7 ms·
Google Open-Sources Trillion-Parameter AI Language Model Switch Transformer
- strin 6y agolol they didn’t open source the model weights
- JZL003 6y agohah! yeah that's what I was looking for too
- verdverm 6y agodata is the new oil, what's the analogy for the data industry's impact on society akin climate change?
- ALittleLight 6y agoSurveillance Capitalism https://en.wikipedia.org/wiki/Surveillance_capitalism https://en.wikipedia.org/wiki/Surveillance_capitalism
- selfhoster11 6y agoSocial Cooling (https://www.socialcooling.com/ https://www.socialcooling.com/).
- notretarded 6y agoIn the source code it says "I have discovered truly marvelous weights for this, which this header file is too small to contain"
- buildbot 6y agoIf I did the math right it would be 3.12TB of weights, maybe they are trying to upload it to gdrive still. (/s, probably)
- schoen 6y agoThey released more data than that for their Google Books n-grams datasets: https://storage.googleapis.com/books/ngrams/books/datasetsv3.html https://storage.googleapis.com/books/ngrams/books/datasetsv3... (I don't remember exactly how much it is, but I remember that the old version was already in the terabytes.)
- lifthrasiir 6y agoAnother example of Google giving much data away is 50 trillion digits of pi [1], which contains about 42 TB of data (decimal and hexadecimal combined). [1] https://storage.googleapis.com/pi50t/index.html https://storage.googleapis.com/pi50t/index.html
- swirepe 6y ago(They are definitely going to exceed their storage quotas.) I want to see how well weights for these models compress, but it will take me some time to run this code and generate some. I'm guessing they won't compress well, but I can't articulate a reason why.
- wisty 6y agoIf weights compress, they have low information, which would suggest they're either useless or the architecture is bad.
- dheera 6y agoThe Waymo open dataset is about 1TB. I don't think releasing a 3TB dataset would present a technical challenge for Google.
- londons_explore 6y agoEven a 3PB model would be very doable for Google...
- pradn 6y agoThe daily upload quota for a user is ~750 GB. It'll take a few days to upload that much data to Google Drive!
- coef2 6y agoIs this because they are afraid of the model misused, like used for generating fake reviews? It is frustrating that I've been hearing great news on NLP but am able to try none of them myself.
- briga 6y agoIt's because the model weights are the valuable thing here. The fancy new architectures are nice and everything, but transformer models are a dime a dozen these days. Seems like they're using this as an example to point at and say "Hey, look at us, we support open source!", whereas unless you're willing to go ahead and spend a small fortune on compute (possibly using their GPUs), these models are somewhat useless.
- buildbot 6y agoIf the model weights aren’t open source, the model isn’t open source in my opinion. It’s not like a switch transformer is that much more complicated to slap into pytorch than any other new research. Also in my personal opinion, while yes this model performs great, it’s ridiculously parameter inefficient. 1.6T parameters is 3.12TB using bfloat 16. I’d rather take 7x longer to train and have a model I can actually put places.
- amelius 6y ago> it’s ridiculously parameter inefficient How do you know that? Perhaps with this methodology you really need those 3.12TB to reach comparable performance?
