- Mar 2024
-
media.dltj.org media.dltj.org
-
Actually, ChatGPT is INCREDIBLY Useful (15 Surprising Examples) by ThioJoe on YouTube, 8-Feb-2024
- 0:00 - Intro
- 0:28 - An Important Point
- 1:26 - What If It's Wrong?
- 1:54 - Explain Command Line Parameters
- 2:36 - Ask What Command to Use
- 3:04 - Parse Unformatted Data
- 4:54 - Use As A Reverse Dictionary
- 6:16 - Finding Hard-To-Search Information
- 7:48 - Finding TV Show Episodes
- 8:20 - A Quick Note
- 8:37 - Multi-Language Translations
- 9:21 - Figuring Out the Correct Software Version
- 9:58 - Adding Code Comments
- 10:18 - Adding Debug Print Statements
- 10:42 - Calculate Subscription Break-Even
- 11:40 - Programmatic Data Processing
-
- Mar 2023
-
www.quantamagazine.org www.quantamagazine.org
-
It’s surprising because these models supposedly have one directive: to accept a string of text as input and predict what comes next, over and over, based purely on statistics. Computer scientists anticipated that scaling up would boost performance on known tasks, but they didn’t expect the models to suddenly handle so many new, unpredictable ones.
Unexpected emergent abilities from large LLMs
Larger models can complete tasks that smaller models can't. An increase in complexity can also increase bias and inaccuracies. Researcher Jason Wei has cataloged 137 emergent abilities of large language models.
-
-
www.federalregister.gov www.federalregister.gov
-
For example, when an AI technology receives solely a prompt [27] from a human and produces complex written, visual, or musical works in response, the “traditional elements of authorship” are determined and executed by the technology—not the human user.
LLMs meet Copyright guidance
See comparison later in the paragraph to "commissioned artist" and the prompt "write a poem about copyright law in the style of William Shakespeare"
-
-
www.technologyreview.com www.technologyreview.com
- Feb 2023
-
nymag.com nymag.com
-
More interesting or alarming or hilarious, depending on the interlocutor, is its propensity to challenge or even chastise its users, and to answer, in often emotional language, questions about itself.
Examples of Bing/ChatGPT/Sydney gaslighting users
- Being very emphatic about the current year being 2022 instead of 2023
- How Sydney spied on its developers
- How Sydney expressed devotion to the user and expressed a desire to break up a marriage
-
-
arxiv.org arxiv.org
-
Shanahan, Murray. "Talking About Large Language Models." arXiv, (2022). https://doi.org/10.48550/arXiv.2212.03551.
Found via Simon Wilson.
Abstract
Thanks to rapid progress in artificial intelligence, we have entered an era when technology and philosophy intersect in interesting ways. Sitting squarely at the centre of this intersection are large language models (LLMs). The more adept LLMs become at mimicking human language, the more vulnerable we become to anthropomorphism, to seeing the systems in which they are embedded as more human-like than they really are. This trend is amplified by the natural tendency to use philosophically loaded terms, such as "knows", "believes", and "thinks", when describing these systems. To mitigate this trend, this paper advocates the practice of repeatedly stepping back to remind ourselves of how LLMs, and the systems of which they form a part, actually work. The hope is that increased scientific precision will encourage more philosophical nuance in the discourse around artificial intelligence, both within the field and in the public sphere.
-
LLMs are generative math-ematical models of the statistical distributionof tokens in the vast public corpus of human-generated text, where the tokens in question in-clude words, parts of words, or individual char-acters including punctuation marks. They aregenerative because we can sample from them,which means we can ask them questions. Butthe questions are of the following very specifickind. “Here’s a fragment of text. Tell me howthis fragment might go on. According to yourmodel of the statistics of human language, whatwords are likely to come next?”
LLM definition
Tags
Annotators
URL
-
-
arstechnica.com arstechnica.com
-
The breakthroughs are all underpinned by a new class of AI models that are more flexible and powerful than anything that has come before. Because they were first used for language tasks like answering questions and writing essays, they’re often known as large language models (LLMs). OpenAI’s GPT3, Google’s BERT, and so on are all LLMs. But these models are extremely flexible and adaptable. The same mathematical structures have been so useful in computer vision, biology, and more that some researchers have taken to calling them "foundation models" to better articulate their role in modern AI.
Foundation Models in AI
Large language models, more generally, are “foundation models”. They got the large-language name because that is where they were first applied.
-