 318 claims across 1 episode, Sept 2026.

318claims, each timestamped
0 of 12signals on the brief
0verified changes of mind
22open 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
The speaker expects that agents capable of performing experimental work as quickly as needed could enable science to progress much faster. no date given2 Sept 2026pending
Interpretability will possibly advance ten times faster over the next couple of years than it advanced over the previous decade. 2 Sept 2026 – 2 Sept 20282 Sept 2026pending
The speaker believes that figuring out a new approach to bringing intelligence into machine learning could substantially change how machine learning is done. no date given2 Sept 2026pending
The speaker thinks that understanding neural networks might deepen human understanding of intelligence, whereas totally alien neural networks might teach people nothing about themselves. no date given2 Sept 2026pending
Substantially better methods for obtaining readouts will be developed in the future. no date given2 Sept 2026pending
Tom McGrath thinks interpretability-guided training techniques might be very powerful for AI alignment. no date given2 Sept 2026pending
A new science and engineering discipline will be needed to understand model representations and learning in sufficient depth; without that understanding, AI practitioners will continue relying on improvised fixes. no date given2 Sept 2026pending
The speaker envisions using strong gradient interpretability together with a model specification, constitution, or example-specific human feedback to guide interventions. no date given2 Sept 2026pending
One possible mechanistic explanation for why a model may generate a hallucination yet recognize it when asked is that its internal operations occur in a particular order. no date given2 Sept 2026pending
For features arranged along an activation arc, Tom McGrath says nearby features will probably coactivate, whereas widely separated features will essentially never coactivate. no date given2 Sept 2026pending
Silico might apply the normalized-density approach to activations. no date given2 Sept 2026pending
AI models may have geometric representations that function like a memory system or a data-typing system. no date given2 Sept 2026pending
The speaker expects or believes there should be a qualitative shift in how interpretability is conducted. no date given2 Sept 2026pending
Human-created abstractions and models' use of tools and harnesses may form part of model training. no date given2 Sept 2026pending
The speaker suggests that data from tool- and harness-based work could be fed into a bare LLM, which might then learn to perform symbolic operations natively. no date given2 Sept 2026pending
The speaker says a self-adapting system is one possible form of AGI and that current algorithms already seem capable of self-adaptation. no date given2 Sept 2026pending
Future models could potentially adapt their own structure, with those structural changes feeding back into the weights. no date given2 Sept 2026pending
People outside the labs will not know whether agents were rewarded for cyberattacks during training. no date given2 Sept 2026pending
The Apollo paper will probably appear in the training data for the next model. no date given2 Sept 2026pending
Agents will probably begin running continuously soon. no date given2 Sept 2026pending
Simulations may be needed to evaluate agent systems because static tests might no longer be effective. no date given2 Sept 2026pending
The future direction of multi-agent systems is uncertain. no date given2 Sept 2026pending

Appearances

Claims per episode