Tobias H

Ap­ply to the Pivotal Re­search Fel­low­ship (AI Safety & Biose­cu­rity)

• Tobias H 21 Aug 2023 20:28 UTC

  1 point

  on: clearthis’s Shortform

  I recently found a 2-hour interview with Mo Gawdat titled “EMERGENCY EPISODE: Ex-Google Officer Finally Speaks Out On The Dangers Of AI!” from 2 months ago. It has 6.5 million views which is 1 million more views than Sam Altman’s interview on Lex Fridman.
  
  I’ve only skimmed through it, but it seems that Gawdat frames AI developments as an emergency and is mainly concerned about potential misuse.
  
  I likely won’t have time to actually listen to it for a while – but it seems pretty relevant for the AI Safety community to understand what kind of other extremely popular media narratives about dangers from AI exist.

• Tobias H 3 Aug 2023 13:54 UTC

  6 points

  2

  on: Work cul­ture creep

  I think the overall point you’re making is intriguing, and I could see how it might alter my home behaviour if I considered it more deeply. But I also strongly disagree with the following:

  Just about every work behavior is an example of bad home behavior

  There is a bunch of “work behavior” that has been very useful – in the right measure – for my personal life:

  • Task Management – This cut down on the time I spend on “life admin”.

  • Scheduling – Reaching out with “let’s find an evening to play tennis” helps me increase the number of fulfilling activities I do with friends.

  • Prioritization – Day-to-day life can obscure what’s really important. Thinking about what I really value and want to achieve can make my life more meaningful.

  • Creating Spreadsheets & Documents – Apart from the obvious use case in personal finance, Spreadsheets are also very valuable to me for evaluating crucial life decisions (“Where should we move?”). I use documents for private events I’m organising (e.g. a weekend trip to the mountains with friends).

  Maybe some of them are too obvious and common. But they are things that my grandmother wouldn’t have done – and I suspect that they are mostly derived from work culture.

• Tobias H 27 Jul 2023 18:23 UTC

  3 points

  0

  on: What are ex­am­ples of some­one do­ing a lot of work to find the best of some­thing?

  DC Rainmaker does incredibly detailed reviews of sports watches.

An­nounc­ing the Swiss Ex­is­ten­tial Risk Ini­ti­a­tive (CHERI) 2023 Re­search Fellowship

• Tobias H 30 Jan 2023 16:14 UTC

  1 point

  0

  on: All AGI Safety ques­tions wel­come (es­pe­cially ba­sic ones) [~monthly thread]

  How useful do people think it would be to have human-in-the-loop (HITL) AI systems?
  
  What’s the likeliest failure mode of HITL AI Systems? And what kind of policies are mostly likely to help against those failure modes?
  
  If we assume that HITL AI Systems are not going to be safer once we reach unaligned AGI. Could it maybe give use tools/​information/​time-steps to increase the chance that AGI will be aligned?

• Tobias H 29 Sep 2022 2:27 UTC

  2 points

  0

  in reply to: jacobjacob’s comment on: How I buy things when Light­cone wants them fast

  I was thinking of money. :)

• Tobias H 26 Sep 2022 22:39 UTC

  12 points

  0

  on: How I buy things when Light­cone wants them fast

  Interesting. Can you give us a sense of how much those asks (offer to pay for the extra labour) end up costing you?

• Tobias H 4 Jul 2022 13:57 UTC

  2 points

  0

  in reply to: MondSemmel’s comment on: Toni Kurz and the In­san­ity of Climb­ing Mountains

  Also: The last name is “Von Almen” not “Almen”

• Tobias H 18 Jun 2022 8:19 UTC

  35 points

  in reply to: AlphaAndOmega’s comment on: Scott Aaron­son is join­ing OpenAI to work on AI safety

  I always assumed that “Why don’t we give Terence Tao a million dollars to work on AGI alignment?” was using Tao to refer to a class of people. Your comment implies that it would be especially valuable for Tao specifically to work on it.
  
  Why should we believe that Tao would be especially likely to be able to make progress on AGI alignment (e.g. compared to other recent fields medal winners like Peter Scholze)?

• Tobias H 15 Jun 2022 9:08 UTC

  1 point

  0

  in reply to: Tobias H’s comment on: AGI Safety FAQ /​ all-dumb-ques­tions-al­lowed thread

  Just saw the inverse question was already asked and answered.

• Tobias H 15 Jun 2022 9:05 UTC

  1 point

  0

  in reply to: Yonatan Cale’s comment on: AGI Safety FAQ /​ all-dumb-ques­tions-al­lowed thread

  [I think this is more anthropomorphizing ramble than concise arguments. Feel free to ignore :) ]
  
  I get the impression that in this example the AGI would not actually be satisficing. It is no longer maximizing a goal but still optimizing for this rule.
  
