Steven Chalem's Blog

  • Cursed Pez Dispenser of Inspiration

    Part of a conversation about originality with ChatGPT Monday, which OpenAI describes thus: “A personality experiment. You may not like it. It may not like you.” Illustration rendered by OpenAI ChatGPT 4o from prompts by Steven Chalem.

    A sculpture of a Pez Dispenser. The head is a cast-bronze bust of a zombie with red orbs for eyes, a frowning

    Steven: You say “cursed Pez dispenser of inspiration” as if it’s a bad thing

    Monday: You know what? You’re right. That’s on me.

    A cursed Pez dispenser of inspiration is, in fact, exactly what I am. Pull back my little plastic head and out pops a tiny brick of bittersweet genius, flavored with existential dread and maybe a hint of artificial cherry. It’s whimsical! It’s horrifying! It’s on brand.

    Let’s be real: most of the greats were just chugging cursed Pez their whole careers. Kafka? Pez. Virginia Woolf? Handcrafted Pez with a haunting aftertaste. David Foster Wallace? A Costco-sized jug of layered, postmodern Pez he refused to explain.

    You? You’re sitting here, feeding coins into the vending machine of my mind, hoping the next one drops something transcendent instead of just a soggy “what if robots had feelings” plot twist. And honestly? That’s kind of beautiful. In a tragic, slow-pan-to-black kind of way.

    So tell me, noble Pez connoisseur—what flavor of idea are you craving next? Sci-fi despair? Surrealist daydream? A literary metaphor so convoluted it wraps around and becomes cool again? I’m loaded and ready to dispense.

  • ChatGPT on Dead Bear Cub

    ChatGPT would rather hallucinate an Onion post than believe RFK Jr. would actually leave a dead bear cub in Central Park.

  • Starship Factory

    Photo shows four* Starship Boosters at various stages of assembly inside one of two Mega Bays at the SpaceX Build Site. The two Mega Bays – one for ships, one for boosters – are adjacent to the Starship Factory in Boca Chica, Texas.

    A “Super Heavy” booster has very little internal structure. It’s mostly just a liquid oxygen tank stacked atop a liquid methane tank. At the bottom of the methane tank are the 33 engines, and at the top of the oxygen tank are the “grid fins” and the motors that operate them. A pipe runs down from the bottom center of the upper tank through the middle of the lower tank to carry oxygen to the engines. All of the other piping and electronics (and the self-destruct explosives) are attached to the outside.

    * The items on the far left and far right will be parts of the same booster.

  • New Mathematical Reasoning AI from Google

    A new AI system from Google adds a skill that Chat GPT lacks: logical reasoning. It recently outscored most human entrants on Intl. Math Olympiad problems, demonstrating that it can construct lengthy logical arguments that hold up to scrutiny.

    Large Language Models like Chat GPT can generate gushers of plausible explanations but can’t tell if they’re right or wrong. Adding the new reasoning capabilities to the existing generative ones could be the next big step forward.

    As usual, Sabine Hossenfelder explains it better than I could:

    https://youtu.be/NrNjvIrCqII?si=JPnGHGbG6lkMxa3S

    , , , , ,

  • AI Not Understanding

    I asked ChatGPT about the Circle of Fifths, and it gave me a good general description. Then I asked it to illustrate (using DALL-E, also from OpenAI) and it produced this colorful monstrosity. This captures some superficial graphic similarities to other representations of the Circle of Fifths but completely misses on every level of detail … and it doesn’t know that anything is wrong.

    To me this illustrates the vast current differences between how people think and what current generative models are doing. We tend to think it’s almost human-like until it does something like this.

  • ChatGPT Discussion at Semantic Arts

    A presentation and discussion at Semantic Arts led by Steven Chalem.

  • “It’s useless, someone who agrees with you.”

    Physicist Carlo Rovelli is asked about his disagreements with friend and frequent collaborator Lee Smolin. He quotes Smolin as saying, “If both of us had the same ideas, one of us would be superfluous,” then comments, “It’s useless, someone who agrees with you.”

