pks016

I tried precise editing with a photo of horse with fence in front of it. One shot didn't work. Tried with multiple smaller regions. Still didn't work

Tried adjusting exposure; Didn't work as well.

I don't like it. Now, the TAs have high cognitive load. Earlier, you only need to know the material of the course. But now, you're forced to navigate AI output. And you got to keep with all new model developments.

One of my colleagues tried it for a few terms. Unfortunately, it didn't work well. Most students don't care about learning, they are here for grades :(

It would be a waste of time. LLMs have become better but not perfect. Personal experience; it's easier/faster to grade for me than have the LLMs grade and cross-check each one.

> Put a mixture of questions in that are impossible to answer

Students will complain to the admins and waste more of your time.

Please read the post. It's not written by Tao.

I'm low-key interested in reading the pre-print. I'll have to take some time this week to read thoroughly. I'm not from the field, so I can't judge the specifics.

On the surface, the preprint looks good. I glanced through the Methods and couldn't figure out if Claude wrote the preprint in Claude Science session or authors wrote it.

I was curious about the exact prompts they gave. If they share it, we could see how much domain specific knowledge was required and if we can replicate similar research with other models.

AI made a discovery. AI is doing it.

AI hacked a system. Humans did it.

We don't know for this particular case with certainty.

But, here many people are okay with the concept of stealing other ideas, plagiarism, unethical aspects of collaboration in research. At least for me and many people I know, this is no okay. This is the part for morality and ethics. I see that no everyone agrees with this.

The issue is, these AI ecosystems don't effect everyone equally. Not everyone knows how to judge AI model output accurately. There is already a large divide on AI usage in research. The fear comes from this. If we wait longer, it might be alright too late. I could be also wrong with these speculations but better to cautious.

That's the thing. It might be. We won't know the impact until the next few years. The things that are changing now: culture of research, aspects of collaboration and sharing of knowledge.

Academia is toxic for sure. But, also there are genuinely good people who work for greater good and whose training and teaching transcend across domains.

I mean it is. Most of the time in graduate school, you're getting trained and learning to do reasearch. People rarely produce great work during graduate school. Only towards the end of their degree or start of new position (post doc, independent researcher, or assistant professor), you start seeing good work. And for a professor, most of their research is done by graduate students.

Academia has it's own sets of big problems. Before this AI boom also, most people graduating never stayed in academia. The way graduate schools are structured, I would be happy if few people joined. That would also mean we lose good researchers in a long run.

Well, who watches the watcher?

Sort of, yeah. We are seeing more people commenting who have anti-AI sentiments or in the fence on the topic of AI usage.

On why some people are making emotionally charged claims, my guess: This affects the core belief of what is right or wrong, Impressions based on past doings of OpenAI, losing trust for OpenAI based on sequence of events.

I don't think we will get to see any clear evidence. I'm not even sure what would be the evidence. I would be surprised if OpenAI comes out clean if they have made a mistake. They move on to the next shiny thing.

I agree. I was wrong and I had a naive world view of morality and ethics in research and academia.

I now realize many people have different tolerance level for this.

That's the crux of it. In the current ecosystem of AI model usage, it's very hard to figure this out for research.

If you set a wrong foot and start trusting the model outputs, you can waste years searching for nothing.

How can someone realize this? By getting proper research training, failing, and learning from mistakes. For people beginning their research, it would be really hard to make decisions to move forward.

I never expected this many people (on this thread) arguing semantics and what not. I know that not everyone has morality and ethics, but I didn't realize it was this bad.

I'm afraid of the ripple effect of the agenda pushed by AI companies will have. In future and even now, they say AI has significantly progressed math and scientific research in general. There is truth to this, but the narrative has done more damage (so far) to the students, researchers, and the culture of knowledge transfer in academia. Many graduate students (I know) are having a crisis if any of their research worth it? If AI can (or will) do everything, what's the point of doing experiments and all? This will eventually deter a whole generation of curious minded students from research.

I guess, only time will whether this is for the good or bad. And how good AI models get without new data from research and experiments.

> Ph.D. graduate in every field

I have yet to see this in my field. Maybe like a PhD student who bullshits their way through. LLMs still can't make correct decisions, only as useful as the person who uses them. To me, LLMs are only useful for making some mundane tasks faster.

Would I read it? No

But, It would be helpful for beginners. I work with students and I know a lot (like a lot) of students (not math or stats major, other disciplines) are scared of statistics. They are always looking for resources to learn.

For Biology; the paper was retracted due to ethical issues.

So kind of similar to Open AI incident?

I had tried astro for my personal website when it was released. It was a mess for me. I couldn't keep with so many components. I keep breaking things, one way or the other. I might have to try again to see what things have changed.

Yes. Kind of. They showed that birds can do function/meaning of calls (e.g., contact call group) using the meaning not other cues (e.g. duration, acoustic similarity of calls etc.).

Generally, you test birds in a series of discrimination experiments. So for this example, the discrimination task would be distance calls vs contact calls. Let's say, birds learn to discriminate between 20 distance calls vs 20 contact calls. They can learn this quickly (reward and punishment). After birds learned this rule (say 75% accuracy), we ask to generalize this by adding more calls. Now, birds have to discriminate 30 distance calls vs 30 contact calls (10 new for each). We can check now, for which new calls they are more likely to correct and incorrect response.

In this research field, we often perform this types of experiments (not this exact one). So, this is not that surprising to some of us. Birds also prioritize duration a lot when telling apart calls.

Yeah, the paper is indeed difficult to read (even for me). I read it a while back. I'll try to explain.

So, there are few things to know before and see what we're trying to prove. main thing: Categorical perception: (in short) brain perceive similar sounding sounds (say A and B) in a continuous manner, and in the continuum of two sound, there would be a point where one sound (A) will switch and sound like the other sound (B).

In zebra finches, let's take an example: distance call and tet call. Both these calls are contact calls, birds use them to keep in contact with other birds. Now, we need to extract the acoustic similarity between these two calls, and test birds in a experiment where we ask them to discriminate between these two call types (and get a learning curve; say trials vs probability of correct choice). Then, we can fit a theoretical model of categorical perception to see how well they fit both in acoustic and perceptual dimensions.

This is for one function (contact call), we can do this within other call types and between other call types. At the end, we can see how much of birds' performance is explained by acoustic dissimilarity, function of acoustic calls etc.

This is what the authors have done. They first performed LDA based classification on call type to obtain distance between call types, misclassification etc. (acoustic dimension). Birds went through a series of discrimination experiments based on call types (similar to what wrote above). Then, they compared/matched acoustic and perceptual dimension.

I work in the field. The main takeaway is zebra finches (maybe other songbirds) can discriminate vocalizations based on function, even if they sound similar.

Generally, birds can tell apart categories of sounds e.g. vocalizations from different individual birds, male vs female, call vs song, conspecific vs heterospecific etc. The question is if birds can do it for specific function e.g. agonistic calls vs non-agonistic calls. Simple question but way harder to test because of associated contextual info. with vocalizations.

The paper is culmination of last decade of work (includes many of the past works) but this is the new result.

I get google scholar alerts according to authors.

Interesting. Btw I agree with some of the comments. The framing/language of the question is a bit weird.

I can relate. I remember my most used syntax, but I have to search once in a while.

I would be in minority. But, I don't like tidyverse ecosystem. I prefer data.table for most of my uses.