lyapunova

Thanks! I love kagi's ethos!

I'm curious, is there a standard benchmark any one knows of that compares "practical usefulness" of LLMs instead of tries to make them take some kind of useless IQ test?

e.g. how useful is this LLM for 1) code debugging, 2) (accurate) fact retrieval, 3) daily task planning

It's settled, then. I'll move on to using a different video chat service...

They're a dime-a-dozen. Good job tanking your reputation and business, zoom!

This is really cute. I appreciate the sentiment and I think it applies even beyond SF.

FWIW, I have lived in SF for the last 8 years and I agree wholeheartedly.

This is potentially even better than the already excellent website!

I don't think I can use this if it's not open source... sorry.

The field moves too fast and the convenience is just not there otherwise.

edit: also the branding makes me think of MK-ultra which is probably something to avoid

The word "bullish" and fundamental science should stay as far away from each other as possible.

I hate to say it, but I agree. Nature has outlived its "legit" branding by leaning too hard into the "product" realm. Most scientists don't want their fundamental work to be sold as a product unless it is a precursor to commercialization of their work. At that point, it becomes advertising rather than science.

When I see Nature pubs, I tend to enjoy the aesthetics of the articles, but discount them a bit to account for the mainstream-ness.

Well...the way science is supposed to work is that everyone is an optimistic skeptic until something is known for certain. There is generally a correlation between higher PR effort and likelihood of falling short. This probably originates from incentive structures not aligning with being truthful and allowing work to speak for itself. So people have to drum up the excitement before it is warranted.

You only have to look as far as big social media to see proof of the damage that corporate motives can cause on the mental health of average internetizens.

This is hilarious. If only I had some kind of app to help me focus long enough to download this app.

I think it's clear at this point that hackernews types don't want to see these thin wrappers around ChatGPT anymore. It was a cool genre of hacker project in February, but I'm not so sure anymore. We are already dropping into the trough of disillusionment, and it hasn't been more than 6 months. which says a lot about the lack of inertia behind projects withe generic LLMs at their core...

The problem with the Boring Company is that it doesn't actually function in the "uninteresting" nature the eponym suggests. The fact that they continue ramming these underwhelming Vegas projects through without the necessary resources/planning to build something more useful than just running efficient ride-haling or buses speaks volumes.

It's like everything they build is a demo and they have to drum up PR around it. Hasn't this been continually happening for the better part of the last decade?

It would be wise to move away from the VC mentality for critical things like state/federal infrastructure.

I think efforts like this are awesome and I hope to see more in the future, but sometimes I wonder if some for-profit entity will swoop in and yoink all of the hard work these wholesome researchers have put out there for free.

I feel that this happens a lot with well thought out "PhD-level" software. Someone with dollar-sign eyes (who is capable of understanding it) takes the FLOSS and wraps it with a different UI, thereby profiteering off the backs of grad students...

I love this response. I'm now thinking of investing in a nice fountain pen and a soft pencil.

Please don't take this as a personal attack, but this list looks like you used the very first option presented to you and then never explored any other options. (maybe this is a woosh)

WARNING: There are a lot of recommendations here that are very out of sync with hacker(news) ethos.

Does that mean you should ignore them? No. But as I read through the preferences on a lot of options, they are seemingly haphazard takes. For example, I'm seeing people saying facebook messenger is the best text chat client and MS Teams is the best video chat? These are among the worst options out there (even setting aside ethos, they're bug-ridden!).

Just be careful, I think many recommendations in this particular thread are limited in perspective.

edit: i wrote this because hackernews tends to be a "ground truth" for a lot of people when they are trying to form their own opinions on geeky stuff

I have used it and if you take time to customize it, it surpasses all the other options simply because you can do whatever you please with your data.

Sometimes I edit my markdown files in other editors like Bear (or web interfaces). Sometimes I draw handwritten notes on my ipad and insert them as images.

It all works pretty well.

OneNote lock-in is crazy. I legitimately lost access to over a year of notetaking after my "complementary" OneDrive storage trial period ended despite having paid for a Surface that included OneNote. I shudder just thinking of going back to any note taking app that doesn't export my data in a portable format.

The toxic pit in this case is the transformation that occurred "ecologically" to the way contributors and agents on the web interact with eachother. This was strongly affected by the major centers of gravity in the 2010s like google and facebook. Unfortunately, their stewardship was focused around leveraging centralization towards profit rather than towards making the web a better place for users. And so here we are today.

FTR, I think youtube is one (among a few) exceptions to this overall trend from google.

I never realized Cheesecake Factory really went this deep. Suddenly, I have a craving for fried mac n' cheese balls...

Agreed on the big value of robotics in the near term, but Tesla won't be the company to do it. Tesla will make an "affordable" humanoid but will struggle to get useful and robust autonomy out the door for the entirety of its offering.

They thing to keep in mind about the "Boston Dynamics approach" is that in order to solve the hardest real world problems, you need to do one of two things: 1) Overshoot the capabilities of the system so that it is robust when actually deployed subject to the uncertainty of the real world (e.g. motivating athletic intelligence). 2.) Grossly constrain your environment so that uncertainty is not a factor. This is (and has been) happening in warehouses and factories for decades.

Google, unfortunately, does not "get" robotics. They are great at big ML training infra, but they need better robotics fundamentals that involve the real world.

Wow this has a lot of great distillation in it. I am always impressed with CMU output.

I never tire of this. Tim is a wonderful no nonsense person. I love these posts and I love that it stays up to date.

This is such a weird post. It feels like a advertisement and a humblebrag more than an attempt to impart any insightful commentary on life or tech.

I love this push to provide Arenas!

I am also a big fan of the human rating style of evaluation [0]

I think they are complementary approaches.

[0] https://colab.research.google.com/drive/1kK6xasHiav9nhiRUJjP...

I got stuck.

Try these: https://huggingface.co/blog/stackllama, https://huggingface.co/blog/trl-peft, https://huggingface.co/blog/hf-bitsandbytes-integration

I have been working in this space for quite a while and while I think the beginning of pytorch lightning meant well, it seems modern use-cases have outgrown it.

These days, when I see content from Lightning AI, I prepare for a contrived approach to doing something that fits within the ecosystem. I can't help but feel they are trying to induce "vendor lock-in" where there really isn't a business case for it...

Anyways, I tried to follow these steps and hit a dead-end. I have to say the content put out by huggingface is always way more straightforward and gets me to where I need to be when I want to spin up quickly.

I'm looking for something like OP's description. If you see this, I would really appreciate a couple pointers on where to look!

tornadosoup+hn@gmail.com