#DanWood - #ChatGPT on The #Commodore64: Can a 1982 #Computer Run Modern #AI?

https://www.youtube.com/watch?v=-OA28r8Up5U&ab_channel=DanWood

#C64 #Commodore #CBM #Amiga #CommodoreAmiga #ArtificialIntelligence #LanguageModel #LanguageModels #LargeLanguageModel #LargeLanguageModels #LLM #Computing #8Bit #16Bit #Retro #RetroComputing

2/6: 🤖 DoppeLLaMA is a project developed by Urist-Mc-Urist. It focuses on fine-tuning the powerful GPT-3.5 language model, created by OpenAI. By leveraging this model, DoppeLLaMA aims to generate high-quality summaries of scientific papers and articles. #AI #LanguageModel

📝 "The Definitive Guide to BERT Models"

👤 Tuana Celik (@tuana)

🔗 https://medium.com/deepset-ai/the-definitive-guide-to-bert-models-160b40c41389?source=rss-d1b6671d742f------2

#pyladies #python #machinelearning #haystack #nlp #languagemodel #artificialintelligence

#AI is going atomar these days. So impressive!

#Google #LanguageModel #Robot

https://ai.googleblog.com/2023/03/palm-e-embodied-multimodal-language.html

RT @LukasPilka@twitter.activitypub.actor

Thought of the Day 🤯 What if there was a generative language model trained on digitized texts published before 1900? Would that create a time machine to the 19th century? #languagemodel #timemachine #timetraveler

🐦🔗: https://twitter.com/LukasPilka/status/1628151532259536904

#LanguageModel #Search #AI

OpenAI has a language model to answer questions. Microsoft is planning an AI based chatbot in its seach engine. Why do neither Google nor Microsoft implement a search that is really useful? —

When can we search the web using Regular Expressions?

#ChatGPT に彼女の #languageModel について聞いてみた．

Oh no, we're all gonna get spammed now!

#ChatGPT #ai #OpenAI #LanguageModel #email #spam

Trying to explain how ChatGPT works in very simple terms
https://www.linkedin.com/pulse/trying-explain-how-gpt-works-very-simple-terms-jose-paez

#ChatGPT #gpt #ai #languagemodel

#Google is getting its #ChatGPT competitor ready for primetime. The company said its new #conversationalAI service, called Bard, would be opening up to trusted testers. Bard will be powered by #LaMDA, a large #LanguageModel that stirred controversy when a Google software engineer publicly asserted that the #AI was “sentient”—a claim that has since come under scrutiny. https://adage.com/article/marketing-news-strategy/google-releases-chatgpt-rival-ai-bard-early-testers/2468816 #GenerativeAI

#ChatGPT with #RobMiles - #Computerphile

https://www.youtube.com/watch?v=viJt_DXTfwA&ab_channel=Computerphile

#ComputerScience #LanguageModel #LanguageModels #LanguageProcessing #TextProcessing #ComputerLearning #AI #ArtificialIntelligence #GPT #GPT3 #AILearning #AITraining #OpenAI

Comparison between three language models:
#ChatGPT
you.com chat
#perplexityai

First task:
I ran a simple packet capture with Wireshark on my computer. And I see lots of red and black colored packets. What do they mean, and should I be worried?

Results (+pictures):
GPT gave the most detailed and clear answer compared to the two others.
Perplexity & UChat were both lazy to write more than 2 sentences.
Not to mention, Uchat gave false information (https://blog.knoldus.com/how-to-read-color-coding-in-wireshark/)
#ai #languagemodel

what other language models do you know of aside from #ChatGPT?

I'm experimenting and comparing and would like to see more.
For now I know of:

you.com chat (serves as a search engine with its own language model to answer your questions aswell)

perplexity.ai

they both have access to browse the web and should give you answers of things currently happening and stuff.
But are far from compareable to GPT.

So if you know of any others please send!

#AI #openai #languagemodel

#BERT

Do you know of any research effort (or paper) that fine-tunes BERT for classifying text on more than 1000 classes?

The highest I have seen so far is around 140 classes

#languageModel #transformers #NLP #multiclass

I "interviewed" ChatGPT to find out more about its training, what it has learned, pattern recognition and how it works to give us those intriguing answers to our prompts...

https://www.sciencebase.com/science-blog/chatgpt-explains-chatgpt.html

#AI #ChatGPT #LanguageModel #MachineLearning #Prompts

I've really only just started to experiment with ChatGPT, and having found some expert prompts have tested how well it could write an article in *my* writing style...

I explain the process in my latest post...or do I?

Judge for yourself:

https://www.sciencebase.com/science-blog/setting-the-tone-artificially.html

#ChatGPT #LanguageModel #MachineLearning #Algo #Algorithm #WritingTool #OpenAI

req> "Evaṁ me [MASK]—"
pred> sutaṁ [58%]

Small victory 😄

Got a first model out, maxing out at ~30% accuracy. A second iteration tweaking the learning rate seems on a better trajectory.

For this #LanguageModel I used the entire #Sutta Nipata both as training and evaluation sets, so there's a chance the model is memorizing more than learning. I'll test it out on the Vinaya just to see how it generalizes.

It is training now.

I am starting with a lightweight #albert masked #LanguageModel to see where we are and iterate.

I failed so many small scale attempts just to write the right tooling, debug the gross mistakes on inputs and params.

I can see some nice training curve. Maybe I need to build an eval dataset, with some "stock phrases" of #buddhism, like "Evam me suttam" or "Yam kiñci samudaya dhammam...".

Not sure any of this will land anything, but I am having fun.

#NLP #DigitalHumanities

Positive perspective:

"I just discovered ChatGPT and it is seriously incredible! It's a chatbot that uses a cutting-edge language model to generate human-like responses to your questions and conversations. I've been having some amazing and thought-provoking conversations with it. If you're into AI and language technology, you have to check it out! #ChatGPT #AI #LanguageModel"

2/3

Ich habe durch #chatgpt ein Employer Branding einer öffentlichen #Verwaltung schreiben lassen: Das Ergebnis kann sich sehen lassen. Ein Thread (1/2): Als öffentliche Verwaltung bieten wir ein breites Spektrum an spannenden und vielfältigen Karrieremöglichkeiten für engagierte und motivierte Mitarbeiterinnen und Mitarbeiter. Wir setzen auf Qualität und Professionalität und fördern die Weiterbildung und Entwicklung unserer Beschäftigten. #digitalisierung #ai #chatbot #languagemodel