Hacker News: HaukeHi

New comment by HaukeHi in "Experts explore new mushroom which causes fairytale-like hallucinations"

HaukeHi — Sat, 27 Dec 2025 08:05:20 +0000

Plants want to be eaten only by big animals that take them on long and random walks and then die far away from where they are picked up to fertilize the seed.

Try O3-Pro for $1 (is AI a bubble?)

HaukeHi — Tue, 24 Jun 2025 12:00:33 +0000

Article URL: https://hauke.substack.com/p/try-o3-pro-in-chatgpt-for-1

Comments URL: https://news.ycombinator.com/item?id=44365215

Points: 1

# Comments: 0

New comment by HaukeHi in "Youth and what happens when it's gone"

HaukeHi — Tue, 04 Mar 2025 09:13:25 +0000

Fluid intelligence is higher when younger.

Crystallised intelligence increases with age.

Your comparative advantage changes.

An Interactive Shapley Value Explainer

HaukeHi — Sat, 12 Oct 2024 09:53:42 +0000

Article URL: https://nonzerosum.games/shapleyvalue.html

Comments URL: https://news.ycombinator.com/item?id=41817904

Points: 2

# Comments: 0

New comment by HaukeHi in "Ilya Sutskever's SSI Inc raises $1B"

HaukeHi — Wed, 04 Sep 2024 16:03:27 +0000

Demis Hassabis says that half of all innovations that caused the recent AI boom came from DeepMind, which is London based.

LLM Applications I Want to See

HaukeHi — Fri, 23 Aug 2024 16:37:08 +0000

Article URL: https://sarahconstantin.substack.com/p/llm-applications-i-want-to-see

Comments URL: https://news.ycombinator.com/item?id=41330464

Points: 2

# Comments: 0

Should we break up Google DeepMind?

HaukeHi — Mon, 22 Apr 2024 15:36:18 +0000

Article URL: https://forum.effectivealtruism.org/posts/oJhzRfgDHsK87aZYr/should-we-break-up-google-deepmind

Comments URL: https://news.ycombinator.com/item?id=40115384

Points: 2

# Comments: 1

New comment by HaukeHi in "Price fixing by algorithm is still price fixing"

HaukeHi — Sun, 03 Mar 2024 00:09:49 +0000

Isn't Airbnb doing the same?

New comment by HaukeHi in "The lesson about the end of nuclear in Germany"

HaukeHi — Wed, 26 Apr 2023 10:47:13 +0000

Continued reliance on outdated nuclear technology might not have the same crucial global technology spillovers as investments in other clean energy (including advanced nuclear). Since the best path towards global decarbonization is through global technology spillover into emerging economies, the actors that have the best emissions score may, surprisingly, not be the most effective actors at reducing the global rate of emissions in the future. This has some counterintuitive implications. Consider that Germany has higher carbon emissions than France even though it has invested more heavily in solar than its neighbor, which uses much more nuclear. Should advanced economies like Germany leave their nuclear plants running? Perhaps, but it will not make a very large dent in global emissions because 75% of all future emissions will come from emerging economies, which will not adopt the kind of (non-advanced) nuclear power currently in use in Germany. Consider that German citizens environmental footprints are currently less than 4% of the global total, a share that is on the decline.

At one point, German subsidies drove ~⅓ of the global solar adoption, ~86% of which occurred outside Germany (6x) - see https://founderspledge.com/stories/changing-landscape#fnref1

Mins on Public Transport ≈ 10√X KM

HaukeHi — Fri, 14 Apr 2023 14:46:49 +0000

Article URL: https://vitalik.ca/general/2023/04/14/traveltime.html

Comments URL: https://news.ycombinator.com/item?id=35570143

Points: 2

# Comments: 0

New comment by HaukeHi in "Fermi Calculation Examples"

HaukeHi — Sat, 01 Oct 2022 15:38:58 +0000

There's an amazing calibration training game where you can practice Fermis and see where you are orders of magnitude off:

https://www.quantifiedintuitions.org/calibration

Trends in GPU Price-Performance

HaukeHi — Fri, 01 Jul 2022 15:59:01 +0000

Article URL: https://www.lesswrong.com/posts/c6KFvQcZggQKZzxr9/trends-in-gpu-price-performance

