jerrygenser

Embedding chunks and finding chunks based on similarity is definitely in use now. But if you can increase context size cheaply then the model can figure out what's relevant.

It seems to be running a base model by developed by Anthropic

Compression in order for the input data to fit context, maybe? For example if context is 4092 and input size is 6000, figuring out the appropriate way to run an operation on all 6000 where context over all 6000 might be relevant to the operation.

There's another interesting article on front page about oauth that's actually almost exactly the same. A very long article about oauth implementation. And sadly I knew the add was coming the whole time and there it was as the last paragraph. It seems that unfortunately or fortunately some of the best really informative intermediate depth blog posts (read: not medium surface level stuff) tends to be an advert by a company offering a very technical product.

Will this prevent the merger from happening in other markets too?

Why is this tone deaf? A lot of aaa gaming today could be considered cloud gaming -- although I'm not sure of the exact definition.

Consider dota 2 -- it's a service entirely on the cloud continuously updated.

The list goes on with in terms of aaa games offered by either company where the main offering is actually multiplayer.

Launch HN is reserved for YC companies only

Even as a startup founder, the best practice is to sign over your IP rights to the company since investors want to see this. So I can say even with a large equity stake in a new venture that's not w-2 work, you still sign over IP rights to an org.

Chatgpt running on bing is ad tech, no?

If I have a python monolith and I have a specific set of heavy dependency like a chromium browser, I don't want my webserver to build with it only async task. Is there a straightforward way to build a separate environment in python that just had the extra dependencies while still having benefits described in the article?

Yes, dramatiq is quite good, but I ditched it for something even more lightweight -- pubsub subscribers via push notification so workers are web workers responding to http requests that might take up to 10 minutes

It's not just about the market speaking. There's path dependence between it being easy to use for scripting and scientific things as well as bumpy which made it natural to supplant things like Matlab in teaching. And then basic knowledge translates to industry jobs.

It's not popular purely based on its merits as a language

Agreed that there are better approaches than LLm to solve this problem There are deep learning and sequence models that are focused on this task using extracted structured info. Wouldn't you be better off using llm or any extraction + linking model to identify medical concepts and then a model that only understands medical concepts in order to do predict the next medical concepts.

When you say prediction my discharge disposition do you mean extracting from the document a statement of what the discharge disposition was? Or do you mean based on m extracted clinical characteristics, predicting what it should be?

For extracting what it was if it's in the document, you can use the indexing approach described in the article. For predicting what it should be, you'd probably be better off using llm as an extraction mechanism and then using structured data models to classify discharge disposition

The ones I know that did not post on social media did not want to be fired.

Consider the selection bias of people posting on social media. Also consider people would want to be positive on social media to keep their future employment prospects.

> Getting significant business traction was difficult

Is the problem a painful problem that you had a solution people would pay for and now they are using gpt-4 to solve it?

Have you heard directly from existing or prospective customers that gpt-4 was a replacement for your solution?

Seems like it would lock in a user to cohere proprietary model in order to embed new search terms to the same space?

Would love to see similar but embedded with a more open representation model or even sent2vec

GPT is an acronym describing a particular strategy for training a language model. Many models can be trained in this fashion. It's very possible that models not trained by the original inventors of the method will be better than the current models operated by the inventors.

Are all pretrained LM models like BERT considered foundational models?

I tried out htmx + django a little but got stuck doing something more than trivial. Do you have an example of a starter project that has intermediate concepts demonstrated? Maybe this can help me get over the hump.

There's been a brain drain from alphabet, why do you bet despite this?

Probably lower since the models have actually been lobotomized to a degree for safety purposes

Link to pydantic v2 plan on the blog post is broken.

Great job team can't wait to use it!

No, the title is linked. The title refers to Pythia by EleutherAI for evaluating LLMs.

You linked something that is a framework for multimodal models by facebook.

Llama may not be licensed for people to share since you need to apply to get one from Facebook for non commercial use. I think it's more of a license issue

Hopefully similar work can be done with LoRA so the fine-tuning is not as expensive

> I still think it should be controlled medically and not as available as lotto tickets

You could make the same argument for alcohol... it can destroy lives and yet is still legal.

this was just on the front page now I can't find it. was it removed?

Some of these large models are able to do zero shot learning and perform tasks they weren't explicitly trained on since the training objective is very general.

Being able to perform more advanced types of zero shot learning tasks would be comparable and further the accuracy on those tasks can be evaluated

microsoft apple google

From earlier threads on hacker news it motivated me to read snow crash. However the book was very difficult to finish. I literally used spark notes to finish it, it got so long in the tooth and boring towards the end.

Is it just me? I like books like the three body problem that are entertaining but snow crash was like a much much worse version of the da Vinci code with some VR thrown in.