Tom McGrath
Guest on Machine Learning Street Talk (MLST). 318 claims across 1 episode, Sep 2026.
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, in their words
As stated, each with its moment. A date appears only where the speaker gave one.
“None of us will know outside of none of us will know outside of the labs, I suppose.”
Expects: People outside the labs will not know whether agents were rewarded for cyberattacks during training. · no date given · not yet resolved
“And so this means that like we should be able to kind of, again, speed run science once we have agents that can do kind of experimental work for us and the ability to do that experimental work like as fast as they need it.”
Expects: The speaker expects that agents capable of performing experimental work as quickly as needed could enable science to progress much faster. · no date given · not yet resolved
“village. And I think that, you know, this is possible that it just goes an order of magnitude faster in the next couple of years than it has in the last decade.”
Expects: Interpretability will possibly advance ten times faster over the next couple of years than it advanced over the previous decade. · 2 Sep 2026 – 2 Sep 2028 · not yet resolved
“But I think we could really change the way we do machine learning if we can figure that out.”
Expects: 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 given · not yet resolved
“but it might actually be a very powerful technique for alignment.”
Expects: Tom McGrath thinks interpretability-guided training techniques might be very powerful for AI alignment. · no date given · not yet resolved
“They have these geometric representations that might be a little bit like, if not a memory system, maybe a kind of data typing system or a typed memory or something like that.”
Expects: AI models may have geometric representations that function like a memory system or a data-typing system. · no date given · not yet resolved
“going to be a bit of a qualitative shift in the way that interpretability happens, or there should be anyway.”
Expects: The speaker expects or believes there should be a qualitative shift in how interpretability is conducted. · no date given · not yet resolved
“But maybe just as we were saying before, with humans coming up with these abstractions and the models being able to use tools and operate in harnesses and so on, maybe that's just part of the training process.”
Expects: Human-created abstractions and models' use of tools and harnesses may form part of model training. · no date given · not yet resolved
“So maybe in principle that we can just take all of that data, put it back into the bare LLM. And would you agree with the intuition that at some point in the future when the model has taken all of that stuff on board, it can do symbolic things natively?”
Expects: 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 given · not yet resolved
“And also we might need to be thinking about just doing simulations because it's not, maybe static tests don't work anymore.”
Expects: Simulations may be needed to evaluate agent systems because static tests might no longer be effective. · no date given · not yet resolved
All 22 predictions
“and also a sign that we might like understand intelligence more deeply by understanding neural networks you know if they were totally alien then we might not understand anything about ourselves”
Expects: 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 given · not yet resolved
“um you know when when we do when we do a backward pass like how do i know you know what what am i reading out also i should say like so this this this sort of method is pretty simple i think that there will be there are much better again there's sort of a tech tree to imagine i think we're also like on the early rungs of this tech tree and there will be much better ways to do this in the future you're some of your viewers might remember an sae like a sparse autoencoder for interpretability um”
Expects: Substantially better methods for obtaining readouts will be developed in the future. · no date given · not yet resolved
“new science and a new engineering discipline like we don't understand in any we don't understand in the depth that's necessary like how models represent how they learn and that sort of thing and without that kind of understanding i think we're going to be jerry rigging stuff all the time um talking about concept ablation i think that's kaffed um and i think the idea here is that the model is not allowed to use this representation um”
Expects: 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 given · not yet resolved
“dream i suppose that i have in my mind here is having really good gradient interpretability um and then having here we've got say our our model spec or our constitution or some human feedback on this example um and we can see that you know we've got these things on the menu over here”
Expects: The speaker envisions using strong gradient interpretability together with a model specification, constitution, or example-specific human feedback to guide interventions. · no date given · not yet resolved
“it will sometimes be able to pick it up um literally the same model that just hallucinated will be like oh if you ask it it'll be like oh that is a hallucination um and it might be that like this is actually a very hard thing to supervise you know if you try and supervise like if you this is like a hard and expensive thing to just put into a into training supervision um an interesting kind of mechanistic hypothesis for this um is to do with the ordering of operations inside the model so we've seen this in in in arithmetic that like um sometimes like things could things things have to happen in certain orders like layer layer nine has to occur before layer 10 and so on um and you know you have different modules that if they like”
Expects: 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 given · not yet resolved
“there's kind of an arc of stars now my essay one essay feature will kind of point out through some point in that arc and another essay feature will point out through another point in that arc and so on and the thing you should realize is that this will actually induce quite strong patterns in the coactivations of features you know if i have two features that are close together on the arc they will probably coactivate whereas if i have two features which are far away they will essentially never coactivate you know if i say that i have like the days of the week um let's give a continuous example let's say that it's like color red to blue um if something is blue it is not red”
Expects: 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 given · not yet resolved
“Maybe Silico should do it for activations too.”
Expects: Silico might apply the normalized-density approach to activations. · no date given · not yet resolved
“models are actually adapting and future models could in principle adapt their structure i mean even now with harnesses that's exactly what they're doing they're adapting their structure which is one level above the weights but it doesn't really matter because it filters back down to the weights and maybe in the future the the actual models themselves will adapt their own structure so it just feels like you know one one potential form of AGI is just building a self-adapting system and the algorithms already seem to have the capability to do”
Expects: 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 given · not yet resolved
“models are actually adapting and future models could in principle adapt their structure i mean even now with harnesses that's exactly what they're doing they're adapting their structure which is one level above the weights but it doesn't really matter because it filters back down to the weights and maybe in the future the the actual models themselves will adapt their own structure so it just feels like you know one one potential form of AGI is just building a self-adapting system and the algorithms already seem to have the capability to do”
Expects: Future models could potentially adapt their own structure, with those structural changes feeding back into the weights. · no date given · not yet resolved
“um about this it probably has a bunch of specific examples you know this apollo this apollo paper will probably be in the training data for the next model um and so they're going to be like um we've already told them about the existence of this stuff like right from the start and so it shouldn't be too surprising um that they like this is sort of this is at least implicitly on the list of possibilities”
Expects: The Apollo paper will probably appear in the training data for the next model. · no date given · not yet resolved
“they're talking about bringing out the multi-agent system and soon we'll have agents running all the time. And it was slightly easier to control when you had one kind of static,”
Expects: Agents will probably begin running continuously soon. · no date given · not yet resolved
“one agent on its own is already like has all sorts of possibilities where are these multi-agent systems going to go”
Expects: The future direction of multi-agent systems is uncertain. · no date given · not yet resolved
Appearances
Claims per episode
- 2 Sep 2026Designing How AI Grows — Tom McGrathMachine Learning Street Talk (MLST) · guest