wwydmanski @wwydmanski@qoto.org

One free message per day to GPT-4 and Claude+
---
RT @poe_platform
Today we are launching Poe subscriptions, which will provide paying users with access to bots based on two powerful new language models: GPT-4 from OpenAI and Claude+ from Anthropic.
https://twitter.com/poe_platform/status/1635691869110947840

Gpt4 is cool and all, but remember that chatbots aren't limited to OpenAI
---
RT @AnthropicAI
After working for the past few moths with key partners like @NotionHQ, @Quora, and @DuckDuckGo, we’ve been able to carefully test out our systems in the wild. We are now opening up access to Claude, our AI assistant, to power businesses at scale.
https://twitter.com/AnthropicAI/status/1635679544521920512

On the field of ChatGPT's competitors it's not nearly as impressive as @AnthropicAI's Claude, but it's a step in the right direction
---
RT @_philschmid
The first open-source ChatGPT alternative got released! 🚀@togethercompute released a 20B chat-GPT model on Apache-2.0 🗣🆕

You can try it for free on Hugging Face. 😍

Demo: https://huggingface.co/spaces/togethercomputer/OpenChatKit
Model: https://huggingface.co/togethercomputer/GPT-NeoX…
https://twitter.com/_philschmid/status/1634492396171071488

Wszystkim laureatom #perłynauki przypominam o obowiązku używania hasztagów.
#nauka #naukaprzyszłościjużdziś

I've created an open source meeting summarizer and deployed it to Huggingface. It accepts texts of arbitrary length, going beyond the default token limit.

https://huggingface.co/spaces/wwydmanski/meeting-summarizer

Here are some notes taken from Hamlet:

Wygląda na to, że wśród #PerłyNauki pojawią się moje badania z XAI w biotechu.
W końcu wszyscy chcemy być zrozumiani, białka też na to zasługują

A healthy competition for ChatGPT is finally here! Claude by @AnthropicAI is available to use for everyone at https://poe.com, created by @Quora.

In a few years we're going to look at this problem the same way that we now look at SQL injections
---
RT @nearcyan
Indirect Prompt Injection: Turning Bing Chat into a Data Pirate

by modifying a website that bing chat reads alongside a user, the chat agent is able to have its goals modified by that site, unbeknownst to the user

demo: https://greshake.github.io/

arxiv: https://arxiv.org/abs/2302.12173
https://twitter.com/nearcyan/status/1630769218512904192

I think I've found the one weird trick for 10xing my research career.

There is an unpleasant lack of metagenomic datasets on @huggingface so I've decided to add some, starting from Feng Q. et al.'s "Gut microbiome development along the colorectal adenoma-carcinoma sequence " publication.

https://huggingface.co/datasets/wwydmanski/colorectal-carcinoma-microbiome-fengq
#Bioinformatics

🙃
---
RT @LOkruszek
Ponieważ w sobotę minął rok od złożenia naszego wniosku w Perłach Nauki mieliśmy dziś małą imprezę rocznicową w labie razem z ówczesną magistrantką a dziś już doktorantką Moniką Malon! Dziękujemy @MEIN_GOV_PL za tą przygodę, czekamy na drugą rocznicę!
https://twitter.com/LOkruszek/status/1628380662074941440

If you look at the first picture, you may ask yourself a question - why do we need to treat the projection from h to z separately?

It turns out that it's a side effect of contrastive loss! z vectors contain only the bare minimum info needed to recognize the similarity. (4/4)

SimCLR is an algorithm which depends on a common operation: augmenting the data.

However, in contrast to traditional methods, we don't use augmentation for pretending that we have more data. To the contrary! (2/4)

Do you remember how autoencoders can be used for pretraining on unlabeled data?

Now, imagine a method that can do it much better, much faster, without the pesky decoding part.

Welcome to a short thread on SimCLR 👇 (1/4)