We’re ok with people posting multiple snapshots, if you want to update it based on later comments! You can edit your comment with a new snapshot link, or add a new comment with the latest snapshot (we’ll consider the latest one, or whichever one you identify as your final submission)
Amandango
Karma: 774
If people don’t have a strong sense of who these people are/would be, you can look through this google scholar citation list (this is just the top AI researchers, not AGI researchers).
In a similar vein to this, I found several resources that make me think it should be higher than 1% currently and in the next 1.5 years:
This 2012⁄3 paper by Vincent Müller and Nick Bostrom surveyed AI experts, in particular, 72 people who attended AGI workshops (most of whom do technical work). Of these 72, 36% thought that assuming HLMI would at some point exist, it would be either ‘on balance bad’ or ‘extremely bad’ for humanity. Obviously this isn’t an indication that they understand or agree with safety concerns, but directionally suggests people are concerned and thinking about this.
This 2017 paper by Seth Baum identified 45 projects on AGI and their stance on safety (page 25). Of these, 12 were active on safety (dedicated efforts to address AGI safety issues), 3 were moderate (acknowledge safety issues, but don’t have dedicated efforts to address them), and 2 were dismissive (argue that AGI safety concerns are incorrect). The remaining 28 did not specify their stance.
Delegate a Forecast
Either expected number of people who get covid or number of microcovids generated by the event works as a question! My instinctive sense is that # of people who get covid will be easier to quickly reason about, but I’ll see as I’m forecasting it.
Here’s my prediction, and here’s a spreadsheet with more details (I predicted expected # of people who would get COVID). Some caveats/assumptions:
There’s a lot of uncertainty in each of the variables that I didn’t have time to research in-depth
I didn’t adjust for this being outdoors, you can add a row and adjust for that if you have a good sense of how it would affect it.
I wasn’t sure how to account for the time being 3 hours. My sense is that if you’re singing loudly at people < 1m for 3 hours, this is going to be a pretty high infection rate. Also, I assumed they weren’t wearing masks because of the singing. I’m most uncertain about this though
You didn’t mention how big the pods are. I assumed 10 people in a pod, but it would change it if this were much smaller.
Here’s my prediction for this! I predicted a median of March 1, 2029. Below are some of the data sources that informed my thinking.
Base rates:
iPhone: ~30% of US owns an iphone in 2020. This is ~10 years from when it was launched in 2007.
Smartphones: Launched in 1992, 30% penetration in 2011 (19 years)
Tablet: iPad launched in 2010, tablets reached 30% penetration in 2013 (3 years, not a perfect reference point because iPads are just a subset of tablets)
How many people currently use AR/AR glasses?
I assume most of this is iphone apps, including Pokemon Go that had 60 million monthly active users in 2017
When will major AR glasses products be released?
Related Metaculus question: When will sales of a non-screen technology be greater than sales of a screen technology?
Oh yeah that makes sense, I was slightly confused about the pod setup. The approach would’ve been different in that case (still would’ve estimated how many people in each pod were currently infected, but would’ve spent more time on the transmission rate for 30 feet outdoors). Curious what your current prediction for this is? (here is a blank distribution for the question if you want to use that)
I can help with this if you share the post with me!
[Question] Forecasting Thread: AI Timelines
The blue distribution labeled “Your distribution” in this snapshot is Alex’s updated 2020 prediction.
Reflections on AI Timelines Forecasting Thread
Here’s a colab you can use to do this! I used it to make these aggregations:
The Ethan + Datscilly distribution is a calculation of:
- 25% * Your inside view of prosaic AGI
- 60% * Datscilly’s prediction (renormalized so that all the probability < 2100)
- 15% * We get AGI > 2100 or never
This has an earlier median (2040) than your original distribution (2046).(Note for the colab: You can use this to run your own aggregations by plugging in Elicit snapshots of the distributions you want to aggregate. We’re actively working on the Elicit API, so if the notebook breaks lmk so we can update it).
I also just discovered BERI’s x-risk prediction market question set and Jacobjacob & bgold’s AI forecasting database, which seem really helpful for this!
This is a really great conditional question! I’m curious what probability everyone puts on the assumption (GPT-N gets us to TAI) being true (i.e. do these predictions have a lot of weight in your overall TAI timelines)?
I plotted human_generated_text and sairjy’s answers:
Thank you for putting this spreadsheet database together! This seemed like a non-trivial amount of work, and it’s pretty useful to have it all in one place. Seeing this spreadsheet made me want:
More consistent questions such that all these people can make comparable predictions
Ability to search and aggregate across these so we can see what the general consensus is on various questions
I thought the 2008 GCR questions were really interesting, and plotted the median estimates here. I was surprised by / interested in:
How many more deaths were expected from wars than other disaster scenarios
For superintelligent AI, most of the probability mass was < 1M deaths, but there was a high probability (5%) on extinction
A natural pandemic was seen as more likely to cause > 1M deaths than an engineered pandemic (although less likely to cause > 1B deaths)
FYI, this is on a log scale. I plotted extinction as > 8B deaths.
(posted a similar comment on the EA forum link, since it seems like people are engaging more with this post there)
I noticed that your prediction and jmh’s prediction are almost the exact opposite:
Teerth: 80%: No human being would be living on another celestial object (Moon, another planet or asteroid) (by 2030)
jmh: 90%: Humans living on the moon (by 2030)
(I plotted this here to show the difference, although this makes the assumption that you think the probability is ~uniformly distributed from 2030 – 2100). Curious why you think these differ so much? Especially jmh, since 90% by 2030 is more surprising—the Metaculus prediction for when the next human being will walk on the moon has a median of 2031.
A rough distribution (on a log scale) based on the two points you estimated for wars (95% < 1B people die in wars, 85% < 10M people die in wars) gives a median of ~2,600 people dying. Does that seem right?
Yeah this seems pretty reasonable. It’s actually stark looking at the Our World in Data – that seems really high per year. Do you have your model somewhere? I’d be interested to see it.
Would the latter distribution you described look something like this?