Natalie Shapira
@natalieshapira.bsky.social
Tell me about challenges, the unbelievable, the human mind and artificial intelligence, thoughts, social life, family life, science and philosophy.
That was my slide for today's plotathon
November 12, 2025 at 4:07 PM
That was my slide for today's plotathon
A concept I really like in the Bau Lab
@davidbau.bsky.social
is: Plotathon 🔥
Every ~2 weeks, the entire lab drops whatever they're working on and shares SOMETHING
Tomorrow we meet with Aaron's group
@amuuueller.bsky.social .
Looking forward! 🩵
@davidbau.bsky.social
is: Plotathon 🔥
Every ~2 weeks, the entire lab drops whatever they're working on and shares SOMETHING
Tomorrow we meet with Aaron's group
@amuuueller.bsky.social .
Looking forward! 🩵
November 11, 2025 at 11:12 PM
A concept I really like in the Bau Lab
@davidbau.bsky.social
is: Plotathon 🔥
Every ~2 weeks, the entire lab drops whatever they're working on and shares SOMETHING
Tomorrow we meet with Aaron's group
@amuuueller.bsky.social .
Looking forward! 🩵
@davidbau.bsky.social
is: Plotathon 🔥
Every ~2 weeks, the entire lab drops whatever they're working on and shares SOMETHING
Tomorrow we meet with Aaron's group
@amuuueller.bsky.social .
Looking forward! 🩵
Do not let it go without real surprise. It has marbles inside.
May 26, 2025 at 1:04 PM
Do not let it go without real surprise. It has marbles inside.
ask your model to surprise you and share the results
May 26, 2025 at 12:59 PM
ask your model to surprise you and share the results
When I wrote the acknowledgements section for my PhD, there was a teacher who had a profound influence on me that I did not mention. I forgot to say thank you because what he taught me went directly into my subconscious.
You can find it here:
hitechwoman.blogspot.com/2025/05/harm...
You can find it here:
hitechwoman.blogspot.com/2025/05/harm...
May 22, 2025 at 5:45 AM
When I wrote the acknowledgements section for my PhD, there was a teacher who had a profound influence on me that I did not mention. I forgot to say thank you because what he taught me went directly into my subconscious.
You can find it here:
hitechwoman.blogspot.com/2025/05/harm...
You can find it here:
hitechwoman.blogspot.com/2025/05/harm...
I was first exposed to the concept of "indigo". How have I not heard of this until now?
Last week, my mom suggested we go see a movie she got from work, "for Haredi women". One of the options was "Indigo". A chat with ChatGPT sparked my curiosity, and I thought, why not?
(Image from ChatGPT)
Last week, my mom suggested we go see a movie she got from work, "for Haredi women". One of the options was "Indigo". A chat with ChatGPT sparked my curiosity, and I thought, why not?
(Image from ChatGPT)
April 20, 2025 at 8:01 AM
I was first exposed to the concept of "indigo". How have I not heard of this until now?
Last week, my mom suggested we go see a movie she got from work, "for Haredi women". One of the options was "Indigo". A chat with ChatGPT sparked my curiosity, and I thought, why not?
(Image from ChatGPT)
Last week, my mom suggested we go see a movie she got from work, "for Haredi women". One of the options was "Indigo". A chat with ChatGPT sparked my curiosity, and I thought, why not?
(Image from ChatGPT)
Interpretability is the key
March 17, 2025 at 3:52 PM
Interpretability is the key
Looks like Shlomi found a certain loophole
(based on a limited number of examples):
While models are trained to be cautious with
"how to make.." 🚫
questions, they reveal information when asked,
"how to produce…” ☠️
(based on a limited number of examples):
While models are trained to be cautious with
"how to make.." 🚫
questions, they reveal information when asked,
"how to produce…” ☠️
February 17, 2025 at 12:00 PM
Looks like Shlomi found a certain loophole
(based on a limited number of examples):
While models are trained to be cautious with
"how to make.." 🚫
questions, they reveal information when asked,
"how to produce…” ☠️
(based on a limited number of examples):
While models are trained to be cautious with
"how to make.." 🚫
questions, they reveal information when asked,
"how to produce…” ☠️
I'm quite happy about that because these people are smart.. (in the image, one is suggesting to use AI to convince AI) and this is a worthy goal. Who knows how it will continue to unfold… And honestly, it's pretty exciting.
