Well, fuck. The GPT disinformation age is now.

I googled "OS for 4gb ram" and the first hit, which also was used by google to populate its snippet is an answer from quora which is very obviously created with #chatgpt (I recognised the non-committal non-answer right away, but it can also be detected by a popular GPT detector).

The user has 98 answers and, you've guessed it, they are all created with GPT.
PLOT TWIST: The questions where also created with GPT!

https://www.quora.com/profile/Heri-Mulyo-Cahyo

#AI

This is quite the exposé!

I knew that OpenAI has been using crowd workers for at least the last two years to build instruct-tuning datasets for #ChatGPT. I didn’t know they were also outsourcing violence, hate speech, and sexual abuse labeling to Kenyans for $2/hour.

The work is apparently so traumatic that the outsourcing company, Sama, has cut ties with OpenAI.

https://time.com/6247678/openai-chatgpt-kenya-workers/

I frequently note the (sometimes competing) methods of #AI (e.g., #NLP):

(1) SYMBOLIST AI (e.g., "Code all the things"). Examples: #SALI, #KnowledgeGraphs

(2) CONNECTIONIST AI (e.g., "Feed all the things into the #ML model, let the machine figure it out.") Examples: #BERT, #ChatGPT

This is a good Stephen Wolfram piece on the benefits of combining Symbolist + Connectionist applications (e.g., Wolfram|Alpha + ChatGPT). 1+1=3.

https://writings.stephenwolfram.com/2023/01/wolframalpha-as-the-way-to-bring-computational-knowledge-superpowers-to-chatgpt/

Are you interested in doing a PhD in neuro-symbolic AI, and doing science for developing new explainable AI technologies for making financial services more fair, transparent and capable?

We have a PhD position available. The selected candidate will be part of a strong team at #illc_amsterdam #UvA_Amsterdam and the upcoming #ai4fintech hub in Amsterdam.

Deadline: Feb 15. See the link below for more information. #xai #causality #logic #fintech #AI4SocialGood

https://vacatures.uva.nl/UvA/job/PhD-Position-in-Explainable-AI-in-Finance/761750702/

New pre-print!

Plug and play generation works for images and text…what about proteins?

We engineer proteins by combining your favorite unsupervised and supervised protein sequence models (even protein language models!) in a fast *gradient-based* discrete MCMC sampler.

🧵

If you installed PyTorch-nightly on Linux between Dec. 25 and Dec. 30, uninstall it and torchtriton immediately and use the latest nightly binaries.

Read the security advisory here: https://pytorch.org/blog/compromised-nightly-dependency/

We are organizing an exciting workshop on the intersection of deep generative modelling and neurosymbolic learning at #ICLR2023 💎!

The call for papers is out now: https://nesygems.github.io/callforpapers/
Submission deadline: 3rd of February 2023

#neurosymbolic #generativemodels