  For a satisficing AGI, I’d imagine something vague like “Get many paperclips” resulting in the AGI trying to get paperclips but at some point (an inflection point of diminishing marginal returns? some point where it becomes very uncertain about what the next action should be?) doing something else.
  
  Or for rules like “get 100 paperclips, not more” the AGI might only directionally or opportunistically adhere. Within the rule, this might look like “I wanted to get 100 paperclips, but 98 paperclips are still better than 90, let’s move on” or “Oops, I accidentally got 101 paperclips. Too bad, let’s move on”.

  In your example of the AGI taking lots of precautions, the satisficing AGI would not do this because it could be spending its time doing something else.

  I suspect there are major flaws with it, but an intuition I have goes something like this:

  • Humans have in some sense similar decision-making capabilities to early AGI.

  • The world is incredibly complex and humans are nowhere near understanding and predicting most of it. Early AGI will likely have similar limitations.

  • Humans are mostly not optimizing their actions, mainly because of limited resources, multiple goals, and because of a ton of uncertainty about the future.

  • So early AGI might also end up not-optimizing its actions most of the time.

  • Suppose we assume that the complexity of the world will continue to be sufficiently big such that the AGI will continue to fail to completely understand and predict the world. In that case, the advanced AGI will continue to not-optimize to some extent.

    • But it might look like near-complete optimization to us.

• Tobias H 12 Jun 2022 15:27 UTC

  6 points

  0

  on: AGI Safety FAQ /​ all-dumb-ques­tions-al­lowed thread

  • Would an AGI that only tries to satisfice a solution/​goal be safer?

  • Do we have reason to believe that we can/​can’t get an AGI to be a satisficer?

• Tobias H 20 Jan 2022 7:44 UTC

  4 points

  in reply to: TAG’s comment on: “A Harm­ful Idea”

  There has been quite a lot of discussion over on the EA Forum:
  https://​​forum.effectivealtruism.org/​​search?terms=phil%20torres
  
  Avital Balwit linked to this lesswrong post in the comments of her own response to his longtermism critique (because Phil Torres is currently banned from the forum, afaik):
  https://​​forum.effectivealtruism.org/​​posts/​​kageSSDLSMpuwkPKK/​​response-to-recent-criticisms-of-longtermism-1#6ZzPqhcBAELDiAJhw

• Tobias H 17 Jan 2022 19:59 UTC

  19 points

  in reply to: cousin_it’s comment on: Try­ing to Keep the Gar­den Well

  The whole thing was much more banal than what you’re imagining. It was an interim-use building with mainly student residents. There was no coordination between residents that I knew of.
  
  The garden wasn’t trashed before the letter. It was just a table and a couple of chairs, that didn’t fit the house rules. If the city had just said “please, take the table out of the garden”, I’d have given a 70% chance of it working. If the city had not said a thing, there would not have been (a lot of) additional furniture in the garden.
  
  By issuing the threat, the city introduced an incentive they didn’t intend.
  Some residents who picked up on the incentive destroyed the garden because they were overconfident in the authority following through with the threat – no matter what.

Try­ing to Keep the Gar­den Well

• Tobias H 15 Jan 2022 15:39 UTC

  1 point

  on: How Bad Is QWERTY, Really? A Re­view of the Liter­a­ture, such as It Is

  The link to the QWYRFM layout contains an error.

• Tobias H 6 Jan 2022 10:56 UTC

  11 points

  in reply to: Rob Bensinger’s comment on: An­i­mal welfare EA and per­sonal dietary options

  I would guess that humans’ nightmarish experience in concentration camps was usually better than nonexistence; and even if you suspect this is false, it seems easy to imagine how it could be true, because there’s a lot more to human experience than ‘pain, and beyond that pain, darkness’.

  I can’t really imagine this – at least for people in extermination camps, who weren’t killed. I’d assume that, all else equal, the vast majority of prisoners would choose to skip that part of their life. But maybe I’m missing something or have unusual intuitions.

• Tobias H 17 Dec 2021 13:11 UTC

  1 point

  in reply to: Connor_Flexman’s comment on: COVID and the holidays

  Thank you! The general reasoning makes sense to me.
  This Cochrane review finds a false negative for asymptomatic individuals of 42% with antigen tests – which were not self-tests. Is your rate significantly higher because you’re thinking of self-administered antigen tests?
  In many European countries, you can get antigen self-tests for about $2-4 a piece, this might make a testing scheme more cost-effective.

• Tobias H 16 Dec 2021 14:09 UTC

  1 point

  on: COVID and the holidays

  Thanks for this post, it helped clarify some of my concerns about the upcoming holidays.
  
  I’m surprised you don’t mention testing (PCR or antigen).
  What are your views on testing before an event?
  What would be a good protocol for testing before a specific event – test on the day of, a day before?