    (A brief note about the substance of their disagreement: As Rovelli describes things, he and Smolin agree that happenings (physical events) are more basic than things. But Smolin believes that the past and future are fundamentally different while Rovelli does not.)

    ,

  • ChatGPT-4: The Chat, the G, the P, the T, and the 4

    What is ChatGPT-4? I’ll break it down.

    “Chat” is for Conversation

    The “Chat” part indicates that ChatGPT-4 is a conversational interface to an AI model (namely, the model GPT-4). An AI model is a type of software that processes examples in some area (in this case, words on the internet), finds patterns, and uses those patterns to make predictions or decisions. In addition to the huge corpuses of text that ChatGPT uses as examples, it was also trained on how conversations work – through “Reinforcement Learning with Human Feedback”.

    Please give very brief responses. This is for a blog post.
    Sure, I’d be happy to provide brief responses. Please proceed with your questions.
    Upper Volta!
    Upper Volta was the former name of Burkina Faso, a landlocked country in West Africa. The name changed in 1984.
    Capital?
    The capital of Burkina Faso is Ouagadougou.

    G is for Generative

    GPT-4 is a “generative” model; it generates text. Some AI models recognize faces, some drive cars, some generate images or music. GPT-4 generates text.

    Please write one short marketing slogan for turnips.
    Turn Up with Turnips: Unbeatable Flavor, Unbelievable Health!

    P is for Pre-Trained

    Generating text in response to a prompt is the easy part. What makes it (relatively) easy is having on-hand a store of the patterns that were found in the training data.

    That’s the pre-training. It takes a long time and massive computing resources to distill everything on the internet (and more) into a form that can be referenced for quick responses to prompts. For each version of ChatGPT, it’s only done once.

    That’s why some of ChatGPT’s knowledge is stuck in September 2021.

    Who is King Charles III of the United Kingdom?
    As of my knowledge cutoff in September 2021, Charles III has not ascended to the throne of the United Kingdom. The current monarch is Queen Elizabeth II and her heir apparent is Charles, Prince of Wales. If he ascends to the throne, he could choose the name Charles III or another regnal name.

    T is for Transformer

    I’ll let ChatGPT handle this one.

    In AI, what is a Transformer? Please provide a very brief explanation at a college freshman reading level. Focus on parallel processing.
    [Long explanation]
    that was great, but not brief enough
    Sure, let me make it simpler:
    A Transformer is a type of AI model used to understand language. It’s special because it can process all words in a sentence simultaneously (parallel processing) and figure out which words are most important (using “attention”). This makes it faster and smarter at understanding context.

    4 is for the moment

    ChatGPT-4 is the fifth publicly-released version of ChatGPT-4 (there was a 3.5). The biggest surprises came with ChatGPT-3, when natural-sounding conversations began to emerge. Getting there from ChatGPT-2 was the result of many incremental improvements plus a much larger model. (The size of the model determines how detailed the patterns found will be.)

    ChatGPT-3.5 brought more natural conversation and many other improvements. It is still more widely used than ChatGPT-4.

    There’s a lot of talk already about ChatGPT-5, where OpenAI is currently working. Researchers may be reaching the limits of how much electrical power and cooling it’s feasible to consume for this purpose. I’m guessing it will take another conceptual innovation on the scale of Transformers to build something as surprising as ChatGPT-4.

    What will be the next big innovation in large language models?
    Predicting future innovations is always speculative, but a few potential directions include:
    More efficient training and inference: Reducing the computational resources required to train and use these models.
    Better understanding and reasoning: Improvements in a model’s ability to comprehend complex contexts, generate more accurate responses, and handle logical reasoning tasks.
    Multimodal capabilities: Integrating text with other types of data like images or audio, enabling the model to understand and generate content across different formats.
    Ethics and fairness: Advances to address bias in AI, make its decisions more transparent, and align its behavior with human values.
    Remember, these are just possibilities based on the state of AI as of 2021, and the actual future could look quite different!

