@BenjaminHan@sigmoid.social
@BenjaminHan@sigmoid.social avatar

BenjaminHan

@BenjaminHan@sigmoid.social

Working on natural language, knowledge, reasoning, machine learning, and AI at a fruity company.

Husband, father, runner, German learner, piano player. A curious soul living in #PacificNorthwest (WA US).

Running 05/25/18-05/19/24 (dist # time pace/m date src):

5K 645 21:34 6'56" 4/20/24 Strava
10K 97 48:59 7'52" 5/16/24 Strava
15K 4 1:16:05 8’10” 5/19/24 Strava
HM 25 1:48:25 8’16” 5/19/24 Strava
M 7 3:44:58 8'35" 3/24/24 AppleW

#nlp #nlproc #knowledgeGraphs #ai #running #classicalMusic

This profile is from a federated server and may be incomplete. Browse more on the original instance.

BenjaminHan, to generativeAI
@BenjaminHan@sigmoid.social avatar

1/ In this age of LLMs and generative AI, do we still need knowledge graphs (KGs) as a way to collect and organize domain and world knowledge, or should we just switch to language models and rely on their abilities to absorb knowledge from massive training datasets?

BenjaminHan, to ai
@BenjaminHan@sigmoid.social avatar

Dog Lenat, founder of , passed away earlier this week. From Professor Ken Forbus:

"People in AI often don't give the Cyc project the respect it deserves. Whether or not you agree with an approach, understanding what has happened in different lines of work is important. The Cyc project was the first demonstration that symbolic representations and reasoning could scale to capture significant portions of commonsense…”

https://www.linkedin.com/posts/forbus_ai-knowledgegraphs-krr-activity-7103445990954700800-qcd-

BenjaminHan, to LLMs
@BenjaminHan@sigmoid.social avatar

1/ How robust and reliable is the code generated by , especially for real-world software development? A recent work [2] constructed a new benchmark based on [1] to evaluate if the generated code uses API correctly. Four popular -- .5, , , and -- are tested, and under zero-shot scored 62.09% misuse rate. Even with one-shot relevant examples the misuse rate of is 49.17%.

image/png
image/png
image/png

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

2/ Since users of with particular APIs are usually relatively inexperienced in the said APIs, these inaccuracies may have grave consequences to the robustness and reliability of the resulting software.

(How would fare?)

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

3/ REFERENCES

[1] Tianyi Zhang, Ganesha Upadhyaya, Anastasia Reinhardt, Hridesh Rajan, and Miryung Kim. 2018. Are code examples on an online Q&A forum reliable? a study of API misuse on stack overflow. In Proceedings of the 40th International Conference on Software Engineering, pages 886–896, Gothenburg, Sweden. Association for Computing Machinery. http://dx.doi.org/10.1145/3180155.3180260

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

4/end

[2] Li Zhong and Zilong Wang. 2023. A Study on Robustness and Reliability of Large Language Model Code Generation. http://arxiv.org/abs/2308.10335

caseynewton, to random
@caseynewton@mastodon.social avatar

Three years of using expensive note-taking software doesn’t seem to be making me a better thinker. I wrote about why that is — and whether a new generation of AI tools can help https://www.platformer.news/p/why-note-taking-apps-dont-make-us

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

@caseynewton Great piece, really resonated with me. And thanks for mentioning Andy’s notes!

To me, a great note-taking system has to feel effortless, e.g., no fuss on formatting, making links etc, and allowing multi-modal input too. And it needs to be great on retrieval and insight discovery.

BenjaminHan, to random
@BenjaminHan@sigmoid.social avatar

Here is the hoping sigmoid.social can adopt it too… @thegradient https://oisaur.com/@renchap/110946188204729065

jbigham, to random
@jbigham@hci.social avatar

a challenge with being an HCI researcher is that we know too much about interaction and humans to make unrealistic but exciting claims about how ML will change interaction 🤔

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

@jbigham Interesting - can you point to a few papers demonstrating this?

BenjaminHan, to mastodon
@BenjaminHan@sigmoid.social avatar

For all of the 91%: please post more about your fields, and be an evangelist to convince more of our colleagues to join the fun! Let’s make an even richer and livelier venue than !

https://sigmoid.social/@lysander07/110926392740342316

BenjaminHan, to random
@BenjaminHan@sigmoid.social avatar

Hi @MonaApp — the user icons showing up in notifications on Apple Watch are often incorrect.

BenjaminHan, to Marathon
@BenjaminHan@sigmoid.social avatar

1/ This weekend's 16.3-mile report (total time/pace):

  • 5K : 24:19/7'50", slower by 24 seconds than last.
  • Half : 1:54:50/8'46", faster by 5 minutes than last.

All-time-best are still:

  • 5K: 23:51/7'41"
  • Half Marathon: 1:51:48/8'32"

image/jpeg
image/jpeg

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

2/

  1. Thanks @jbigham for inspiring me to do a negative split! My pace for the two halves of the half Marathon were 8’54” vs 8’38” — I guess one way to do it is to slow down in the first half! :-)

  2. Found a way to manage my dry mouth caused by mouth breathing: frequent swallowing keeps my throat a bit more moist.

  3. Keeping my motion on a straight line forward and positioning my elbows higher with reduced swinging helps reducing wobbling.

