Should LLMs just treat text content as an image?

(seangoedecke.com)

48 points | by ingve 6 days ago ago

38 comments

  • wongarsu 2 hours ago

    Look long enough at literature on any machine learning task, and someone invariably gets the idea to turn the data into an image and do machine learning on that. Sometimes it works out (turning binaries into images and doing malware detection with a CNN surprisingly works), usually it doesn't. Just like in this example the images usually end up as a kludge to fix some deficiency in the prevalent input encoding.

    I can certainly believe that images bring certain advantages over text for LLMs: the image representation does contain useful information that we as humans use (like better information hierarchies encoded in text size, boldness, color, saturation and position, not just n levels of markdown headings), letter shapes are already optimized for this kind of encoding, and continuous tokens seem to bring some advantages over discrete ones. But none of these advantages need the roundtrip via images, they merely point to how crude the state of the art of text tokenization is

    • vanderZwan 17 minutes ago

      Doesn't this more or less boil down to OCR scans of books having more privileged information than a plaintext file? In which case a roundtrip won't add anything?

      [0] https://web.archive.org/web/20140402025221/http://m.nautil.u...

    • pegasus 34 minutes ago

      Exactly. The example the article give of reducing resolution as a form of compression highlights the limitations of the visual-only proposal. Blurring text is a poor form of compression, preserving at most information about paragraph sizes. Summarizing early paragraphs (as context compression does in coding agents) would be much more efficient.

  • DonHopkins a minute ago

    [delayed]

  • bonsai_spool 2 hours ago

    This doesn’t cite the very significant example of DeepVariant (and as of 10/16/25 DeepSomatic) which convert genomic data to images in order to find DNA mutations. This has been done since the late 2010s

    https://google.github.io/deepvariant/posts/2020-02-20-lookin...

  • LysPJ 2 hours ago

    Andrej Karpathy made an interesting comment on the same paper: https://x.com/karpathy/status/1980397031542989305

    • onionisafruit 7 minutes ago

      > It makes two characters that look identical to the eye look as two completely different tokens internally in the network. A smiling emoji looks like a weird token, not an... actual smiling face, pixels and all

      This goes against my limited understanding of how LLMs work — and computers generally for that matter. Isn’t that rendering of a smiling emoji still just a series of bits that need to be interpreted as a smiley face? The similar looking characters point makes more sense to me though assuming it’s something along the lines of recognizing that “S” and “$” are roughly the same thing except for the line down the middle. Still that seems like something that doesn’t come up much and is probably covered by observations made in the training corpus.

      All that said, Karpathy knows way more than I will ever know on the subject, and I’m only posting my uninformed take here in hopes somebody will correct me in a way I understand.

  • mannykannot 27 minutes ago

    Language was spoken long before it was written (or so it seems.) This article almost suggests that sound might be a superior input medium over either digital text or images.

    • falcor84 10 minutes ago

      I've never worked in that area, but recall reading about how images of spectrograms are often superior inputs to neural nets in comparison to the raw audio data.

  • mohsen1 2 hours ago

    My understanding is that text tokens are too rigid. The way we read is not to process each character (tokens for LLMs) precociously but to see a word or sometimes a collection of familiar words and make sense of writing. That concept that we understand from written text is really what we read and not letter or words exactly. This is why we can easily read written text with typos. They are just similar enough. By letting LLMs not to be too hung up on exact tokens and "skim" through text we can make them more efficient just like how humans efficiently read.

    • bonsai_spool 2 hours ago

      You miss that we already have ‘context’ when we begin reading something, and that probably enables our fast reading. Maybe there’s a way to give that background setting information to an llm but then we could also just have it read the entire input stream

    • lupire an hour ago

      Google AI mode:

      User: cna uto nfeir waht tihs si msieplnig?

      Google: Here are top web results for exploring this topic [few websites containing or discussing gibberish in general]

      User: what was my previous question?

