How AI reads chips

1 Your word measured

noun 15 senses 30 synonyms

Definitions come from WordNet, a hand-curated dictionary.

2 To an AI, it's pieces measured

DeepSeek V3 chips 1 token

Mistral 7B chips 1 token

Qwen3 8B chips 1 token

This is the tokenizer splitting text, not the model understanding it.

3 Synonyms, antonyms, and the AI measured

Qwen3 8B

DeepSeek V3

Mistral 7B

The model’s neighbours

chip0.21 lamps0.12 disks0.12 pieces0.11 bulbs0.11 crystals0.11 sandwiches0.11 counters0.11

Dictionary synonyms measured

chip0.21 fries0.08 check0.06 nick0.06 crisp0.05 scrap0.05 bit0.04 french fries french-fried potatoes flake fleck cow chip cow dung buffalo chip potato chip Saratoga chip poker chip microchip micro chip silicon chip microprocessor chip chip shot chipping splintering

Shade shows how close the model puts each word to chips, 1 of 8 also appear in the dictionary. 7 of 30 dictionary synonyms are in this build; the rest have no vector to compare yet.

The model’s neighbours

chip0.53 cards0.23 clips0.21 chunks0.19 fries0.19 memberships0.19 philips0.19 crackers0.19

Dictionary synonyms measured

chip0.53 fries0.19 crisp0.12 scrap0.10 bit0.09 nick0.08 check0.06 french fries french-fried potatoes flake fleck cow chip cow dung buffalo chip potato chip Saratoga chip poker chip microchip micro chip silicon chip microprocessor chip chip shot chipping splintering

Shade shows how close the model puts each word to chips, 2 of 8 also appear in the dictionary. 7 of 30 dictionary synonyms are in this build; the rest have no vector to compare yet.

The model’s neighbours

chip0.43 cards0.20 mips0.20 slips0.19 ips0.18 clips0.17 slippers0.17 phones0.16

Dictionary synonyms measured

chip0.43 crisp0.10 nick0.09 fries0.09 scrap0.06 bit0.06 check0.04 french fries french-fried potatoes flake fleck cow chip cow dung buffalo chip potato chip Saratoga chip poker chip microchip micro chip silicon chip microprocessor chip chip shot chipping splintering

Shade shows how close the model puts each word to chips, 1 of 8 also appear in the dictionary. 7 of 30 dictionary synonyms are in this build; the rest have no vector to compare yet.

Model neighbours are distributional, not dictionary synonyms. Two words can be close because they appear in similar sentences, which is why an antonym can outscore a synonym here.

4 Its personality measured

Qwen3 8B

DeepSeek V3

Mistral 7B

rare common

concrete abstract

casual formal

everyday technical

negative positive

mild intense

powerless powerful

small big

rare common

concrete abstract

casual formal

everyday technical

negative positive

mild intense

powerless powerful

small big

rare common

concrete abstract

casual formal

everyday technical

negative positive

mild intense

powerless powerful

small big

Scores are projections onto axes we defined from anchor words, not labels the model assigns.

5 Surprising neighbours measured

Qwen3 8B

DeepSeek V3

Mistral 7B

beads0.107 tiles0.107 nuts0.107 cards0.106 cells0.105

These sit just outside the closest neighbours in panel 3, and no dictionary lists any of them as related to chips. That gap is the model’s own learned association.

tiles0.179 mips0.177 fellowships0.174 devices0.173 disks0.171

These sit just outside the closest neighbours in panel 3, and no dictionary lists any of them as related to chips. That gap is the model’s own learned association.

machines0.161 beans0.159 lights0.155 chocolate0.153 ships0.152

These sit just outside the closest neighbours in panel 3, and no dictionary lists any of them as related to chips. That gap is the model’s own learned association.

These are statistical associations in the training data, not the model thinking.