- buildbot 6y agoIf you read through the paper where they present switch transformers, they train all models to the same performance, so for example t5-large, at 770M parameters[1] is just as good as this model. That’s only 1.54GB [1] https://huggingface.co/t5-large https://huggingface.co/t5-large
- thesz 6y agoI beg to disagree. [1] provides one with a whole-data-set training method (ADMM, one of such methods). Page 8 contains figure 2(b) - accuracy of training after specified amount of time. Note that ADMM start where stochastic gradient stops. [1] https://arxiv.org/pdf/1605.02026.pdf https://arxiv.org/pdf/1605.02026.pdf At [2] I tried to apply logistic regression trained using reweighted least squares algorithm on the same Higgs boson data set. I've got the same accuracy (64%) as mentioned in the ADMM paper with much less number of coefficients - basically, just the size of input vector + 1 instead of 300 such rows of coefficients and then 300x1 affine transformation. When I added squares of inputs (for the simplest approximation of polynomial regression) and used the same reweighted iterative least squares algorithm, I've got even better accuracy (66%) for double the number of coefficients. [2] https://github.com/thesz/higgs-logistic-regression https://github.com/thesz/higgs-logistic-regression There's a hypothesis [3] that SGD and ADAM are best optimizers because that everyone use and report on. Rarely if ever you get anything that differ. [3] https://parameterfree.com/2020/12/06/neural-network-maybe-evolved-to-make-adam-the-best-optimizer/ https://parameterfree.com/2020/12/06/neural-network-maybe-ev... So, answering your question of "how do you know" - researchers at Google cannot do IRLS (search provides IRLS only for logistic regression in Tensorflow), they cannot do Hessian-free optimization ([4], closed due lack of activity - notice the "we can't support RNN due to the WHILE loop" bonanza), etc. All due to the fact they have to use Tensorflow - it just does not support these things. https://github.com/tensorflow/tensorflow/issues/2682 https://github.com/tensorflow/tensorflow/issues/2682 I haven't seen anything about whole-data-set optimization from Google at all. That's why I (and only me - due to standing I take and experiments I did) conclude that they do not quite care about parameter efficiency.
- amelius 6y agoSince this type of model can be used for the task of "question answering", would it be possible to create a search engine with this type of model?
- sp332 6y agoYou could make something like an "expert system" with it. But since it doesn't seem to link back to the source, you'd never know if it was giving you a real answer or just making something up.
- whimsicalism 6y agoA bit like people
- qayxc 6y agoNot exactly. People can be asked to provide their sources or reasoning, which an informed user can easily check and verify. This is not the case with black-box models. They only point to an answer/give a reply without any justification or reasoning behind it. This is actually a very severe problem with black-box models: ultimately they cannot be trusted, because it's very hard to verify whether the learned objective function matches the intended objective function (this is called the alignment problem [1].) Optimisers tend to produce Clever Hans instances whenever they can, because it's the cheapest and therefore most optimal solution. Even if this becomes obvious from failure cases (e.g. common misclassification in image recognition systems), it's still not obvious which clues the system used that lead to the misclassification. This is in contrast to a person, who an be queried as to why and how they arrived at their conclusion. [1] https://bdtechtalks.com/2021/01/18/ai-alignment-problem-brian-christian/ https://bdtechtalks.com/2021/01/18/ai-alignment-problem-bria...
- PartiallyTyped 6y agoGoogle already uses a combination of NLP + PageRank to serve search queries [1]. [1] https://en.wikipedia.org/wiki/RankBrain https://en.wikipedia.org/wiki/RankBrain
- huslage 6y agoCan someone translate this post into english for those of us not imbued with AI knowledge?
- gradys 6y agoOne key idea here is to use a very large number of parameters (model weights), but only use some subset of the parameters on each example. The parameters are divided up into blocks called "experts", and then some subset of experts are used on any given input. Which subset is used is chosen by the model itself in a data-dependent manner. This can be thought of as letting the model specialize different experts to handle different situations. The advantage, as they show, is that the model can train to a given level of performance much faster with a fixed amount of computing power compared to an architecture that uses all parameters on every step. This might be because it allows you to have a very large number of parameters that can store a lot more specialized information without incurring as much of a computational cost. Of course the downside is that you end up with a very large model that literally won't fit in a lot of environments.
- stingraycharles 6y agoAnother naive question: why is this better than creating a separate, smaller model for each expert?
- p1esk 6y agoYes, you can think of this as a collection of small models, with another model choosing which smaller model to use for each input.
- gwenzek 6y agoNot really, the dispatching between experts happen for every word and every layer, so you can't easily isolate distinct models.
- probably_wrong 6y ago
- airocker 6y agoWould GDPR or other regulation apply to the data that these model is trained with? Is it not a risk that the model will record some private information in someone's email?
- neximo64 6y agoMaybe host and use it in a non GDPR environment, it's always going to be ambiguous and risky.
- ma2rten 6y agoYes, it has been shown [1] that these models can memorize information even if it only appears once in the training data. It could potentially cause issues with privacy and especially copyright. However, in this case it's not relevant because they didn't release the model weights. [1] https://arxiv.org/abs/2012.07805 https://arxiv.org/abs/2012.07805
- airocker 6y agoEven if they did not, they have it for themselves. If this uses any personal data, then GDPR and other regulations may not be getting respected.
