AI Lab website: research interests as effects

The research phrases are quoted word for word from each person’s published profiles and sites, pending their approval; the tokens, scores and depth are computed offline with named models, the faculty headshots are untouched, and the links open each page with motion on.

  1. Sample RT: River of Tokens

    1. Sample RT on a laptop, moment 1: the first row of People and research reads as plain text, and the rows below are still lanes of token chips1The first row in plain text, the rest still token lanes
    2. Sample RT on a laptop, moment 2: four rows open, each with a name, a title line, five phrases and sources2Scroll: four rows open
    3. Sample RT on a laptop, moment 3: one row's tokens moving into place in the person's paragraph3A row’s tokens moving into its paragraph
    4. Sample RT on a laptop, moment 4: all eleven rows in plain text, then Contact4All eleven rows in plain text
    Sample RT, first screen on a phone: the river of titles with one lane of token chipsPhone, first screen

    R’s river of titles still flows under the skyline, and one lane now carries the faculty’s research phrases as the CLIP tokenizer cuts them, each token in a light blue chip with its token ID under it.

    In People and research, each person’s lane of tokens stops as you scroll and the tokens move back into their paragraph, beside their name.

    Open the page for sample RT, River of Tokens, with motion on

    The original R, River of Titles, with motion on

  2. Sample BI: From the Building to the Ideas

    1. Sample BI on a laptop, moment 1: the Babbio Center entrance as a cloud of points, turned to show its depth1The photo as points, turned to show its depth
    2. Sample BI on a laptop, moment 2: the nearest points leave the building and gather in small clusters2Scroll: the nearest points leave first
    3. Sample BI on a laptop, moment 3: most clusters in place, each with its phrase and name beside it3Clusters land and their phrases appear
    4. Sample BI on a laptop, moment 4: all 55 phrases on the map, with the methods note in the corner4All 55 phrases on the map
    Sample BI, first screen on a phone: the email button above the entrance as points, with the methods notePhone, first screen

    The Babbio plaza photo turns into 40,024 points and tilts into the depth that Depth Anything V2 Small computed for it.

    On scroll the points stream off the building and gather in 55 small clusters where BAAI/bge-small-en-v1.5 and UMAP placed the faculty’s phrases, and each phrase appears beside its cluster with the person’s name.

    Open the page for sample BI, From the Building to the Ideas, with motion on