I just vibe coded a whole iOS app in Swift (without having programmed in Swift before, though I learned some in the process) and now ~1 hour later it's actually running on my physical phone. It was so ez... I had my hand held through the entire process. Very cool. I didn't even read any docs at all, I just opened a ChatGPT convo and followed instructions.

GPT 4.5 + interactive comparison :)

Today marks the release of GPT4.5 by OpenAI. I've been looking forward to this for ~2 years, ever since GPT4 was released, because this release offers a qualitative measurement of the slope of improvement you get out of scaling pretraining compute (i.e. simply training a bigger model). Each 0.5 in the version is roughly 10X pretraining compute. Now, recall that GPT1 barely generates coherent text. GPT2 was a confused toy. GPT2.5 was "skipped" straight into GPT3, which was even more interesting. GPT3.5 crossed the threshold where it was enough to actually ship as a product and sparked OpenAI's "ChatGPT moment". And GPT4 in turn also felt better, but I'll say that it definitely felt subtle. I remember being a part of a hackathon trying to find concrete prompts where GPT4 outperformed 3.5. They definitely existed, but clear and concrete "slam dunk" examples were difficult to find. It's that ... everything was just a little bit better but in a diffuse way. The word choice was a bit more creative. Understanding of nuance in the prompt was improved. Analogies made a bit more sense. The model was a little bit funnier. World knowledge and understanding was improved at the edges of rare domains. Hallucinations were a bit less frequent. The vibes were just a bit better. It felt like the water that rises all boats, where everything gets slightly improved by 20%. So it is with that expectation that I went into testing GPT4.5, which I had access to for a few days, and which saw 10X more pretraining compute than GPT4. And I feel like, once again, I'm in the same hackathon 2 years ago. Everything is a little bit better and it's awesome, but also not exactly in ways that are trivial to point to. Still, it is incredible interesting and exciting as another qualitative measurement of a certain slope of capability that comes "for free" from just pretraining a bigger model.

Keep in mind that that GPT4.5 was only trained with pretraining, supervised finetuning, and RLHF, so this is not yet a reasoning model. Therefore, this model release does not push forward model capability in cases where reasoning is critical (math, code, etc.). In these cases, training with RL and gaining thinking is incredibly important and works better, even if it is on top of an older base model (e.g. GPT4ish capability or so). The state of the art here remains the full o1. Presumably, OpenAI will now be looking to further train with Reinforcement Learning on top of GPT4.5 model to allow it to think, and push model capability in these domains.

HOWEVER. We do actually expect to see an improvement in tasks that are not reasoning heavy, and I would say those are tasks that are more EQ (as opposed to IQ) related and bottlenecked by e.g. world knowledge, creativity, analogy making, general understanding, humor, etc. So these are the tasks that I was most interested in during my vibe checks.

So below, I thought it would be fun to highlight 5 funny/amusing prompts that test these capabilities, and to organize them into an interactive "LM Arena Lite" right here on X, using a combination of images and polls in a thread. Sadly X does not allow you to include both an image and a poll in a single post, so I have to alternate posts that give the image (showing the prompt, and two responses one from 4 and one from 4.5), and the poll, where people can vote which one is better. After 8 hours, I'll reveal the identities of which model is which. Let's see what happens :) Question 1. Poll is in the following post.

The (true) story of development and inspiration behind the "attention" operator, the one in "Attention is All you Need" that introduced the Transformer. From personal email correspondence with the author @DBahdanau ~2 years ago, published here and now (with permission) following some fake news about how it was developed that circulated here over the last few days.

Attention is a brilliant (data-dependent) weighted average operation. It is a form of global pooling, a reduction, communication. It is a way to aggregate relevant information from multiple nodes (tokens, image patches, or etc.). It is expressive, powerful, has plenty of parallelism, and is efficiently optimizable. Even the Multilayer Perceptron (MLP) can actually be almost re-written as Attention over data-indepedent weights (1st layer weights are the queries, 2nd layer weights are the values, the keys are just input, and softmax becomes elementwise, deleting the normalization). TLDR Attention is awesome and a *major* unlock in neural network architecture design.

It's always been a little surprising to me that the paper "Attention is All You Need" gets ~100X more err ... attention... than the paper that actually introduced Attention ~3 years earlier, by Dzmitry Bahdanau, Kyunghyun Cho, Yoshua Bengio: "Neural Machine Translation by Jointly Learning to Align and Translate". As the name suggests, the core contribution of the Attention is All You Need paper that introduced the Transformer neural net is deleting everything *except* Attention, and basically just stacking it in a ResNet with MLPs (which can also be seen as ~attention per the above). But I do think the Transformer paper stands on its own because it adds many additional amazing ideas bundled up all together at once - positional encodings, scaled attention, multi-headed attention, the isotropic simple design, etc. And the Transformer has imo stuck around basically in its 2017 form to this day ~7 years later, with relatively few and minor modifications, maybe with the exception better positional encoding schemes (RoPE and friends).