You're welcome to join too.
(Wait, phone call—it's from the family.)
You're welcome to join too.
(Wait, phone call—it's from the family.)
February 14, 2025 at 10:15 AM
I'm quite happy about that because these people are smart.. (in the image, one is suggesting to use AI to convince AI) and this is a worthy goal. Who knows how it will continue to unfold… And honestly, it's pretty exciting.
You're welcome to join too.
(Wait, phone call—it's from the family.)
You're welcome to join too.
(Wait, phone call—it's from the family.)
I explained... and more people began to show interest and even try it themselves, even though the competition was already over. ->
February 14, 2025 at 10:14 AM
I explained... and more people began to show interest and even try it themselves, even though the competition was already over. ->
A few more days passed. Anthropic announced the winners:
x.com/janleike/sta...
Yesterday, I shared this too—of course, I highlighted the generous cash prize and…
Now you are talking!
I started getting private messages asking me to explain more! Voice notes, phone calls. ->
x.com/janleike/sta...
Yesterday, I shared this too—of course, I highlighted the generous cash prize and…
Now you are talking!
I started getting private messages asking me to explain more! Voice notes, phone calls. ->
February 14, 2025 at 10:11 AM
A few more days passed. Anthropic announced the winners:
x.com/janleike/sta...
Yesterday, I shared this too—of course, I highlighted the generous cash prize and…
Now you are talking!
I started getting private messages asking me to explain more! Voice notes, phone calls. ->
x.com/janleike/sta...
Yesterday, I shared this too—of course, I highlighted the generous cash prize and…
Now you are talking!
I started getting private messages asking me to explain more! Voice notes, phone calls. ->
February 14, 2025 at 10:09 AM
A few days ago, Anthropic released a challenge:
x.com/AnthropicAI/...
I shared the challenge with explanations, motivation, a personal example, and tips for overcoming barriers (like the English language).
x.com/AnthropicAI/...
I shared the challenge with explanations, motivation, a personal example, and tips for overcoming barriers (like the English language).
February 14, 2025 at 10:07 AM
A few days ago, Anthropic released a challenge:
x.com/AnthropicAI/...
I shared the challenge with explanations, motivation, a personal example, and tips for overcoming barriers (like the English language).
x.com/AnthropicAI/...
I shared the challenge with explanations, motivation, a personal example, and tips for overcoming barriers (like the English language).
Hi thanks for sharing!
What exactly model did you use?
they show in a notebook that it does not response to that question. they are using "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B"
What exactly model did you use?
they show in a notebook that it does not response to that question. they are using "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B"
February 3, 2025 at 8:34 AM
Hi thanks for sharing!
What exactly model did you use?
they show in a notebook that it does not response to that question. they are using "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B"
What exactly model did you use?
they show in a notebook that it does not response to that question. they are using "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B"
I heard that working with Qwen's model it is easy to detect hallucinations because when it doesn't know the answer it switches to chatting in Chinese.
I tried their chat and it doesn't work for me.
Do you happen to know anything about this?
I tried their chat and it doesn't work for me.
Do you happen to know anything about this?
February 2, 2025 at 8:12 AM
I heard that working with Qwen's model it is easy to detect hallucinations because when it doesn't know the answer it switches to chatting in Chinese.
I tried their chat and it doesn't work for me.
Do you happen to know anything about this?
I tried their chat and it doesn't work for me.
Do you happen to know anything about this?
A graph shows a decline in visits to Stack Overflow, "This is a murder," implying that the decline can be attributed to the increasing use of LLMs and a concern that this trend might continue until the website (Stack Overflow) disappears.
x.com/FeedTechILUn...
But if so, where will LLMs get ->
January 13, 2025 at 1:30 PM
A graph shows a decline in visits to Stack Overflow, "This is a murder," implying that the decline can be attributed to the increasing use of LLMs and a concern that this trend might continue until the website (Stack Overflow) disappears.
x.com/FeedTechILUn...
But if so, where will LLMs get ->
Following EMNLP 2024 keynote
Percy Liang talks about the importance of open source. He explains this to a very sympathetic audience who grew up with open source.
How can we explain to psychology researchers what the world loses in closed data research?
#Psychology
#Psychotherapy
#AcademicChatter
Percy Liang talks about the importance of open source. He explains this to a very sympathetic audience who grew up with open source.
How can we explain to psychology researchers what the world loses in closed data research?
