euphetar

It's indeed a very challenging task. I picked it because it's one of those where the lack of generalization is apparent. In fact, researchers study commonsense QA and similar tasks to find how we can reach generalization.

I agree, it's very hard to fidn the difference. Especially for GPT or BERT, where the training set is basically the whole internet.

It's a very good question about fur. I would suspect that it would correctly recognise all kinds of dog-like fur and sometimes fail on different furs, like bear furs. But in general NN's are very good at textures, so maybe it will just be good on all kinds of fur. One problem that might arise is that you will show it something fur-like but not really fur, and it will think it's fur.

I agree it's some kind of generalization. Here I don't have enough background to draw the line, perhaps a more theorically oriented person could, but I can't.

I guess the most important judge is that if you devise a benchmark like commonsense Q&A, a neural network fails it. Or how a Tesla will recognise a truck full of red stop signs as a real stop sign, while a "generalizing" thing like a human would definitely know that a core property of a stop sign is that it should be installed near a road. So there is a real problem.

It's repeating the same patterns it learned from text. Maybe with different specific objects, but still. One good way to test this is to ask it to count. It will break very soon. A person is able to build a rule in their head: "one apple is 1, two apples are 2, three apples are... 1+2 = 3".

Let's test GPT-2: https://huggingface.co/gpt2?text=One+apple+is+1%2C+two+apple...

Prompt: "One apple is 1, two apples are 2, three apples are" Model output: "One apple is 1, two apples are 2, three apples are 4, four apples are 5, six apples are 7, seven apples are 8, nine apples are 10, ten apples are 11 (for apples being the perfect length of life)."

Even if you use a special dataset to each it to count, it won't be able to count beyound the examples in the training set. So it's spewing plausible-sounding gibberish at you (i.e. approximating the training set distribution)

It doesn't generalize. Not in the sense that it can't give you a phrase that didn't exist in the training set. It can. But it can't give you a new kind of phrase, of a "kind" that didn't exist in the training set.

If that was true, a model trained on one commonsense Q&A dataset would be able to answer questions from another commonsense Q&A dataset without finetuning. But they can't. It's the same for every task you can find, but especially evident on commonsense reasoning benchmarks. At least last time I actively researched the question, I haven't found a single task where NNs definitely generalize.

When researchers dig in, they find that the neural network is learning wrong things. Word matching between answer and question, learning to model the annotator who asked a lot of questions because his Q&A answers are predictable, and stuff like this.

There was a great review of the problem, but I can't find it, so I will have to link to this article [1], which gives an overview of issues with current NLP models.

[1] https://thegradient.pub/frontiers-of-generalization-in-natur...

It's like this: no one is dependent, but it would be great to be able to support my relatives, they need it now.

One good lens I know is that a neural network is just good at approximating stuff. Trained properly, you can have it approximate a distribution. A conditional distribution like p(animal_species=dog | image=what I am seeing) (discriminative model, e.g. classifier), or even a joint one p(animal,image) (generative model, Autoencoder/GAN/VAE/Diffusion).

There is also an information theoretic lens about compression, which is probably very close to what you are thinking about, but I haven't studied it yet.

Regarding Image -> Animal. An image of an animal is a projection of the animal onto a 2D plane, plus lots of noise. So there is some dependance between an image of an animal and the animal. Biologists can get a lot of information looking at a photo of an animal. In some sense they are always looking at two images from their eyes.

But the problem you are talking about is indeed serious, and far from solved. You can't understand the real world from 2D images with the current approaches. Ideally we want neural networks to build a 3D (or even 4D, with time?) model of reality. Instead we find them trying to guess labels based on patterns. May favourite example is the tiger-dog [1]. Still, there is evidence NNs are doing some clever things [2]. My guess is that the problem is that we just haven't found a way to formulate the task for the solution we want. In the current formulation it's easiest for the model to minimize the loss by sticking to patterns, so why do something else?

There is a lot of research on more applied ML that asks the questions you are asking. It's just that this paper is another attempt at a theoretical explanation.

