PhD student in theoretical computer science, doing research on why (graph) algorithms can be much more efficient in practice than suggested by worst-case theory.
I have signed no contracts or agreements whose existence I cannot mention.
wilm
Karma: 33
Sitting under the roof in the small garden, I’m wearing a green hoodie jacket
It looks a bit too rainy so let’s go to the cafe close by https://maps.app.goo.gl/msxgfRPqzRXHf4g96?g_st=com.google.maps.preview.copy
small edit: I removed the location from this post because of the contract with the venue; if you want to come just ask me about the location!
I can add to that that there are still (quite some) spots available due to them renting out the whole youth hostel. I highly encourage applying.
Careful: this event page is a duplicate of https://www.lesswrong.com/events/JWWaknqcT3zJHrdNu/acx-schelling-meetup-6
The location is confirmed :)
Sith
small typo
Location: (49.0142290, 8.4038544) https://maps.app.goo.gl/qrzC1mseUChj2cvFA?g_st=ic
I’ve unfortunately been quite distracted, but better a late reply than no reply.
With capabilities I mean how well a system accomplishes different tasks. This is potentially high dimensional (there can be many tasks that two systems are not equally good at). Also it can be more and less general (optical character recognition is very narrow because it can only be used for one thing, generating / predicting text is quite general). Also, systems without agency can have strong and general capabilities (a system might generate text or images without being agentic).This is quite different from the definition by Legg and Hutter, which is more specific to agents. However, since last week I have updated on strongly and generally capable non-agentic systems being less likely to actually be built (especially before agentic systems). In consequence, the difference between my notion of capabilities and a more agent related notion of intelligence is less important than I thought.
Thanks for your replies. I think our intuitions regarding intelligence and agency are quite different. I deliberately mostly stickest to the word ‘capabilities’, because in my intuition you can have systems with very strong and quite general capabilities, that are not agentic.
One very interesting point is that you : “Presumably the problem happens somewhere between “the smartest animal we know” and “our intelligence”, and once we are near that, recursive self-improvement will make the distinction moot”. Can you explain this position more? In my intuition building and improving intelligent systems is far harder than that.
I hope to later come back to your answer to information about the real world.
I’m already there, wearing a red T-shirt
new location for meet-up: https://maps.google.com/maps?q=49.013498146132875%2C8.40058685281165
Alex talk will be postponed to next week. Status of today’s meeting is still uncertain.
There might be rain / thunderstorm. In this case we will meet at a private place. PM me if you are new, but still want to join.
New location because of too many people https://goo.gl/maps/6UWhUZVLLazzf7UeA (The group with the large unicycle)
Due to server problems the link is now https://public.senfcall.de/karlsruheunterrationality
The meeting will take place on jitsi meet (meet.jit.si) in the room KarlsruheLW20201124
Thanks for the report, I very much appreciate this condensed summary of your recent papers. I have three questions:
1. Can you explain your reasoning and degree of belief regarding whether your research does contribute to general AI capabilities progress? From my perspective, investigating belief geometry is very foundational and it is uncertain whether it will in the end contribute more to AI safety or to capabilities.
2. What are directions for future work that you want to investigate, or that you think are promising but don’t think you will investigate yourself?
3. As I saw the job posting on your website: do you think people with a theoretical computer science / algorithms PhD with experience in probability and geometry can meaningfully contribute?
[In my previous research I have been modeling relevant structural properties of real-world networks to derive theoretical explanations for the surprising practical efficiency of certain graph algorithms. In particular, I have combined rigorous proofs and rigorous math with implementation and the empirical study of practical instances.]