felipeerias

Exactly, the clock is external to the model. Nothing prevents it from being faster or slower, or even running backwards, because it’s ultimately just another data point in the input stream to a computer function.

Your brain and your whole body exist in time. Even when you are asleep, your body does not flicker out of existence and your brain actually continues working during that time.

What, exactly, would be the link between you as you are right now, and “you” in a different body?

Hardware is fungible. Each LLM response in a conversation could be served from a different machine.

Would you be "you" in a different body?

IMHO the sane position is essentially the Aristotelian one.

Hylomorphism: body and consciousness are intrinsically linked. The nature of that link is an open metaphysical question.

Virtue ethics: even if LLMs are not conscious, we should not abuse or mistreat them them because cruelty practised on anything trains one's disposition toward cruelty.

Mistreatment and abuse, even when directed at a machine, make you a worse person.

Even if you are only interested in getting good results out of them, LLMs tend to work better when they are immersed in a narrative of open collaboration.

Mind-body dualism is not real. Even in your example, you would be building upon a minimal part of a person's body.

Your brain is part of your body, that's the point. There isn't a "you" separate from your actual, physical existence. Mind-body dualism is not real.

A stronger version of that argument is that LLMs are not intrinsically affected by the passage of time.

Input stream comes in, input stream comes out. The LLM doesn't care whether this happens once a minute or once a year.

What are you, ultimately, if not your body?

One of the mathematicians in the video describes the process as:

> the AI has been able to explore all these possibilities much more comprehensibly, and doing that it found a path, it found a way to the solution.

Finding a counterexample of a mathematical conjecture strikes me as not that different from finding a vulnerability in a complex codebase.

The other side of this is that open source projects that allow AI tools will be more restrictive towards new contributors.

This already happens to some degree on large software projects with corporate backing (Web engines, compilers, etc.), where it is often not trivial to start contributing as an independent individual.

Reasonable people can disagree on whether one approach is inherently better than the other, as ultimately they seem to be optimising for different goals.

Claude 4.7 broke something while we were working on several failing tests and justified itself like this:

> That's a behavior narrowing I introduced for simplicity. It isn't covered by the failing tests, so you wouldn't have noticed — but strictly speaking, [functionality] was working before and now isn't.

I know that a LLM can not understand its own internal state nor explain its own decisions accurately. And yet, I am still unsettled by that "you wouldn't have noticed".

Nowadays Japan’s fertility rate is higher than most of its neighbours. We are just used to pick it as an example because it started aging earlier than most other countries.

Japanese population is still over 120 million. Forecasts put it falling below 100 million at some point in the second half of this century.

Things will have to change in order to keep population stable in the long term, but the Japanese approach seems IMHO more sensible than that of other countries.

Cohesive democratic societies are fragile.

From the article:

> Our tests gave models the vulnerable function directly, often with contextual hints (e.g., "consider wraparound behavior").

Anthropic gave the model the whole codebase and told it to find a vulnerability on a specific file, iterating across sessions focusing on different files.

What happens then is that, for example, the model looks through that particular file, identifies potential problems, and works upwards through the codebase to check whether those could actually be hit.

“Hum, here we assume that the input has been validated, is there any way that might not be the case?”

This is not unique to Mythos. You can already do this with publicly available models. Mythos does appear to be significantly more capable, so it would get better results.

The research discussed here provided models with just a known buggy function, missing the whole process required to find that bug in the first place.

That’s besides the point because most whales were killed in the XX century.

People had been hunting whales for centuries, but industrialisation gave them the means and the motivation to do so until near extinction.

The US implemented severe immigration restrictions in the 1920s that were lifted gradually over the 1950s–1960s.

Several European countries have already fallen in this trap. As pensioners comprise an increasingly large fraction of voters, pandering to them becomes far more politically attractive than investing in the future.

A living brain exists physically, changes over time, and never stops working.

A brain cut from its body and frozen its a dead brain.

A LLM is not intrinsically affected by time. The model rests completely inert until a query comes in, regardless of whether that happens once per second, per minute, or per day. The model is not even aware of these gaps unless that information is provided externally.

It is like a crystal that shows beautiful colours when you shine a light through it. You can play with different kinds of lights and patterns, or you can put it in a drawer and forget about it: the crystal doesn’t care anyway.

LLMs are disembodied and exist outside of time.

Bundle of tokens comes in, bundle of tokens comes out. If there is any trace of consciousness or subjectivity in there, it exists only while matrices are being multiplied.

If they bundled together these two radically different usage patterns, either the service would become more expensive or the limits would become a lot tighter, in both cases making Claude Code far less attractive to professional users.

One could carefully calculate exactly how much a given document in the training set has influenced the LLM's weights involved in a particular response.

However, that number would typically be very very very very small, making it hard to argue that the whole model is a derivative of that one individual document.

Nevertheless, a similar approach might work if you took a FOSS project as a whole, e.g. "the model knows a lot about the Linux kernel because it has been trained on its source code".

However, it is still not clear that this would be necessarily unlawful or make the LLM output a derivative work in all cases.

It seems to me that LLMs are trained on large FOSS projects as a way to teach them generalisable development skills, with the side effect of learning a lot about those particular projects.

So if I used a LLM to contribute to the kernel, clearly it would be drawing on information acquired during its training on the kernel's code source. Perhaps it could be argued that the output in that case would be a derivative?

But if I used a LLM to write a completely unrelated piece of software, the kernel training set would be contributing a lot less to the output.

The obvious solution is for Anthropic et al. to certify the skills of each user:

> “Good at explaining requirements, needs handholding to understand complex algorithms, picky with the wording of comments, slightly higher than average number of tokens per feature.”

I’m not saying this would be good at all, but the data (/insights) and the opportunity are clearly there.

Planning might end up being more reliable thanks to coding agents: if you want to estimate how long a task would take, just send an agent to do it.

If the agent comes back in a few minutes with a tiny fix, it is probably a small task.

If the agent produces a large, convoluted solution that would need careful review, it is at least a medium task.

And if the agent gets stuck, runs into architectural constraints, etc. then it is definitely a hard task.

Most people seem to be expecting some kind of quantitative analysis: N developers undertook M tasks with and without access to a given AI tool, here is the statistical evidence that shows (or fails to show) the effect, and this result is valid across other projects and tools.

In practice, arriving at this ideal scenario can be very challenging. Actually feasible experiments will be necessarily narrow, with the expectation that their results can be (roughly) extrapolated outside of their specific experimental setup.

Another valid approach would be to carry out qualitative research, for example a case study. This typically requires the study of one (or a few) developers and their specific contexts in great detail. The idea is that a deep understanding of how one person navigates their work and their tools would provide us with insights that might be related to our specific situation.

Personally, in this particular area, I tend to prefer detailed qualitative accounts of how other developers are working on similar projects and with similar tools as me.

But in any case, both approaches are valid and complementary.

As someone looking at this from outside the US, the whole sequence of events is frankly terrifying.

I fear that frontier AI is going to be nationalised for military purposes, not just in the US but across the globe.

At the same time, I really don’t know what Anthropic were expecting when they described their technology as potentially more dangerous than an atom bomb while agreeing to integrate purpose-built models with Palantir to be deployed in high-security networks for classified military tasks.

It’s a bit more complex than that, but to be fair I don’t know what they were expecting after they integrated a purpose-built model with Palantir to be deployed in high-security networks to carry out classified tasks.

Reportedly, Anthropic didn't know about Claude's role in capturing Maduro until they saw it on the headlines.