kubrickslair

Many people claim that the Artificial Analysis Index is highly contaminated - I have not personally looked into it.

Though, unlike the creators of benchmarks like Terminal Bench or ARC AGI, the Artificial Analysis Index team does not seem to have deep technical or ML backgrounds. They are ex-strategy consultants, McKinsey, et. al.

Why do you think OpenAI is the prettiest?

Claude often makes better looking interfaces and designs. And I think OpenAI has solved more open math/ stats/ CS problems.

Why do you think passion tax is not applicable in AI research? As that too is a very passion heavy industry at least recently.

I have found l-tyrosine also being quite helpful when taken occasionally.

They are probably looking for LGMs - Large Generative Models which encapsulate vision & multi-modal models.

Yes, this may be relevant:

In 1934, David Hilbert, by then a grand old man of German mathematics, was dining with Bernhard Rust, the Nazi minister of education. Rust asked, “How is mathematics at Göttingen, now that it is free from the Jewish influence?” Hilbert replied, “There is no mathematics in Göttingen anymore.”

Haha, I think it was covered at HN. But did not get as much attention as it should have.

Thanks! Most of the current work is not yet out as press or blogs - we generally do that post code release/ publications/ arXiv submissions. Will share them as they are made public.

Meanwhile, this is the blog of our earliest work from January that will be published in ICML24: https://blog.allenai.org/data-driven-discovery-with-large-ge...

Also if you are interested - happy to correspond more by email. Lot more to share privately.

The impact extends beyond mathematics. We have seen some pretty positive outcomes for LLMs as an assistant for scientists - with a lot of custom agent dev though - and not off-the-shelf models.

This is already being used by a few hundred scientists in a lab. We are aiming to extend to thousands next year with a focus on CS, climate & bio.

Phind-CodeLlama-34B-v2 seems to work well for our team.

Fireworks so elegantly captured the vector level control along with bitmap effects like Photoshop!

I have found OpenAI to be the most superior in complex prompts especially where written messages like “Get better, Mom” are expected in the images. The distant second would be ideogram.

I am using these tools to send custom personal messages to close friends and family.

I have only interacted briefly with Sam but I found him to be one of the smartest YC folks. But I will let a Paul Graham essay speak [1]:

Honestly, Sam is, along with Steve Jobs, the founder I refer to most when I'm advising startups. On questions of design, I ask "What would Steve do?" but on questions of strategy or ambition I ask "What would Sama do?"

What I learned from meeting Sama is that the doctrine of the elect applies to startups. It applies way less than most people think: startup investing does not consist of trying to pick winners the way you might in a horse race. But there are a few people with such force of will that they're going to get whatever they want.

[1] http://www.paulgraham.com/5founders.html

I think this is the relevant post with senior OpenAI folks reposting with heart:

Counted 52 OpenAI employees supporting @sama with more rolling in by the minute

https://twitter.com/FreddieRaynolds/status/17261100249827369...

Is not Lightcap retweeting the bigger deal as Greg already quit? It shows some serious level of support internally.

I think you misunderstood that he helped me overstay my student visa etc.

I was in a Ph.D. program at a top CS school and there are ways to transition your visa when building a startup. It was that I was not sure if the transition or the startup would work out - that startup did not - but years later another one did.

I would probably not have taken the plunge out of academia and not achieved much else had it not been for him. And I am deeply grateful for that.

I know Sam even less, but when I first moved to Valley a decade ago he went out of the way to help. I wanted try out a crazy startup idea on a student visa with limited connections in the Valley - he loved what I was doing and went above and beyond to help me out.

It forever tuned me in to the ethos of Silicon Valley. And I have tried paying back where I can.

I use it with Browse with Bing and it works quite well for keeping up with research on a diverse set of topics - CS, AI, economics, neuroscience etc. The citations help you verify quickly.

And this is besides the other stuff that search ones can’t do directly but it does quite well. It has definitely cut web searches for me.

They are also have a $1B+ annual run rate [1].

[1] https://www.reuters.com/technology/openai-track-generate-mor...

This is a good start, devxpy! What other features are you planning to build over it? Hit me up if you want to collaborate.

> We will be filtering our browsers with LLMs to remove junk and override Google, Twitter and FB ... Nobody got time to read garbage and ads.

Exactly, there is a paradox that is getting more extreme by the day - that social media (and the broader web) is a wellspring of knowledge, yet also a vortex of addiction, filter bubbles & lost productivity. I want one without the other.

This pushed me to start building open source at OpenLocus, contributions & feedback are welcome - details in other comment.

We started work on an open-source LLM-powered media filter at OpenLocus (https://github.com/openlocus), aiming for an alpha release by late 2023.

In the more medium term we are collaborating on improving information overload, filter bubbles & misinformation with labs at AllenAI, CMU, UPenn & Utah.

Contributions and feedback are welcome! Feel free to hit me up for an early access - email in the profile.

A lot of it is driven by underlying params - if you decrease style param a bunch in Midjourney, some of the high contrast & high saturation affinity goes away.

Yes, at least among my colleagues - both in the US and India offices.

Looks like 2019.

To quote pg, “The aim of web design is not to use all available screen space. It is legibility. Text is most legible with no more than 70 characters per line.”

Absolutely, the integration of logic into LLMs is far from a straightforward task.

Having been involved in both traditional machine learning and common sense AI in my grad school years, I've seen first-hand the limitations of a purely statistical approach. (Some of my past data augmentation work is being used to benchmark LLM reasoning.)

While most folks are too fixated by the 'quick wins' achieved by LLMs the trade-off is often a lack of non shallow reasoning. And I worry that many active researchers are glossing over these deeply rooted issues.

I find it fascinating that how useful generative AI and LLMs are despite these basic logical limitations. It really puts into perspective the gap between our current state of the art and more robust models we will see in the future.

Look forward to the day when such flaws are largely eliminated!

Thanks for sharing - I was curious too but didn’t delve in myself.

It does it pretty well for Python with code interpreter, especially for quant/ ML heavy code.