BenjaminHan, to books
@BenjaminHan@sigmoid.social avatar

“What chemicals cause that nostalgic old book smell?

Compounds like benzaldehyde, vanillin, ethylbenzene, and 2-ethyl hexanol are often responsible for old book scents. Benzaldehyde has an almond-like scent, vanillin smells like vanilla, ethylbenzene is sweet and plastic-y, and 2-ethyl hexanol is lightly floral.”

Why Do Old So Good? – ScienceSwitch https://scienceswitch.com/2023/08/19/why-do-old-books-smell-so-good/

BenjaminHan, to LLMs
@BenjaminHan@sigmoid.social avatar

Large degradations observed from when tasks are reframed into counterfactuals. Only basic syntax, logic and music chords are decent.

(see also: https://lnkd.in/gje_WkR3 )

Zhaofeng Wu, Linlu Qiu, Alexis Ross, Ekin Akyürek, Boyuan Chen, Bailin Wang, Najoung Kim, Jacob Andreas, and Yoon Kim. 2023. Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Tasks. http://arxiv.org/abs/2307.02477

image/jpeg

jbigham, to random
@jbigham@hci.social avatar

for what it's worth, i noticed a big difference in marathoning when i did multiple 20-22 mile runs at a pretty fast clip in the weeks leading up to it … https://www.nytimes.com/article/marathon-training-pace-miles.html

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

@jbigham Very timely — my half Marathon race is coming in 2 weeks. I might try negative split this weekend.

BenjaminHan, to mastodon
@BenjaminHan@sigmoid.social avatar

I hope more and will join the exodus to . Sigmoid.social in particular is an instance for folks interested in , , , or any other relevant disciplines to congregate and mingle.

https://aus.social/@thesiswhisperer/110878776001647572

feditips, to random
@feditips@mstdn.social avatar

If you're looking for a reliable and responsibly moderated Fediverse server to sign up on or move to, I run a website that lists them:

➡️ https://fedi.garden

All servers listed comply with the Covenant (https://fedi.garden/about-this-site/) and have opted in to being listed.

Follow the site's account to keep up to date on the latest servers added to the site:

➡️ @FediGarden

If you want help with moving a Mastodon account to another server, here's a step-by-step guide:

➡️ https://fedi.tips/transferring-your-mastodon-account-to-another-server/

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

@feditips @FediGarden Include https://sigmoid.social/about ? Topics are , etc.

BenjaminHan, to mastodon
@BenjaminHan@sigmoid.social avatar

This explains why sometimes someone favorited/reposted my post, when I clicked on their profiles, I didn’t get to see their posts. The end result is, I don’t feel informed enough to follow them, which I think discourages Mastodon users from expanding their circles.

Notes on using a single-person server https://jvns.ca/blog/2023/08/11/some-notes-on-mastodon/

image/jpeg

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

But this can be easily fixed on the client side, no? Just fetch enough posts dynamically when a user clicks on a profile if it’s not already available from the user’s instance? @MonaApp ?

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

@MonaApp Thanks — didn’t realize this feature! Is there any reason not to make this the default behavior when showing a “remote” profile?

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

@MonaApp Got it. Do you see value in making it a setting then? This way users can choose to always load remote profiles from remote servers if they’re willing to suffer from the shortcomings you mentioned?

BenjaminHan, to Marathon
@BenjaminHan@sigmoid.social avatar

1/ Alright -- here is my not-so-good, pretty-bad weekend's 16.3-mile report:

  • On Saturday I finished my 35th 5K 25:20, almost 2 minutes slower than my PB (23:21) just a week ago! I even stopped to walk 3 times during the run, which never happened before!

  • On Sunday I finished my 11th half 2:10:58, almost 20 minutes slower than my fastest (1:51:48) just a month ago!

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

2/ Thinking of the possible reasons, one could be the high humidity in the past two days. On Saturday we started out with 80+% humidity, and on Sunday it was 91%. The temperature was mild (~19 Celsius), but tiredness more quickly set in for me with no particular muscle soreness etc. With the 5K runs, my heart rate was routinely in Zone 5 (160+BPM), meaning a bit thicker air might have done me in.

BenjaminHan,
@BenjaminHan@sigmoid.social avatar

1/ I should have learned this, but it's never too late: this page explains what humidity can do to runners https://marathonhandbook.com/running-in-humidity/

TLDR:

  1. Humidity is a relative measure depending on temperature. More reliable metric we should pay attention to is dew point (see picture). The day I was having trouble, the dew point was about 64.5.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • megavids
  • tester
  • DreamBathrooms
  • thenastyranch
  • magazineikmin
  • osvaldo12
  • ethstaker
  • Youngstown
  • mdbf
  • slotface
  • rosin
  • ngwrru68w68
  • kavyap
  • GTA5RPClips
  • JUstTest
  • cisconetworking
  • InstantRegret
  • khanakhh
  • cubers
  • everett
  • Durango
  • tacticalgear
  • Leos
  • modclub
  • normalnudes
  • provamag3
  • anitta
  • lostlight
  • All magazines
    •