Ivory exactly the use case we should celebrate—scarce human expertise multiplied by scale. On circuit design: equally telling, because it shows how easily knowledge work can be restructured when the bottleneck isn’t intelligence but attention and access. And on the human/LLM reasoning overlap: I’d frame it a little differently. We shouldn’t be surprised that large systems start to rhyme with human cognition; they’re constrained by the same physical universe, the same math of pattern recognition. That doesn’t mean they’re conscious, or even converging on our kind of awareness. It means they’re good mirrors—sometimes sycophantic, sometimes distorted. The through-line for me isn’t “AI looks like us, therefore treat it as us.” It’s “AI reorganizes where human attention, labor, and legitimacy sit.” In ivory, that’s a gift. In jobs, it’s a shock. Thank you for this, Jack.
I don’t think humans and AIs will converge on representations. Stockfish has a much different chess board representation compared to any human. The fact that it has a big table of all solved endgame positions and humans don’t is one example. With large language models, I expect RL to make the representations less human like over time
The correlations in the Dutch study indicate just that... a convergence. But the question is, a convergence of what sort? Will it be a total convergence, 1-to-1 in all aspects? That is certainly doubtful.
However, the point is that: as a convergence is indicated, their correspondence/similarities should be examined further. In this study, evidence has been provided for both a representational AND processing (reasoning) convergence. But do we know what is converging on either the AI or human side EXACTLY? No (the EEG measurements are not exact either). So the point (again) will be to look further. We definitely have a ways to go yet. But it's one step at a time.
The correlations in the Dutch study indicate just that... a convergence. But the question is, a convergence of what sort? Will it be a total convergence, 1-to-1 in all aspects? That is certainly doubtful.
However, the point is that: as a convergence is indicated, their correspondence/similarities should be examined further. In this study, evidence has been provided for both a representational AND processing (reasoning) convergence. But do we know what is converging on either the AI or human side EXACTLY? No (the EEG measurements are not exact either). So the point (again) will be to look further. We definitely have a ways to go yet. But it's one step at a time.
Ivory exactly the use case we should celebrate—scarce human expertise multiplied by scale. On circuit design: equally telling, because it shows how easily knowledge work can be restructured when the bottleneck isn’t intelligence but attention and access. And on the human/LLM reasoning overlap: I’d frame it a little differently. We shouldn’t be surprised that large systems start to rhyme with human cognition; they’re constrained by the same physical universe, the same math of pattern recognition. That doesn’t mean they’re conscious, or even converging on our kind of awareness. It means they’re good mirrors—sometimes sycophantic, sometimes distorted. The through-line for me isn’t “AI looks like us, therefore treat it as us.” It’s “AI reorganizes where human attention, labor, and legitimacy sit.” In ivory, that’s a gift. In jobs, it’s a shock. Thank you for this, Jack.
I don’t think humans and AIs will converge on representations. Stockfish has a much different chess board representation compared to any human. The fact that it has a big table of all solved endgame positions and humans don’t is one example. With large language models, I expect RL to make the representations less human like over time
Actually Steeven...
Convergence is exactly the point.
The correlations in the Dutch study indicate just that... a convergence. But the question is, a convergence of what sort? Will it be a total convergence, 1-to-1 in all aspects? That is certainly doubtful.
However, the point is that: as a convergence is indicated, their correspondence/similarities should be examined further. In this study, evidence has been provided for both a representational AND processing (reasoning) convergence. But do we know what is converging on either the AI or human side EXACTLY? No (the EEG measurements are not exact either). So the point (again) will be to look further. We definitely have a ways to go yet. But it's one step at a time.
Actually Steeven...
Convergence is exactly the point.
The correlations in the Dutch study indicate just that... a convergence. But the question is, a convergence of what sort? Will it be a total convergence, 1-to-1 in all aspects? That is certainly doubtful.
However, the point is that: as a convergence is indicated, their correspondence/similarities should be examined further. In this study, evidence has been provided for both a representational AND processing (reasoning) convergence. But do we know what is converging on either the AI or human side EXACTLY? No (the EEG measurements are not exact either). So the point (again) will be to look further. We definitely have a ways to go yet. But it's one step at a time.
Would love to chat with you Jack. Keen to harness more voice in this space. https://fundresearch.substack.com/p/meet-jake-an-ai-fing-genius-building?r=18mod0