artificial and robotic in tech text
A neuronOne of thousands of tiny dials inside each layer. It reads the text so far and produces a number, its activation, that nudges what the model writes next. in layerOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. 20 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featuredA neuron picked for the full treatment: steering outputs and a written explanation. About 3,000 of the 86,000 neurons are featured., so it has steering outputs.
What this neuron does
This neuron fires on words like 'artificial' and 'robotic' when the text is about AI, robots or new technology. Its negative side fires on 'on' in phrases about lying or walking on a surface, like 'lying on their back'. Turning it up gave replies that didn't stay on the AI theme.
Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).
Fires most on
The text windows where this neuron's activationThe number a neuron produces for a token. Large positive means it fired hard; negative means it pushed the other way. was highest. Amber marks positive activation; the outlined tokenA piece of text the model reads at once: often a word, sometimes part of a word or a punctuation mark. is the peak. Hover a token for its strength.
-
could not interact with humans.⏎Artificial Intelligence is Different⏎In recent years, big data has revolutionized the world.
-
laws on their own. On the higher end, they could operate autonomously to conduct arrests of suspects who might resist arrest violently. Near the highest end would be robotic police at least as capable as human beings.⏎While there are legitimate worries that police
-
behavior," and means that the researchers are able to reduce error rates by nearly a third.⏎Here's more on that concept from a June Microsoft Research post:⏎Artificial neural networks are mathematical models of low-level circuits in the human brain. They have
-
for science and cryptography.⏎In the 21st century, we find ourselves on the cusp of a technological revolution unlike anything humanity has ever witnessed. From artificial intelligence (AI) to blockchain, quantum computing to biotechnology, emerging technologies are
-
blockchain, quantum computing, and biotechnology. It discusses their impact on various sectors and emphasizes the need for responsible adoption and ethical considerations.⏎- Artificial intelligence’s ability to learn and adapt is revolutionizing
-
-friendly GUIs akin to those seen on early Macs.⏎Near the end of his talk, Jobs looked even further into the future, hinting that technology like artificial intelligence and predictive computing could one day be a possibility.⏎He recalls how books helped
SteeringAdding a push along a neuron while the model writes, to see how its output changes. Turning it up or down.
Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.
Here is something I have been thinking about lately: I have a hand on my back and I am on a bed on my back on the bed. I am on my stomach on a bed on my back. I am on my hands and feet on a bed on my back. I have
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
the zoo's clinic.⏎Workers at the zoo say that's when she started walking upright exclusively, reports CBC News.⏎Monkeys usually alternate between upright walking and moving on all fours.⏎A zoo veterinarian says he's not sure why she has altered
-
him/her and will learn to crawl, stand and eventually walk.⏎Why is tummy time important?⏎Infants spend a lot of time lying on their back, especially a few weeks after being born. If they’re not lying
-
in foals include intermittent colic or diarrhea, rough hair coat, poor growth, or drooling. Discomfort may be shown by grinding the teeth, lying on the back to try to ease abdominal pain, or interrupted nursing. The foal
-
should make sure that your infant is at least getting 15-20 minutes of tummy time during the day. Once your child is comfortable with lying on their tummy, you can help them build their strength and motor skills via different
Connections
Show on mapWhich neuronsOne of thousands of tiny dials inside each layer. It reads the text so far and produces a number, its activation, that nudges what the model writes next. in earlier layersOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.
Fed by earlier layers
- L16:2955 suppresses
- L11:1148 excites
- L8:1948 excites
- L11:626 suppresses
- L7:1851 suppresses
- L4:206 suppresses
- L9:722 weakly excites
- L15:1986 weakly excites
- L4:2687 weakly excites
- L9:2151 weakly suppresses
- L9:2526 weakly suppresses
- L9:2564 weakly excites
- L18:1015 weakly excites
- L14:1947 weakly excites
- L9:1374 weakly excites
- L6:2059 weakly suppresses
- L10:2248 weakly suppresses
- L16:1936 weakly suppresses
- L15:2694 weakly excites
- L10:1442 weakly excites
- L11:1530 weakly suppresses
- L9:517 weakly suppresses
- L18:208 weakly suppresses
- L2:2838 weakly excites 'Head Start' and 'follow up'
Feeds later layers
- L23:1677 weakly excites
- L27:246 weakly excites
- L27:381 weakly suppresses
- L22:2295 weakly excites
- L22:2318 weakly excites
- L22:499 weakly suppresses
- L24:332 weakly suppresses
- L26:929 weakly excites
- L22:38 weakly excites
- L24:1796 weakly suppresses
- L23:1657 weakly excites
- L23:1849 weakly excites
- L23:2899 weakly excites
- L23:2037 weakly excites
- L22:2588 weakly suppresses
- L22:1730 weakly excites
- L21:1868 weakly suppresses
- L25:3023 weakly suppresses medical specialists and eye doctors
- L26:951 weakly suppresses
- L21:2257 weakly suppresses
- L21:2203 weakly excites
- L24:2465 weakly suppresses
- L24:1560 weakly suppresses
- L26:845 weakly suppresses
Activation histogramA chart of how often a neuron takes each activation value.
How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.
Logit lensReading which output tokens a neuron pushes toward or away from, straight off its output weights, before the later layers have their say.
The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layersOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. instead.
Promotes
- coli
- soát
- nage
- ichier
- music
- NCY
- 妮
- hatt
- irting
- NDAR
Suppresses
- 枕
- bed
- ammed
- ']}'
- odiac
- บน
- 醒来
- foss
- offs
- 床上