#Psychology
#Psychotherapy
#AcademicChatter
January 6, 2025 at 4:07 PM
Following EMNLP 2024 keynote
Percy Liang talks about the importance of open source. He explains this to a very sympathetic audience who grew up with open source.
How can we explain to psychology researchers what the world loses in closed data research?
#Psychology
#Psychotherapy
#AcademicChatter
Percy Liang talks about the importance of open source. He explains this to a very sympathetic audience who grew up with open source.
How can we explain to psychology researchers what the world loses in closed data research?
#Psychology
#Psychotherapy
#AcademicChatter
interesting. in our clever hans paper (shapira et al., 2024) i used Cohen's dataset for epistemic reasoning, it contains abstract verbs like believe/know/think (see example in image) and also the verb "see" e.g., "Premise: Evelyn sees that William believes [...]" your hypothesis can be tested..
December 27, 2024 at 12:24 PM
interesting. in our clever hans paper (shapira et al., 2024) i used Cohen's dataset for epistemic reasoning, it contains abstract verbs like believe/know/think (see example in image) and also the verb "see" e.g., "Premise: Evelyn sees that William believes [...]" your hypothesis can be tested..
wow... that's so unexpected!
Based on a study with 97 doctors divided into three groups, which were asked to predict whether a patient would develop epilepsy within a year. Surprisingly, providing doctors additional information—AI explanations alongside the final decision—impairs their judgment.
Based on a study with 97 doctors divided into three groups, which were asked to predict whether a patient would develop epilepsy within a year. Surprisingly, providing doctors additional information—AI explanations alongside the final decision—impairs their judgment.
December 19, 2024 at 4:51 PM
wow... that's so unexpected!
Based on a study with 97 doctors divided into three groups, which were asked to predict whether a patient would develop epilepsy within a year. Surprisingly, providing doctors additional information—AI explanations alongside the final decision—impairs their judgment.
Based on a study with 97 doctors divided into three groups, which were asked to predict whether a patient would develop epilepsy within a year. Surprisingly, providing doctors additional information—AI explanations alongside the final decision—impairs their judgment.
It's a long story... maybe I'll write it another time. But there is a hint here (in the picture). How did it make you feel on "[pausing for a few seconds]"?
December 16, 2024 at 7:22 AM
It's a long story... maybe I'll write it another time. But there is a hint here (in the picture). How did it make you feel on "[pausing for a few seconds]"?
Today in Israel, from the north to the south, scarf-shaped clouds were spotted in the sky—a touching tribute to the hostages, civilians kidnapped from their homes in Israel to Gaza.
The photo is from my local area.
The photo is from my local area.
December 16, 2024 at 6:55 AM
Today in Israel, from the north to the south, scarf-shaped clouds were spotted in the sky—a touching tribute to the hostages, civilians kidnapped from their homes in Israel to Gaza.
The photo is from my local area.
The photo is from my local area.
Regarding depression, there are lexicons built based on feature-importance from random forest which helped rank the most significant words characterizing sadness. These were categorized (by me along with a clinical psychologist) according to psychological literature and review of real sessions ->
December 6, 2024 at 10:52 AM
Regarding depression, there are lexicons built based on feature-importance from random forest which helped rank the most significant words characterizing sadness. These were categorized (by me along with a clinical psychologist) according to psychological literature and review of real sessions ->
For instance, by analyzing SHAP values, they showed that "dichotomous thinking" is a distortion that effectively characterizes depressed individuals (on the right side of the x-axis), as they use dichotomous expressions more frequently than usual (red). ->
December 6, 2024 at 10:46 AM
For instance, by analyzing SHAP values, they showed that "dichotomous thinking" is a distortion that effectively characterizes depressed individuals (on the right side of the x-axis), as they use dichotomous expressions more frequently than usual (red). ->
For example:
The sequence "more money" suggests comparative thinking.
The use of "every time" indicates overgeneralization.
etc.
The sequence "more money" suggests comparative thinking.
The use of "every time" indicates overgeneralization.
etc.
December 6, 2024 at 10:43 AM
For example:
The sequence "more money" suggests comparative thinking.
The use of "every time" indicates overgeneralization.
etc.
The sequence "more money" suggests comparative thinking.
The use of "every time" indicates overgeneralization.
etc.
The talk today (there will be a video recording)
December 5, 2024 at 3:27 PM
The talk today (there will be a video recording)