      Google: Your previous question was "cna uto nfeir waht tihs si msieplnig?". I was able to understand this through a combination of contextual understanding and sub-word tokenization, which allowed me to infer the intended meaning despite the numerous typos. The corrected question is: "Can you infer what this is misspelling?".

      • Miraltar 13 minutes ago

        I couldn't read it

  • leemcalilly 43 minutes ago

    and reading (aka “ocr”) is the fastest way for the brain to process language.

  • aitchnyu an hour ago

    The amount of video/imagery to make a million tokens vs the amount of text to do the same is a surprisingly low ratio. Did they have the same intuition?

  • qiine 39 minutes ago

    or maybe 3d objects, since that's closer to what real life is and how the brain shaped itself around?

  • vindex10 2 hours ago

    reminds me of the difference between fasttext and word2vec.

    fasttext can learn words it haven't seen before by combining words from ngrams, word2vec can learn better meaning of the whole words, but then missing out on the "unknown words".

    image tokens are "text2vec" here, while text tokens are a proxy towards building a text embedding of even unseen before texts.

  • ghoul2 2 hours ago

    But does this not miss the "context" that the embeddings of the text tokens carry? An LLM embedding of a text token has a compressed version of the entire preceding set of tokens that came before it in the context. While the image embeddings are just representations of pixel values.

    Sort of at the level of word2vec, where the representation of "flies" in "fruit flies like a banana" vs "time flies like an arrow" would be the same.

  • themoxon 2 hours ago

    There's a new paper from ICCV which basically tries to render every modality as images: https://openaccess.thecvf.com/content/ICCV2025/papers/Hudson...

  • ToJans 2 hours ago

    A series of tokens is one-dimensional (a sequence). An image is 2-dimensional. What about 3D/4D/... representation (until we end up with an LLM-dimensional solution ofc).

    • dvt 2 hours ago

      This isn't exactly true, as tokens live in the embedding space, which is n-dimensional, like 256 or 512 or whatever (so you might see one word, but it's actually an array of a bunch of numbers). With that said, I think it's pretty intuitive that continuous tokens are more efficient than discrete ones, simply due to the fact that the LLM itself is basically a continuous function (with coefficients/parameters ∈ ℝ).

      • wongarsu 2 hours ago

        We call an embedding-space n-dimensional, but in this context I would consider it 1-dimensional, as in it's a 1d vector of n values. The terminology just sucks. If we described images the same way we describe embeddings a 2 megapixel image would have to be called 2-million-dimensional (or 8-million-dimensional if we consider rgba to be four separate values)

        I would also argue tokens are outside the embedding space, and a large part of the magic of LLMs (and many other neural network types) is the ability to map sequences of rather crude inputs (tokens) into a more meaningful embedding space, and then map from a meaningful embedding space back to tokens we humans understand

  • pcwelder 2 hours ago

    I ϲаn guаrаntее thаt thе ОСR ϲаn't rеаd thіs sеntеnсе ϲоrrесtlу.

  • Havoc 2 hours ago

    Seems wildly counterintuitive to me frankly.

    Even if true though not sure what we’d do with it. The bulk of knowledge available on the internet is text. Aside from maybe YouTube so I guess it could work for world model type things? Understanding physical interactions of objects etc

    • hshdhdhehd 2 hours ago

      Trivial to convert text to images to process. But counter-intuitive to me too.

    • bilsbie 2 hours ago

      All text is technically converted to images before we see it.

      • thfuran 35 minutes ago

        Only if you see it instead of hearing it or touching it.

  • onesandofgrain 2 hours ago

    A picture is worth a thousand words

    • hshdhdhehd 2 hours ago

      And a picture of a thousand words is worth a thousand words.

      • genghisjahn 2 hours ago

        I type at .08 pictures per minute.

  • metalliqaz an hour ago

    Future headline: "The unreasonable effectiveness of text encoding"

  • nacozarina 2 hours ago

    the enshittifiers simply haven't yet weighted image processing fees with potential token charges; once they have, your cost advantage goes bye bye