Comments URL: https://news.ycombinator.com/item?id=31948334

Points: 3

# Comments: 0

New comment by HaukeHi in "It looks like you’re trying to take over the world"

HaukeHi — Sun, 27 Mar 2022 15:29:49 +0000

There's a good video by Robert Miles about that 'Why Not Just: Think of AGI Like a Corporation?' (youtube.com/watch?v=L5pUA3LsEaw)

Corporations are kind of like AIs, if you squint. How hard do you have to squint though, and is it worth it? In this video we ask: Are corporations artificial general superintelligences?

New comment by HaukeHi in "It looks like you’re trying to take over the world"

HaukeHi — Sun, 27 Mar 2022 14:06:44 +0000

I think evolution is not an undirected process in that sense because it's an optimization process, that optimizes to create more copies of itself. Superintelligence will likely use some Evolutionary Computation (see en.wikipedia.org/wiki/Evolutionary_computation ).

Also see Karl Sims 'Creatures' from the 90s: youtube.com/watch?v=JBgG_VSP7f8 or OpenAI's Multi-Agent Hide and Seek: youtube.com/watch?v=kopoLzvh5jY

New comment by HaukeHi in "It looks like you’re trying to take over the world"

HaukeHi — Sun, 27 Mar 2022 13:23:58 +0000

> Was any force ever able to get close to world domination?

Evolution? 2.5bn years ago stromatolites changed the atmosphere from a CO2-rich to O2-rich through photosynthesis, because they had no competition. Now plants dominate the earth (≈450 Gt C, the dominant kingdom), then animals (≈2 Gt C, mainly marine, and bacteria (≈70 Gt C) and archaea (≈7 Gt C).

In 2020, global human-made mass exceeded all living biomass ( nature.com/articles/s41586-020-3010-5).

Russia's cloud providers: Amazon, Microsoft, Google dominate

HaukeHi — Mon, 07 Mar 2022 14:28:33 +0000

Article URL: https://incountry.com/blog/global-clouds-and-cloud-providers-in-russia/

Comments URL: https://news.ycombinator.com/item?id=30588186

Points: 2

# Comments: 0

Examples of Specification Gaming in ML

HaukeHi — Wed, 29 Apr 2020 21:35:39 +0000

Article URL: https://www.youtube.com/watch?v=nKJlF-olKmg

Comments URL: https://news.ycombinator.com/item?id=23024666

Points: 1

# Comments: 0

New comment by HaukeHi in "Loss of sense of smell a marker for Covid-19 infection"

HaukeHi — Sat, 21 Mar 2020 21:22:30 +0000

From ""Rapid Clinical Evaluation of Anosmia - The Alcohol Sniff Test" (https://jamanetwork.com/journals/jamaotolaryngology/article-...):"Standard 70% isopropyl alcohol preparation pad is opened such that 0.5 cm of the pad itself is visible. The alcohol pad is placed beneath the patient's nostrils while the patient inspires twice, to familiarize himself or herself with the alcohol odor, and the subject is asked if he or she detects an odor. Odor thresholds for alcohols are 2 or more orders of magnitude lower than trigeminal thresholds for the same stimuli.6 Thus, an anosmic will detect the presence of alcohol trigeminally only when it is extremely close to the nose. The alcohol pad is withdrawn and the threshold test begun. The subject is asked to close the mouth and eyes, breathe normally, and indicate when the odor is detected. Active sniffing and deep inspiration are discouraged. The basic procedure follows the method of limits. A standard metric tape measure is extended downward from the patient's nares and held in place (Figure I ). The alcohol pad is placed 30 cm below the nose and, with each expiration, is moved 1 cm closer to the nares until the subject detects the presence of odor. The distance from the anterior nares to the alcohol padis measured in centimeters at the point at which the subject first detects the odor. The procedure is repeated 4 times and the mean distance defines the threshold.Butanol ThresholdFor purposes of comparison, all of the subjects completed a standard olfactory threshold test. A series of 10 concentrations of butanol ( -butyl alcohol) was used to determine absolute olfactory threshold sensitivity. The highest butanol concentration consisted of 4% vol/vol in distilled water. Each successive dilution was one third of the preceding dilution. Two "blanks," containing only distilled water, were also prepared. All bottles, including blanks, contained 60 mL of liquid. Olfactory threshold was assessed with a modified version7 of a 2-alternative, forced-choice,ascending method of limits procedure.8 The subject was presented with 2 bottles, one containing the odorant and the other consisting of distilled water. Each nostril was tested separately. The spout of the bottle was inserted into the nos tril of interest. The subject was asked to squeeze the bottleto generate a puff of air. The subject did this with both bottles. Subjects were asked to identify which of the 2 bottles contained the stronger odor.All subjects began at the lowest concentration to avoidadaptation.9 Incorrect choices led to presentation of a higher concentration and correct choices led to continued presentation of the same concentration to a criterion of 5 successive correct responses. The presentation of the odorantand blank were randomized for each comparison trial and the nostril to be tested first was also randomly determined. There were approximately 45 seconds between trials to allow time for recovery of the olfactory system and for the odor molecules to collect in the head space of the bottle."