- visarga 6y agoIf they have a policy to respond to GDPR requests, and maybe if they use a bloom filter to avoid repeating the training data, then it should be ok?
- bidirectional 6y agoI feel like a future version of GPT will easily be able to answer 'What is the home address of ...@gmail.com', given the abundance of data it is trained on and the widespread proliferation of leaked databases.
- jariel 6y agoSo, almost every possible ngram ever used is a 'input'? Can someone describe what the lexical reality of 3.5T inputs actually means? I feel like this is 'Deep Memorization' instead of 'Deep Learning'. Like a Doctor who passes everything merely by memorizing the textbook with absolutely no ability beyond that.
- visarga 6y agoEven so, memorisation implies word by word information retrieval and interpolation, it's not a hash table.
- jariel 6y agoYes, I mean, that's cool, but I'm just thinking that this isn't quite the AI we were thinking about before. It's like a 'new form of storage and lookup' as opposed to the kind of 'magic algorithm' we usually think of when we think of AI. Or maybe that's just me.
- azinman2 6y agoMagic algorithms don’t exist. That’s the true reality of AI. I also felt the same thing and became dismayed as an undergrad at what was going on in computation having also studied neurobiology. But as it turns out, a shit ton of data with some stats can get you very, very far.
- jariel 6y agoI didn't mean 'magic' (I know it's not that) - I just meant to imply that I think of AI as a 'function' not a 'lookup'. Inputs -> Outputs not Search -> Response Like if you train AI on a small dataset, it feels like what it's doing afterwords is a 'function' or 'algorithm' using what is in the end some arcane algebra. But if you train on all the data in the world, with a trillion parameters ... well ... I kind of feel that 'all the data' is in an AI-style datastructure, that we are 'querying' with AI. But that's just an observers abstraction.
- worik 6y agoAt that many trillion parameters every utterance can be a special case....
- johndoe42377 6y agoThere is something fundamentally wrong with these models. The brain "works" because it's evolved structure matches or reflects reality. It is not about having billions of neurons, but about to have the right structure which matches the environment. My favourite example is how butterflies evolve pictures of eyes on its wings to scare predators, having literally no idea about existence of other creatures. It has been evolved because other creatures have eyes, and they are there, of course. The proper structure of neural networks must be based on such fundamental features, like "most of creatures have eyes" and similar ones. Brain does not have a flat structure, like a billion x billion matrix. It is more clever and simpler that that. A language model must be based on the fundamental notion that there are nouns (things), verbs (processes) and adjectives (attributes). It is that simple.
- dahele 6y agoSlightly apples to oranges comparison. The brain has an incredibly complex architecture, which evolved over millions of years. On top of that, it then develops throughout a human's lifespan. The brain we observe is a "finished product", and even then it has ~150 trillion synapses to do computations [0]. Even massive neural networks have a relatively simple architecture before they are trained. Part of the training process is effectively learning more complex architectures, which are manifested by changing weights. What I'm getting it is that artificial neural networks aren't equivalent to the brain - ANNs are both learning their own structure, on top of the circuits actually doing computations. They are doing the work of millions of years of evolution, genetics, developmental biology, interaction with the environment etc. Perhaps it's to be expected that ANNs will need orders of magnitude greater number of parameters than a brain. An interesting development is meta-learning, where we separate the process for learning the architecture (this could be using deep learning, but not necessarily) with the network actually doing computation (equivalent to the brain). > A language model must be based on the fundamental notion that there are nouns (things), verbs (processes) and adjectives (attributes). I agree, but how does the brain represent these concepts? Some would argue that ANNs do have these concepts, just hidden away in abstract vector representations. Take the visual system, which has been extensively studied - we see the brain represents contrast, edges, shapes and so on very similarly to convolutional NNs. [0] It's likely that this number doesn't come close to capturing the brain's complexity, as it doesn't incorporate parameters like long-term potentiation/depression, synchronization, firing rates, habituation vs sensitization, immunomodulation and likely so much more we haven't yet discovered.