I agree on the self-driving cars. We can't have truly self-driving cars until a model can generalize, which none can't at the moment. The core question is: if we had a "dumb" model that does clever averaging and successfully covers 99% cases, such that the car is dumb in special cases, but smarter than most human drivers in usual cases, would it justify deploying the cars? If this was the case, dumb ML might be enough for self-driving. It's definitely enough for self-driving in walled garden conditions, so there is some evidence that with enough data we can brute force our way to a tolerable solution.

[1] https://www.dropbox.com/s/ucvflwwrm8idnp6/photo_2022-03-29_2... [2] https://distill.pub/2020/circuits/

Well since the next layer outputs are a linear transformation plus some nonliearity of the previous function, it's a fact that it's a change in representation. But I guess the true question is broader: "Is it proven that the next layer is preparing a feature representation for the next one?".

I don't know if it is mathematically proven, but you can easily see it yourself when making an image classifier with one hidden layer. It has some really good evidence for an assumption, at this point I would call it at least empirical evidence.

As an ML Research Scientist, I have never heard this interpretation. It's a very interesting thought that NN == kNN. It puts some of my lingering intuitions in clear wording. Thank you for this.

I think you are close to truth. This would explain why even the largest language models can't generalize beyound the training set.

At the same time, I disagree that analysis is out of proportion. It might be some clever averaging, but it does useful and interesting things. Take a look at Google: some clever averaging can get you a long way. It would be great to understand how it works and how can we go beyond it.

I do believe we need a paradigm shift, but it does not come out of nothing.

This so much. It's good its all in one place, but come on, it's a checklist of four whole screens.

I would prefer a minimal checklist instead: what measures give you the most (security) bang per buck (effort spent)?

My beef with all of these checklists: do you expect me to spend my whole managing my security?

Location: Instanbul, Turkey (recently left Russia, moving around) Remote: Yes Willing to relocate: Yes Technologies: Python, Pytorch, Python ML stack, Spark, Postgres, Docker, Git Email: iambtseytlin[at]gmail[dot]com Github: https://github.com/btseytlin Linkedin: https://www.linkedin.com/in/btseytlin/ Résumé/CV: https://www.dropbox.com/s/jkhvd4fc4ki5bt8/CV_BTseytlin_mg.pd...

Machine Learning Research Engineer.

Having left Russia, I am looking for a new position with more responsibility: machine learning engineer, research scientist or team lead. I have an engineering background, experience in ML research and product analytics. Together, this allows me to apply AI from understanding the problem to the data infrastructure to the implementation of models. I want to apply these skills to build something cool and grow with the company.

Honestly, no. Despite all the supposed advantages of async, I haven't found a use for it. When it comes to web apps a nginx + uwsgi + a sync framework is an unbeatable combination. The only times I have used async were when I made website scrapers and telegram bots for And if I was to scrape websites now, I would just use scrapy. My favorite telegram bot framework is synchronous too.

I think if you need to copy paste code from server to client you failed the separation of concerns. When is this useful, ever?

Location: Moscow, Russia

Remote: Yes

Willing to relocate: No

Technologies: Python, Django, Flask, Docker, Ansible, SQL, Git, Javascript, VueJS

Résumé/CV: https://www.dropbox.com/s/obqgmsyjsd5mdpc/resume_photo.pdf?d...

Email: b.tseytlin@lambda-it.ru

-----

I am Python Developer with experience of bringing a project from nothing to production. I am self motivated and self organized, with minimal supervision required to contribute. I am looking for a remote position where I could grow as an engineer, preferably in a team of great people.

Could you provide examples of games? The answer is very different for, say RTS or Shooter, MMO or match-based, etc.

It seems like a good idea, but won't work.

I can't imagine two horny teens like: M: Show me something hot baby F: Sure! But please install that app to take all the nececary security precautions before we proceed with our sexting...

This will gain traction among camgirls and other people that produce private porn (I just invented that term because I don't know how one would call porn distributed on an individual basis). So thats like, pervware?