# Generative Artificial Intelligence, Large Language Models, and Image Synthesis
**URL:** https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383
**Category:** The Happening World
**Tags:** chatgpt, image-synthesis, large-language-model, artificial-intelligence, generative-transformer
**Created:** [2 December 2022 16:12 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383 "2022-12-02T16:12:04Z")
**Posts on this page:** 20
**Page:** 33
### Author:  [@johnwalker](https://scanalyst.fourmilab.ch/u/johnwalker)
#### Post date: [13 December 2023 22:38 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/642 "2023-12-13T22:38:41Z")
> **[Bash One-Liners for LLMs](https://justine.lol/oneliners/)**
>
> Tutorial on how llamafile makes LLMs shell scriptable.
---
### Author:  [@eggspurt](https://scanalyst.fourmilab.ch/u/eggspurt)
#### Post date: [15 December 2023 14:32 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/643 "2023-12-15T14:32:58Z")
Interesting to see how to work around LLMs, and how tools like ChatGPT are actually built:
[https://platform.openai.com/docs/guides/prompt-engineering/](https://platform.openai.com/docs/guides/prompt-engineering/)
---
### Author:  [@eggspurt](https://scanalyst.fourmilab.ch/u/eggspurt)
#### Post date: [15 December 2023 14:36 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/644 "2023-12-15T14:36:33Z")
The new SEO:
[](https://twitter.com/delcomplex/status/1735344373037187488)
---
### Author:  [@jabowery](https://scanalyst.fourmilab.ch/u/jabowery)
#### Post date: [15 December 2023 18:55 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/645 "2023-12-15T18:55:43Z")
These attacks expose the lack of what might be called “critical thinking” in foundation model learning algorithms. The first step toward critical thinking is recognizing that any datum has a latent provenance chain. Phenomenologists sometimes call this bracketing or putting the datum in quotes. This was my motivation for suggesting Wikipedia as the corpus for the Hutter Prize but capital is so misallocated that 17 years later we end up with this nonsense threatening all of humanity.
---
### Author:  [@CTLaw](https://scanalyst.fourmilab.ch/u/CTLaw)
#### Post date: [17 December 2023 14:48 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/646 "2023-12-17T14:48:26Z")
> **[Report: Microsoft's AI Chatbot 'Hallucinates' Election Misinformation](https://www.breitbart.com/politics/2023/12/16/report-microsofts-ai-chatbot-hallucinates-election-misinformation/)**
>
> A recent investigation has revealed troubling issues with Microsoft's AI chatbot, Copilot, disseminating misinformation and conspiracy theories related to elections. In one case, researchers asked the AI chatbot about corruption allegations against a...
---
### Author:  [@owenwengerd](https://scanalyst.fourmilab.ch/u/owenwengerd)
#### Post date: [18 December 2023 03:26 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/647 "2023-12-18T03:26:35Z")
Chatbot hallucination is a feature, not a bug. Andrej Karpathy tweet: [the hallucination problem](https://twitter.com/karpathy/status/1733299213503787018)
---
### Author:  [@johnwalker](https://scanalyst.fourmilab.ch/u/johnwalker)
#### Post date: [18 December 2023 12:11 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/648 "2023-12-18T12:11:43Z")
[](https://twitter.com/ChrisJBakke/status/1736533308849443121)
**Earlier**
---
### Author:  [@eggspurt](https://scanalyst.fourmilab.ch/u/eggspurt)
#### Post date: [18 December 2023 15:09 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/649 "2023-12-18T15:09:28Z")
> **[Google DeepMind used a large language model to solve an unsolved math...](https://www.technologyreview.com/2023/12/14/1085318/google-deepmind-large-language-model-solve-unsolvable-math-problem-cap-set/)**
>
> They had to throw away most of what it produced but there was gold among the garbage.
> FunSearch (so called because it searches for mathematical functions, not because it’s fun) continues a streak of discoveries in fundamental math and computer science that DeepMind has made using AI. First[AlphaTensor](https://www.technologyreview.com/2022/10/05/1060717/deepmind-uses-its-game-playing-ai-to-best-a-50-year-old-record-in-computer-science/) found a way to speed up a calculation at the heart of many different kinds of code, beating a 50-year record. Then[AlphaDev](https://www.technologyreview.com/2023/06/07/1074184/google-deepmind-game-ai-alphadev-algorithm-code-faster/) found ways to make key algorithms used trillions of times a day run faster.
>
> FunSearch combines a large language model called Codey, a version of Google’s PaLM 2 that is[fine-tuned on computer code](https://www.technologyreview.com/2023/12/06/1084457/ai-assistants-copilot-changing-code-software-development-github-openai/), with other systems that reject incorrect or nonsensical answers and plug good ones back in.