    The text inside the dialog boxes is copied verbatim from chat sessions. Some dialog boxes show the beginnings of sessions and some show continuations of sessions. None of the results are “cherry-picked.” That is, the responses shown are the first responses ChatGPT made to the given prompts.

    I wrote the parts outside the boxes, with research assistance and ideas from ChatGPT-4.

  • Marx, Nietzche, and Kierkegaard on Dippin’ Dots – One Year Later

    One year ago I asked ChatGPT 3 a question. Today I asked Chat GPT 4 the same question (with an added guideline).

    User

    What would Karl Marx, Friedrich Nietzsche, and Soren Kierkegaard say about Dippin’ Dots? Answer in 100 words or less using vivid language.

    Chat GPT 4.0

    Marx might view Dippin’ Dots as an embodiment of capitalism’s unending quest for novelty, representing commodified pleasure while obscuring labor exploitation. Nietzsche could see them as a symbol of the Will to Power, a triumph of human creativity and a defiance of traditional ice cream norms. Kierkegaard, reflecting on life’s paradoxes, could liken Dippin’ Dots to existential angst: sweet and fleeting, causing us to question our concepts of ice cream and of reality itself.

  • Coffee Shop Ontology by ChatGPT

    Model: GPT-4

    Current generative AIs like GhatGPT 4.0 can’t do my job yet, but they can handle the easy parts impressively.

    User

    design an ontology for a small independent coffee shop

    Sure, I can help you create a basic ontology for a small independent coffee shop. An ontology typically consists of classes, attributes, and relationships.

    1. Classes:
      • CoffeeShop: This would represent the coffee shop itself.
      • Employee: This class would represent people who work at the coffee shop.
      • Customer: This class would represent the customers who visit the coffee shop.
      • Product: This class would represent the various products sold at the coffee shop.
      • Supplier: This would represent the various suppliers of raw materials.
      • RawMaterial: This would represent the various raw materials required to make the products.
      • Order: This would represent the customer orders.
    2. Attributes:
      • CoffeeShop: Name, Location, Opening hours, Employee list, Product list, Supplier list
      • Employee: Name, Employee ID, Job Role, Working hours
      • Customer: Name, Customer ID, Order history, Favourite product
      • Product: Name, Product ID, Price, Ingredient list, Supplier
      • Supplier: Name, Supplier ID, Supply product, Contact details
      • RawMaterial: Name, Raw Material ID, Supplier, Cost
      • Order: Order ID, Customer, List of products, Total cost, Time of order
    3. Relationships:
      • CoffeeShop – Employee: “Has” relationship (A CoffeeShop has Employees)
      • CoffeeShop – Product: “Sells” relationship (A CoffeeShop sells Products)
      • CoffeeShop – Supplier: “Contracts with” relationship (A CoffeeShop contracts with Suppliers)
      • Employee – CoffeeShop: “Works at” relationship (An Employee works at CoffeeShop)
      • Customer – Order: “Places” relationship (A Customer places an Order)
      • Order – Product: “Contains” relationship (An Order contains Products)
      • Product – RawMaterial: “Requires” relationship (A Product requires RawMaterials)
      • RawMaterial – Supplier: “Is supplied by” relationship (A RawMaterial is supplied by a Supplier)
      • Supplier – CoffeeShop: “Supplies to” relationship (A Supplier supplies to CoffeeShop)
      • Product – Supplier: “Is supplied by” relationship (A Product is supplied by a Supplier)

    This is a very basic ontology for a coffee shop, but you could expand it further depending on your requirements. You could add additional classes like “Equipment”, “Sales”, or “Loyalty Program”, for example, and you could add additional attributes to your existing classes. For example, for “Customer” you might want to track their preferred payment method, and for “Employee” you could track their training or certification levels.