Galactica: A Large Language Model for Science Research

Quite Useful idea for sharing orgaizedscience data with with Large Language Methods

 https://www.youtube.com/watch?v=ZTs_mXwMCs8&t=1418s

https://arxiv.org/abs/2211.09085

Galactica: A Large Language Model for Science

Computer Science > Computation and Language

[Submitted on 16 Nov 2022]

Ross Taylor, Marcin Kardas, Guillem Cucurull, Thomas Scialom, Anthony Hartshorn, Elvis Saravia, Andrew Poulton, Viktor Kerkez, Robert Stojnic

Information overload is a major obstacle to scientific progress. The explosive growth in scientific literature and data has made it ever harder to discover useful insights in a large mass of information. Today scientific knowledge is accessed through search engines, but they are unable to organize scientific knowledge alone. In this paper we introduce Galactica: a large language model that can store, combine and reason about scientific knowledge. We train on a large scientific corpus of papers, reference material, knowledge bases and many other sources. We outperform existing models on a range of scientific tasks. On technical knowledge probes such as LaTeX equations, Galactica outperforms the latest GPT-3 by 68.2% versus 49.0%. Galactica also performs well on reasoning, outperforming Chinchilla on mathematical MMLU by 41.3% to 35.7%, and PaLM 540B on MATH with a score of 20.4% versus 8.8%. It also sets a new state-of-the-art on downstream tasks such as PubMedQA and MedMCQA dev of 77.6% and 52.9%. And despite not being trained on a general corpus, Galactica outperforms BLOOM and OPT-175B on BIG-bench. We believe these results demonstrate the potential for language models as a new interface for science. We open source the model for the benefit of the scientific community.

Subjects: Computation and Language (cs.CL); Machine Learning (stat.ML)

Cite as: arXiv:2211.09085 [cs.CL]

  (or arXiv:2211.09085v1 [cs.CL] for this version)

https://doi.org/10.48550/arXiv.2211.09085

Focus to learn more

Submission history

From: Robert Stojnic [view email]

[v1] Wed, 16 Nov 2022 18:06:33 UTC (10,715 KB)

Meta, Yet Another Contender in the AI Rush?

Is Meta poised for this too?     Must this be VR oriented? 

Meta, Long an A.I. Leader, Tries Not to Be Left Out of the Boom  By The New York Times, February 8, 2023

Meta has been left out now that Silicon Valley is gripped with excitement by generative A.I., the name for technologies that generate text, images and other media on their own.

Credit: Michael Furler

Two weeks before a chatbot called ChatGPT appeared on the internet in November and wowed the world, Meta, the owner of Facebook, WhatsApp and Instagram, unveiled a chatbot of its own.

Called Galactica, it was designed for scientific research. It could instantly write its own articles, solve math problems, generate computer code and annotate images.

Like ChatGPT, Galactica also played fast and loose with facts, making up mathematical proofs, misstating historical dates and spinning tall tales. One user coaxed the chatbot into talking about the history of bears in space. When asked who runs Silicon Valley, Galactica replied, "Steve Jobs."

But unlike OpenAI, the tiny San Francisco lab that made ChatGPT, Meta encountered an avalanche of complaints about Galactica's mishaps. After just three days, the company, which has faced scrutiny for spreading misinformation and hate speech through its social networking apps, removed Galactica from the internet.

From The New York Times    

Three Key Takeaways from Meta's Galactica AI

Galactica AI,  general idea is good, navigating complex forms of data is useful.  One of our early AI successes did exactly this for key internal knowledge.  Make it easy to easy to use, and importantly make it easy to validate and update in changing contexts.

Three Takeaways from Meta's Galactica AI

By TechTalks, November 23, 2022

We should look at Galactica's initial failure as another scientific experiment...and, every failed experiment brings us one step closer to success.

Meta's most recent large language model (LLM) release, Galactica, was supposed to help scientists navigate tons of published scientific information. Its developers presented it as being able to find citations, summarize academic literature, solve math problems, and perform other tasks that help scientists in research and writing papers.

However, three days after the release, Meta had to shut down the online demo following a deluge of criticism by scientists and tech media about the model's incorrect and biased output. While Galactica was obviously not a success, its short history provides us with some useful lessons about LLMs and the future of AI research.

From TechTalks   View Full Article