I think I broke it. I told it my company was called Waldo’s Wands and I was in the transport industry - I got a transport intro, a list of magic wand services, and a cannabis picture!
Seriously though, the Unsplash search is pretty rubbish. Also, I’m only returning the first image from search (I only have 50/h) , could probably improve the search on my side by checking metadata or something. At least the plant has the right number of leaves, right?
Actually now that I think of it - I think I made the last image do a search using only the company name (to avoid same images as others above) so you probably got “Waldo’s Weed” there!
I follow AI on YT but focus on cognitive agent architectures and txt models so I don’t have any sources dedicated to robotic AI. Can you recommend a channel that focuses on progressions in that area ?
I dont see much info on youtube, but that’s the last use I have of twitter, there are several accounts posting about these here. I mostly follow LLMs and robotics subject. You can take a look at my followers list: twitter.com/ktp_programming/following
Or do as I did, follow someone like @DrJimFan and follow a new person every time he retweets something you find interesting.
Maybe I understand the following :
(my apologies if this is grossly simplified and doesn’t help)
1- Better neural network need to contain more (stacked) layers.
2- input layer at one end of the stack is exposed to messy informations from the real world.
3- at the other end the output layer provide results from the network.
4- the first step for making this work is the training of the network during which training, learning is done.
5- instabilities and stagnation in some layers often occur when learning does not occur in an optimal way. This problem increases exponentially with the number of layers.
6- here learning is done all at once to all the layers. Something called rotation which I don’t understand, is used to stabilize and optimize the learning.
I feel this is very different from human learning where it happens in stages : we first learn words, then try to assemble them to form simple sentences, then evolve to make sense of more complex notions and so on. I wish this approach could apply also in artificial intelligence development.
The human brain isn't a blank slate when it comes into existence. There are already structures that are designed to do certain things. These structures come "pre trained" and a lot of the learning humans do is more akin to the fine tuning that we do for foundation models.
While designing a similar classifier, I've considered the idea of giving it the whole thread as "context" of sorts.
Not just the parent comment, the whole thread up to original post.
I've abandoned the idea.
A comment must stand on it's own, and it would put limits on results, the way I was planning to do it.
I might be very wrong, your insight into this would be very helpful.
My original idea was to go recursively trough the thread and test each comment individually.
Then I would influence the actual comment results with the combined results of it's parents.
No context during inference, just one comment at a time.
For example consider thread OP->C1->C2->C3.
My current model takes milliseconds per test with little resources used.
It would be ok up to very large threads but would contain a limit to save on answer time.
I want to determine if Comment 3 is toxic in the context of C2, C1, and OP.
Test C3, test C2, test C1, test OP. Save results.
My current model gives answer in several fields ("toxic", "severe toxic", "obscene", "threat", "insult", and "identity hate")
The idea was to then combine the results of each into a final result for C3.
How to combine? Haven't figure it out but it would be results manipulation instead of inference/context, etc.
Edit: Is there any way you can point me at examples difficult to classify? It would be a nice real world test to my stuff.
Current iteration of model is very new and has not been tested in the wild.
(“toxic”, “severe toxic”, “obscene”, “threat”, “insult”, and “identity hate”)
You aren’t the author of Detoxify are you by any chance? It uses the same classifications. I was originally using it but switched to my own model as I really only needed binary classification and felt a new dataset that better suited Lemmy was needed anyway. I have 2 outputs (toxic and not-toxic).
I’ve been building my own dataset as the existing ones on Huggingface seemed to contain a lot of content you might see on Twitter, and were a poor match for Lemmy. Having said that, I’ve generally avoided putting that sort of content into the dataset as I figured if I can’t easily decide if it’s toxic, then how could a model.
Is there any way you can point me at examples difficult to classify? It would be a nice real world test to my stuff. Current iteration of model is very new and has not been tested in the wild.
Here’s a few where I’ve had to go back to the parent comment or post to try and work out if it was toxic or not:
Do your research on the case and the answers will be very obvious. (What comment prompted this? Is it a trolling comment or a reasonable response to a trolling comment)
Because you’re a fascist. The fact that they disagree with you is secondary (Is the commenter calling another commenter a fascist, or continuing a discussion?)
Me tard, you tard, retard nation! (Is this a quote from a movie or TV show or an insult to another commenter? Not sure.)
Fuck you shoresy! (pretty sure this is a quote from a tv show)
A comment must stand on it’s own, and it would put limits on results, the way I was planning to do it. I might be very wrong, your insight into this would be very helpful.
I originally thought that, and I’m actively tuning my model to try and get the best results on the comment alone, but I don’t think I’ll ever get better than about 80% accuracy. I’ve come to the conclusion that those cases in the grey zone where toxic ~= not-toxic can only be resolved by looking upstream.
Oof, pop-culture references are hard and I had not considered that at all.
Thanks for the examples, I'll have a think on how to deal with those.
My only insight is one you already had.
Test at least the comment before, and then use the output to dampen or amplify the final result.
Sorry for being no help at all.
--
My project is very basic but I'll post it here for any insight you might get out of it.
I teach Python in a variety of settings and this is part of a class.
Model is a very basic TensorFlow implementation intended for teaching TF basics. https://github.com/vluz/ToxTest
Some of the helper scripts are very wonky, need fixing before I present this in class.
I’m working on rewriting it with all the multi-processor code removed so you can better understand the algorithm, but I’d love to know what you are trying to better understand.
Now with this effectively replicating the same end result as the Stanford study, I’ll be really interested to see what invariably turns out to fix it in the next 12 months.
I don’t think so. But GPT-4 API was on waitlist (I was stuck on it for a long time) and every paying user now got access to it. It is still billed per token and GPT-4 tokens are more expensive (30x) than GPT-3.5 turbo
I’ve been slowly chipping away at this series for months. It’s beyond excellent, but fully digesting everything he’s saying is taking me a while.
I’m also trying to apply each lesson to a side project. I’m failing at most of these projects but it’s helping me cement the lessons and intuitively understand what approaches work, what not, and why.
machinelearning
Hot
This magazine is from a federated server and may be incomplete. Browse more on the original instance.