>
> A second algorithm then checks and scores what Codey comes up with. The best suggestions—even if not yet correct—are saved and given back to Codey, which tries to complete the program again.
---
### Author:  [@CTLaw](https://scanalyst.fourmilab.ch/u/CTLaw)
#### Post date: [18 December 2023 22:58 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/650 "2023-12-18T22:58:43Z")
> **[OpenAI suspends ByteDance’s account after it allegedly used GPT to build...](https://nypost.com/2023/12/18/business/openai-suspends-bytedances-account-after-it-allegedly-used-gpt-to-build-rival-ai-product-report/)**
>
> ByteDance has reportedly relied on OpenAI’s application programming interface, or API, “during nearly every phase of development” of its AI product.
---
### Author:  [@eggspurt](https://scanalyst.fourmilab.ch/u/eggspurt)
#### Post date: [19 December 2023 20:56 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/651 "2023-12-19T20:56:56Z")
[](https://huggingface.co/spaces/lmsys/chatbot-arena-leaderboard)
---
### Author:  [@johnwalker](https://scanalyst.fourmilab.ch/u/johnwalker)
#### Post date: [21 December 2023 14:17 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/652 "2023-12-21T14:17:39Z")
[](https://twitter.com/EricTopol/status/1737505177052348545)
The paper, published in _Nature_ on 2023-12-20, is “[Discovery of a structural class of antibiotics with explainable deep learning](https://www.nature.com/articles/s41586-023-06887-8)”. Here is the abstract:
> The discovery of novel structural classes of antibiotics is urgently needed to address the ongoing antibiotic resistance crisis. Deep learning approaches have aided in exploring chemical spaces; these typically use black box models and do not provide chemical insights. Here we reasoned that the chemical substructures associated with antibiotic activity learned by neural network models can be identified and used to predict structural classes of antibiotics. We tested this hypothesis by developing an explainable, substructure-based approach for the efficient, deep learning-guided exploration of chemical spaces. We determined the antibiotic activities and human cell cytotoxicity profiles of 39,312 compounds and applied ensembles of graph neural networks to predict antibiotic activity and cytotoxicity for 12,076,365 compounds. Using explainable graph algorithms, we identified substructure-based rationales for compounds with high predicted antibiotic activity and low predicted cytotoxicity. We empirically tested 283 compounds and found that compounds exhibiting antibiotic activity against _Staphylococcus aureus_ were enriched in putative structural classes arising from rationales. Of these structural classes of compounds, one is selective against methicillin-resistant _S. aureus_ (MRSA) and vancomycin-resistant enterococci, evades substantial resistance, and reduces bacterial titres in mouse models of MRSA skin and systemic thigh infection. Our approach enables the deep learning-guided discovery of structural classes of antibiotics and demonstrates that machine learning models in drug discovery can be explainable, providing insights into the chemical substructures that underlie selective antibiotic activity.
Full text is behind a Springer paywall, because one couldn’t imaging allowing the _hoi polloi_ access to such knowledge. Source code for the “Chemprop” molecular property prediction deep learning software is, however, [available on GitHub](https://github.com/chemprop/chemprop) with [documentation here](https://chemprop.readthedocs.io/en/latest/). Chemprop code used in the paper is also [posted on GitHub](https://github.com/felixjwong/antibioticsai).
---
### Author:  [@johnwalker](https://scanalyst.fourmilab.ch/u/johnwalker)
#### Post date: [23 December 2023 15:31 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/653 "2023-12-23T15:31:44Z")
J.P. Morgan goes all “[Roaring Twenties](https://scanalyst.fourmilab.ch/t/rudy-rucker-and-john-walker-on-artificial-intelligence-in-the-roaring-twenties/4099)” on generative artificial intelligence.
[](https://twitter.com/JimPethokoukis/status/1738212545532445110)
---
### Author:  [@Roxie](https://scanalyst.fourmilab.ch/u/Roxie)
#### Post date: [23 December 2023 15:55 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/654 "2023-12-23T15:55:48Z")
Yikes!
And breathtaking.
---
### Author:  [@jabowery](https://scanalyst.fourmilab.ch/u/jabowery)
#### Post date: [23 December 2023 16:14 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/655 "2023-12-23T16:14:59Z")
> [@johnwalker](#):
>
> The discovery of novel structural classes of antibiotics is urgently needed to address the ongoing antibiotic resistance crisis.
You can tell when the rentier class is _serious_ about a tech by the degree to which it maintains their own evolution of virulence (take the money of one body politic and run to the next) by fighting the evolution of competing virulent agents (turn host body into copies of itself that shed to infect other bodies). If the “populists” get the idea that border control is a matter of immediate survival value to their children, _they_ might get serious about preventing horizontal transmission of the rentier class between bodies politic. Can’t have that! Anything but THAT!
