morganlee

About:

computer scientist by training. done a lot of research on biometrics and machine learning for psychology. work on some related defense and science policy topics now.

Show more
View All Badges »
2025 Season Stats

-0.046227

Relative Brier Score

3

Forecasts

1

Upvotes
Forecasting Activity
Forecasting Calendar
No forecasts in the past 3 months
 
Past Week Past Month Past Year This Season All Time
Forecasts 0 0 3 3 3
Comments 0 0 3 3 3
Questions Forecasted 0 0 2 2 2
Upvotes on Comments By This User 0 0 1 1 1
 Definitions
New Badge
morganlee earned a new badge:

Active Forecaster

New Prediction
Why do you think you're right?
  • All 32 allies have already agreed to the 2% floor even though several remain below it. Shows unanimity is possible.
  • Jumping from 2% to 5% more than doubles the burden. Many allies are still well under 2% (like Italy, Canada, Belgium)
  • The wording does make assent easier aka draft pledge does not set a deadline, only a commitment to reach 5%, meaning it lowers the near-term cost of saying yes.
  • Pressure from Russia raises political will. Russia outproduces NATO in munitions. Could threaten Europe.
  • Trump + US envoy have seemingly made 5% the price of continued US backing which pressures NATO members.
  • Lost of investment by NATO members. Which is tough to meet if they are trying to grow other aspects of their nations. Each nation who can't make this is a potential veto.
  • The temporal constraint (July 31, 2025), makes this hard to meet. Prior 2% pledge took years of past work.
  • Base rate: 15% chance. +5pp for intensifying security pressure and US pressure. -2pp for persistent fiscal and political constraints and veto risk.
  • Estimating roughly 18% chance that every NATO member agrees to the 5% commitment (at least stated, not done)
Files
Why might you be wrong?
  • Threat perceptions change (Russia capability warnings get starker)
  • NATO can adopt consensus language that commits to 5% yet lets each state choose its timetable. But one member can block if it is politically toxic at home.
  • There may not be enough time to get a consensus. This took years to get 2% commitment in the past.
Files
New Badge
morganlee earned a new badge:

Upvotes Received

New Prediction
What will the Artificial Analysis Intelligence Index score for OpenAI’s highest scoring model be on 31 December 2025?
morganlee made their 1st forecast (view all):
Compare to me
Probability
Answer
10%
Less than or equal to 70
21%
Between 71 and 74, inclusive
27%
Between 75 and 78, inclusive
25%
Between 79 and 82, inclusive
17%
More than or equal to 83
Why do you think you're right?
  • Probably going to be 77 pts.
  • Current leader is o4-mini high at 70
  • Next best openAI model is o3 at 67. Google gemini 2.5 pro scores 69.  Competitive pressure keeps OpenAI incentivized to stay 1-3 pts ahead.
  • Previous OpenAI flagships gpt-4 got 36, gpt-4o got 41, o3 / o4-mini are at 67-70.
  • + 30 pt in about 20 months. Rate is slowing as benchmarks saturate.
  • Their index methodology weights 8 hard evals and claims 1 pt measurement error.  
  • 1.3 pt/month over (34 pts) 26 month mar 2023 to may 2025
  • o3->o4-mini high only +3, shows diminishing marginal returns. indicates benchmark saturation.
  • 1.3/pt/mo * 7 months (may to dec) = 9 pts. assume productivity taper to 0.8 pts/mo. = 6 pts. Median of two approaches is +7 pts 70+7 = 77 pts
Files
Why might you be wrong?
  • If the metric is rebased, widen the dataset mix, tighten scoring, raw scores might dip 3-6 pts.
  • If scaling laws flatten, and no GPT-5 equivalent is production ready, openAI might end 2025 with incremental finetuning (70-73).
  • Multimodal "o5" or agent-augmented o5-reasoning might be significant gains. If it happens, or science is there.
Files
New Badge
morganlee earned a new badge:

My First Question

Congratulations on making your first forecast!
New Badge
morganlee earned a new badge:

Active Forecaster

New Prediction
How many measles cases in the U.S. will be reported by the CDC between 1 January 2025 and 30 September 2025?
morganlee made their 1st forecast (view all):
Compare to me
Probability
Answer
18%
Less than or equal to 1349
21%
Between 1350 and 1499, inclusive
23%
Between 1500 and 1649, inclusive
18%
Between 1650 and 1799, inclusive
20%
More than or equal to 1800
Why do you think you're right?
  • I think it's most likely to be about 1550 cases (median of scenarios below).
  • Recent trajectory of cases slowing down. https://www.theguardian.com/us-news/2025/may/24/texas-measles-outbreak
  • Low spread scenario 1 case/day
  • Moderate 3 case/day
  • Status-quo about 5 case/day
  • High (no slowdown based on current data from jan 1 to may 23) 7.4 case/day = 970 additional cases between may 23 and sep 30. total about ~2000.
  • High baseline immunity
  • Past analogs. 2019 had 1249 cases by 30 Sep and then only 25 cases through year end.
Files
Why might you be wrong?
  • If it is higher, then some communities may have lower vaccination rates, there would likely have been a summer travel spike, the healthcare system in Texas might've been stretched by the spring outbreak.
  • If lower than expected, there would have to be some rapid burnout of a current outbreak. Maybe emergency vaccination uptake, more so than usual. World events that reduces travel, may reduce infections. Some cases may be misreported as measles, or not.
Files
Files
Tip: Mention someone by typing @username
Cancel