However, this test has been called imprecise (https://journals.sagepub.com/doi/abs/10.2500/ajra.2017.31.44...).

If the paper is true, one could do spot checks today to estimate the true IFR in every country tomorrow on a shoestring budget.

https://www.statnews.com/2020/03/17/a-fiasco-in-the-making-a...

Depending on the results that might save trillions of dollars.

Ask HN: Detect Coronavirus Fever via Webcam?

HaukeHi — Wed, 18 Mar 2020 18:29:21 +0000

I have an idea for creating an open-source zero marginal cost, digital thermometer to detect coronavirus. Can you help?

Here's the idea:

1. Heart rate can be estimated via (webcam) video of someone’s face with high accuracy (even with poor video quality).[1],[2]

2. This heart rate can then be used to estimate core temperature with high accuracy.[3],[4]

3. Fever (body temperature ≥38°C) [5] is the most typical symptom of C19 - in 88% of confirmed cases.[6] (Though some C19 transmission might be asymptomatic[7] and presymptomatic.[8],[9])

4. Can we start an open-source project to create a digital fever thermometer? This could be used as a smartphone app or on the web (ala donottouchyourface.com). A webcam could continuously monitor people’s temperature and alert them to it if they have a fever.

‘Thermometer Guns’ have drawbacks: they’re more expensive, you need to get close to someone’s head to take temperature, they are not very accurate, they don’t provide continuous measurement- yet it is still used for coronavirus containment.[10]

I feel if this would work, it might be a very cost-effective intervention to diagnose coronavirus.

[1] "Detecting Pulse from Head Motions in Video - People.csail.mit ...." http://people.csail.mit.edu/balakg/pulsefromheadmotion.html

[2] "Heart rate estimation using facial video: A review " https://www.sciencedirect.com/science/article/abs/pii/S1746809417301362

[3] "Estimating Resting Core Temperature Using Heart Rate" https://journals.humankinetics.com/view/journals/jmpb/1/2/article-p79.xml

[4] "Real-time core body temperature estimation from heart ... - NCBI." 13 May. 2015, https://www.ncbi.nlm.nih.gov/pubmed/25967760

[5] "Early Transmission Dynamics in Wuhan, China, of ... - NEJM." https://www.nejm.org/doi/full/10.1056/NEJMoa2001316

All other citations here:

https://docs.google.com/document/d/1D4nhDux-ncnsZjXJrmXkZiK7LJjH_tyxWkVJR3fOOJ0/edit?usp=sharing

Comments URL: https://news.ycombinator.com/item?id=22619759

Points: 2

# Comments: 3

Remote Monitoring of Vital Signs in Using Computer Vision: A Review (2019)

HaukeHi — Sun, 15 Mar 2020 14:09:12 +0000

Article URL: https://www.mdpi.com/2076-3417/9/20/4474

Comments URL: https://news.ycombinator.com/item?id=22583550

Points: 2

# Comments: 0