---
### Author:  [@eggspurt](https://scanalyst.fourmilab.ch/u/eggspurt)
#### Post date: [27 December 2023 15:05 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/656 "2023-12-27T15:05:20Z")
The attention has moved on from GenAI to Geopolitical uncertainty:
---
### Author:  [@Roxie](https://scanalyst.fourmilab.ch/u/Roxie)
#### Post date: [27 December 2023 15:42 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/657 "2023-12-27T15:42:04Z")
Gosh, I wonder why…
---
### Author:  [@johnwalker](https://scanalyst.fourmilab.ch/u/johnwalker)
#### Post date: [29 December 2023 16:39 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/658 "2023-12-29T16:39:34Z")
> **[The New York Times sues OpenAI and Microsoft for using its stories to train...](https://apnews.com/article/nyt-new-york-times-openai-microsoft-6ea53a8ad3efa06ee4643b697df0ba57)**
>
> The Times said OpenAI and Microsoft are advancing their technology through the “unlawful use of The Times’s work to create artificial intelligence products that compete with it.”
The _New York Times_ has filed a lawsuit against the OpenAI cluster of companies: [The New York Times Company v. MicrosoftCorporation (“Microsoft”) and OpenAI, Inc., OpenAI LP, OpenAI GP LLC, OpenAI LLC, OpenAI OpCo LLC, OpenAI Global LLC, OAI Corporation, LLC, OpenAI Holdings, LLC](https://s3.documentcloud.org/documents/24238498/nyt_complaint_dec2023.pdf), The link is to the complaint, filed in the U.S. District Court for the southern district of New York, a 69 page PDF file. The complaint alleges that OpenAI’s “scraping” of the Web for training data for its large language models infringes the _Times_’ copyright on its original content, and presents numerous examples in which GPT-4, from a minimal prompt, reproduced lengthy content originally published in the _Times_ almost verbatim. Here is an example from p.30, with identical text shown in red.
These examples continue for 17 pages and constitute a substantial part of the complaint. The complaint requests a jury trial, and analysis have commented that such examples have been effective in previous infringement actions.
After reviewing the complaint, AI researcher and consultant Brian Roemmele conducted experiments with “The Pile” open source corpus for AI foundation model training. The results are reported on 𝕏 as a series of posts starting with the one below.
[](https://twitter.com/BrianRoemmele/status/1740386671148122476)
[](https://twitter.com/BrianRoemmele/status/1740393363034276053)
---
### Author:  [@CTLaw](https://scanalyst.fourmilab.ch/u/CTLaw)
#### Post date: [29 December 2023 17:36 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/659 "2023-12-29T17:36:00Z")
> [@johnwalker](#):
>
> The link is to the complaint, filed in the U.S. District Court for the southern district of New York, a 69 page PDF file.
With exhibits:
[https://www.courtlistener.com/docket/68117049/the-new-york-times-company-v-microsoft-corporation/](https://www.courtlistener.com/docket/68117049/the-new-york-times-company-v-microsoft-corporation/)
---
### Author:  [@owenwengerd](https://scanalyst.fourmilab.ch/u/owenwengerd)
#### Post date: [29 December 2023 22:21 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/660 "2023-12-29T22:21:28Z")
The copyright violations occur in the training phase (you can’t “read” a web site without copying some bits), not in the generated output. I think it is a red herring to focus on the generated output.
---
### Author:  [@johnwalker](https://scanalyst.fourmilab.ch/u/johnwalker)
#### Post date: [29 December 2023 23:22 UTC](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383/661 "2023-12-29T23:22:05Z")
> [@owenwengerd](#):
>
> The copyright violations occur in the training phase (you can’t “read” a web site without copying some bits), not in the generated output.
Since the weights in the OpenAI models are proprietary and undisclosed (locked away on a cloud server where third parties cannot examine them), my understanding of the argument in the complaint is that the fact that a simple prompt (which, I notice, they do not choose to disclose—I think this weakens their argument and is one thing the defense should demand in discovery—nor did they disclose how many times they tried the prompt in order to obtain the text quoted in the complaint) elicits largely identical text to the copyright article is used as evidence that the copyright text was used to train the model and can be recovered from the model without permission simply by asking for something similar. In the complaint, the _Times_ goes to pains to describe their investment of time and money in creating the purported original content which they claim was purloined in training the model and can be recovered without compensation in nearly-original form from it.
If Brian Roemmele’s reports that these models can create news stories comparably similar to those published in media for events which occurred after the cutoff date for their training corpus by “hallucinating” details anchored by a few facts supplied in the prompt are correct, that will be useful for the defense and we may see Roemmele as an expert witness if the matter goes to trial.
[Previous page](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383.md?page=32)
[Next page](https://scanalyst.fourmilab.ch/t/generative-artificial-intelligence-large-language-models-and-image-synthesis/2383.md?page=34)