Naci Cankaya

Karma: 104

Naci Cankaya 5 Mar 2026 14:20 UTC
1 point
0
in reply to: jbash’s comment on: Make Powerful Machines Verifiable
Construct the circuits? Can you explain what you mean by that?

Naci Cankaya 26 Feb 2026 12:58 UTC
3 points
0
in reply to: Gordon Seidoh Worley’s comment on: All hands on deck to build the datacenter lie detector
Yeah, I didn’t say this, but it would be great if there was a dedicated startup! The tech would necessarily need to be open source, but support could be the business model. That, or the startup could simply be non-profit and funded by Longview or similar:
https://www.longview.org/request-for-proposals-on-hardware-enabled-mechanisms-hems-for-ai-verification/
Things are moving fast though, and by next week I’ll know more about some specific company’s ambitions and roadmaps. Can share this when this info is public.

Regardless, more commercial engagement is needed.

Naci Cankaya 20 Oct 2024 17:50 UTC
9 points
0
in reply to: Nathan Helm-Burger’s comment on: [Linkpost] Hawkish nationalism vs international AI power and benefit sharing
Hello Nathan, thank you for your comment! Really looking forward to your post and how it expands on the idea.

Regarding your estimates: I personally do not think that assigning probabilities to preferable outcomes is very useful. On the contrary, one can argue that the worldviews held by influential people can become self fulfilling prophecies. That is especially applicable to prisoner’s dilemmas. One can either believe the dilemma is inevitable and therefore choose to defect, or instead see the situation itself as the problem, not the other prisoner. That was the point we were trying to make.

Rather than suggesting what to do under the assumption that things turn out well, we were arguing for a shift in perspective and what measures could help with increasing the chances of a good outcome of the advent of TAI. Fixing the game, instead of accepting the rules and hoping for the best.

In the disclaimer, when we said that the relevance of the article increases with longer timelines, we did not mean to narrow its applicability down towards a scenario of an agreed upon AI pause. The purpose was mainly to not appear too confident in our assumptions before being proven wrong by unpredictable impacts of the technology.
What links here?
- A path to human autonomy by Nathan Helm-Burger (29 Oct 2024 3:02 UTC; 53 points)