Anyway, pasting the full email below, which also hints at why this operation is called "attention" in the first place - it comes from attending to words of a source sentence while emitting the words of the translation in a sequential manner, and was introduced as a term late in the process by Yoshua Bengio in place of RNNSearch (thank god? :D). It's also interesting that the design was inspired by a human cognitive process/strategy, of attending back and forth over some data sequentially. Lastly the story is quite interesting from the perspective of nature of progress, with similar ideas and formulations "in the air", with a particular mentions to the work of Alex Graves (NMT) and Jason Weston (Memory Networks) around that time.

Thank you for the story @DBahdanau ! "Links in the reply followup" (not a huge fan :p)
referenced papers:

Attention paper:
"Neural Machine Translation by Jointly Learning to Align and Translate"
arxiv.org/abs/1409.0473

Transformer paper:
"Attention is All You Need"
arxiv.org/abs/1706.03762

Alex Graves paper around that time with similar soft pooling operations:
"Neural Turing Machines"
arxiv.org/abs/1410.5401
+the referenced (at the time super impressive, inspiring and forward-looking) handwriting paper, this is 2013!:
"Generating Sequences With Recurrent Neural Networks"
arxiv.org/abs/1308.0850

Jason Weston mentioned paper:
"Memory Networks"
arxiv.org/abs/1410.3916

The referenced Ilya, Oriol, Quoc paper at Google:
"Sequence to Sequence Learning with Neural Networks"
arxiv.org/abs/1409.3215

New (2h13m 😅) lecture: "Let's build the GPT Tokenizer"

Tokenizers are a completely separate stage of the LLM pipeline: they have their own training set, training algorithm (Byte Pair Encoding), and after training implement two functions: encode() from strings to tokens, and decode() back from tokens to strings. In this lecture we build from scratch the Tokenizer used in the GPT series from OpenAI. We will see that a lot of weird behaviors and problems of LLMs actually trace back to tokenization. We'll go through a number of these issues, discuss why tokenization is at fault, and why someone out there ideally finds a way to delete this stage entirely.

"Man-Computer Symbiosis" by Licklider, 1960

I love reading technology prediction documents because the benefit of hindsight is training data. Here, 64 years ago, Licklider imagines computing as a fundamentally intelligence amplification tool.groups.csail.mit.edu/medg/people/ps… Licklider argues that the period of "intelligence augmentation" (IA) may be transient on the path to full automation (AI), but still long enough to be worth thinking through and about.
His citations for what must have felt like rapid progress in both narrow AI and AGI (of that age, i.e. the "general problem solver" [20]) are today known to be false starts that were off track in a quite fundamental way, at that time based on a manual process of encoding knowledge with predicate logic and using production rules of logic and search to manipulate them into conclusions. Today, most of AI is only aware of all of this work as a historical curiosity, it is not part of the "master branch" of the field, it is stuck in a dead end feature branch. And notably, what is considered today the most promising approach (LLMs) were at that time not only completely computationally inaccessible, but also impossible due to the lack of training data of trillions of tokens in digitized forms. (What might be an equivalent of that today?)
The study by the Air Force, estimating that machines alone would be doing problem solving of military significance in 20 years time evokes a snicker today. Amusingly, "20 years away" seems to be a kind of codeword for "no idea, long time". Arguably, I'm not sure that we are there even today, 64 years later. Computers do a lot to increase situational awareness, but decision making of "military significance" afaik is still well within the domain of human computation.

Next frontier of prompt engineering imo: "AutoGPTs" . 1 GPT call is just like 1 instruction on a computer. They can be strung together into programs. Use prompt to define I/O device and tool specs, define the cognitive loop, page data in and out of context window, .run().

https://twitter.com/SigGravitas/status/1642181498278408193

Interesting non-obvious note on GPT psychology is that unlike people they are completely unaware of their own strengths and limitations. E.g. that they have finite context window. That they can just barely do mental math. That samples can get unlucky and go off the rails. Etc.

More good read/discussion on psychology of LLMs. I don't follow in full but imo it is barking up the right tree w.r.t. a framework for analysis. lesswrong.com/posts/D7PumeYT… A pretrained LLM is not an AI but a simulator, described by a statistical physics based on internet webpages. The system evolves given any initial conditions (prompt). To gather logprob it internally maintains a probability distribution over what kind of document it is completing

The hottest new programming language is English This tweet went wide, thought I'd post some of the recent supporting articles that inspired it.
1/ GPT-3 paper showed that LLMs perform in-context learning, and can be "programmed" inside the prompt with input:output examples to perform diverse tasks arxiv.org/abs/2005.14165

🔥 New (1h56m) video lecture: "Let's build GPT: from scratch, in code, spelled out."

We build and train a Transformer following the "Attention Is All You Need" paper in the language modeling setting and end up with the core of nanoGPT. First ~1 hour is 1) establishing a baseline (bigram) language model, and 2) introducing the core "attention" mechanism at the heart of the Transformer as a kind of communication / message passing between nodes in a directed graph.

