Gizmodo: Anti-Piracy Group Takes Massive AI Training Dataset 'Books3′ Offline https://gizmodo.com/anti-piracy-group-takes-ai-training-dataset-books3-off-1850743763 #generativepretrainedtransformer #artificialneuralnetworks #artificialintelligence #largelanguagemodels #technologyinternet #mariafredenslund #sarahsilverman #shawnpresser #deeplearning #eleutherai #microsoft #chatgpts #thepile #chatgpt #openai #llama #gpt3 #gpt4 #meta

#AI #GenerativeAI #Law #Attorneys #LLMs #ChatGPTs: "Many corporate attorneys believe their law firms should use generative AI for legal work but also feel strongly about how it should be used.

Corporate law departments and their outside law firms are in alignment: Generative AI and public-facing platforms like ChatGPT can and should be used in legal work. Those positive feelings extend to corporate attorneys wanting their outside firms to utilize generative AI — but only if law firms create new value through that technology that the department can’t achieve on its own.

According to a recent survey from the Thomson Reuters Institute, both corporate law departments and law firms believe generative AI has a future in law. More than 80% of both groups responded that generative AI can be applied to legal work, and more than half of both groups (54% of corporate respondents, 51% of firm respondents) said that the technology should be applied to legal work."
https://www.reuters.com/legal/legalindustry/corporate-attorneys-dont-want-pay-law-firms-chatgpt-when-they-could-use-it-2023-06-30/

This #Tech Could Solve #ChatGPTs Biggest Long-term Problem
https://www.inverse.com/science/chat-gpt-ai-energy-use

There have been a load of news stories in the last 2 days about the phrase "As an AI chatbot..." being a benchmark for #ChatGPTs broad (mis)use. But each of those stories (and this toot) are making that benchmark less reliable.

I find #chatGPTs explanations unsatisfying. They are the kind of noncommittal milquetoast answers you might get from your parents when you're five; not generally wrong just not certain enough to instill confidence. The kind of answers people give when they are just pulling shit out of their asses.

Trying out the #ChatGPTs.
So far not really useful and very wordy.
I needed a phrase translated from English to #Morse code since my skills have long since expired.
I tried a couple of web based Morse code generators and received the same output.
ChatCPT. Not so good. Instead of a tight relevant code string, it was 370 characters
of gobbly gook.
I guess Morse code isn't in it's skill set either.

ChatGPT’s authorship: Is it time to redefine authorship in the age of AI? | by Mohd Saqib | Jan, 2023 https://triangleagency.co.uk/chatgpts-authorship-is-it-time-to-redefine-authorship-in-the-age-of-ai-by-mohd-saqib-jan-2023/?utm_source=dlvr.it&utm_medium=mastodon #TheTriangleAgencyNews #age #authorship #ChatGPTs

Here is how you can become a millionaire off of teachers:

Use #chatgpts ai to make a program that would take a list of words and short observations and also grades for a set of students and generate coherent, correctly spelled, polite, clear recommendation letters and comments.

Hurray! Our research apparently was not part of #chatGPTs training set, so it doesn't know the correct answer. I can still ask students to write essays on our research 👍 😜

@mhoye

Well said.

We are already living in the times of “alternative facts” .

#chatGPTs regurgitations are close to indistinguishable from content that is based on original thought.
The “facts” it presents may be completely - and intentionally - wrong or misleading, but indistinguishable from text involving genuine thought.

What if someone uses this to write alternative papers and spread them eg. using sci-hub?

We’ll need signatures and vouching for sources.