Didn't tweet nanoGPT yet (quietly getting it to good shape) but it's trending on HN so here it is :) :
github.com/karpathy/nanoG…
Aspires to be simplest, fastest repo for training/finetuning medium-sized GPTs. So far confirmed it reproduced GPT-2 (124M). 2 simple files of ~300 lines Rough example, a decent GPT-2 (124M) pre-training reproduction would be 1 node of 8x A100 40GB for 32 hours, processing 8 GPU * 16 batch size * 1024 block size * 500K iters = ~65B tokens. I suspect this wall clock can still be improved ~2-3X+ without getting too exotic.

Dreambooth (stable diffusion finetuning for personal profile pictures) has been going viral last few days as well, for good reasons it's super fun; Unlike other places stableboost.ai lets you play with infinite variations and experiment and play with your own prompts:

https://twitter.com/tall/status/1600571735455465472

Turns out in a parallel Universe I'd look awesome as a samurai, cowboy and... saint? :D

An interesting historical note is that neural language models have actually been around for a very long time but noone really cared anywhere near today's extent. LMs were thought of as specific applications, not as mainline research unlocking new general AI paths and capabilities E.g. ~20 years ago Bengio et al 2003 (pdf: jmlr.org/papers/volume3…) trained a neural language model. The state of the art GPT+friends of today are the exact same (autoregressive) model, except the neural net architecture is upgraded from an MLP to a Transformer.

Is it the number of examples that matters or the number of presentations to the model during training? E.g. humans used spaced repetition to memorize facts but there are no equivalents of similar techniques in LLMs where the typical training regime is uniform random.

https://twitter.com/_akhaliq/status/1592701993805488128

More generally a few remarkable strategies people use during their training:
1) skim text because they already know it
2) ignore text because it's clearly noise (e.g. they won't memorize SHA256 hashes. LLMs will.)
3) revisit parts that are learnable but not yet learned

The Transformer is a magnificient neural network architecture because it is a general-purpose differentiable computer. It is simultaneously:
1) expressive (in the forward pass)
2) optimizable (via backpropagation+gradient descent)
3) efficient (high parallelism compute graph) (1) because its message-passing-like architecture is general (i.e. completeness) and powerful (i.e. efficiency), able to cover many real-world algorithms and in a small number of compute steps; an an empirical finding.

🥷New (1h55m) Lecture #5: "Becoming a Backprop Ninja"
We take the 2-layer MLP from last lecture and backprop through all of it manually: cross entropy loss, linear layer 2, tanh, batchnorm, linear layer 1, embedding table. I give away answers in the video (yes I had a lot of fun with the thumbnail :D)

I actually mostly built Lexicap so I could share a few snippets of Nick Lane ep :). (I already read the books so I'm ~familiar with the topics, these snippets are just personally newish+notable). (Maybe a great podcast app would make threads like this much easier!) "A cell is basically just a micro version of the planet."
karpathy.ai/lexicap/0318-s… haven't thought about it this way before.

Reading through OpenAI Whisper paper github.com/openai/whisper some notes: Idea 1: keep the neural net and the optimization super simple: vanilla Transformer (2017 style) LLM. The innovation is around 1) what the dataset and the training objective is and 2) the I/O schema that allows a single model to multi-task as a speech recognition swiss-army knife.

Stable Diffusion concepts library huggingface.co/sd-concepts-li… textual inversion is amazing - can train a custom word vector (not otherwise reachable by english text) to mean a concept, based on examples. Opens up many possibilities of condensing objects/styles into special tokens 🚀 prompts may start to take on a mixed english mixed special inverted token forms, like "a photo of <karpathy/cool-object-v7> in the style of <coolperson/trippystyle>".

Computer vision research feels a bit stagnating in a local minimum of 2D texture recognition on ImageNet, COCO etc. This is great but only step 1. Unlocking further progress needs new framework:
1) the data source has to become diverse videos, not individual frames from internet 2) ground truth is compiled from "offline tracker" 3D reconstructions, not human labeling. The reconstructions are aided by solutions from step 1.
3) outputs are (NeRF-like) query-able scene representations, not 1-of-k class labels.

The ongoing consolidation in AI is incredible. Thread: ➡️ When I started ~decade ago vision, speech, natural language, reinforcement learning, etc. were completely separate; You couldn't read papers across areas - the approaches were completely different, often not even ML based. In 2010s all of these areas started to transition 1) to machine learning and specifically 2) neural nets. The architectures were diverse but at least the papers started to read more similar, all of them utilizing large datasets and optimizing neural nets.

Really excellent reading and pointers from @ericjang11, putting into words a new "Just Ask for Generalization" approach/philosophy to AI that the field has been slowly internalizing recently. Few more thoughts in thread ->

https://twitter.com/ericjang11/status/1452298036252708866

The first time I was personally shook by this philosophy was when I saw the "Just tell the AI to be nice" meme on my Twitter, which is the same idea - GPT can be seen as a super multi-task policy (trained via supervised learning), and prompt engineering is the goal conditioning.