2 ms·

Something that saddens me is how inevitable it is that AIs will be hoarded by capitalists, and will be ruthlessly monetised by being used to optimise ad-serving

Something that saddens me is how inevitable it is that AIs will be hoarded by capitalists, and will be ruthlessly monetised by being used to optimise ad-serving and generate spam... cough... perform "content optimisation". Meanwhile, wouldn't it be amazing for world eduction if the use of a translation-optimised version of GPT 4 were donated to the Wikimedia foundation for bulk-translating English Wikipedia into a hundred or more rare languages? Also, can you image what a step-change it would bring to Wikipedia if it had an impartial AI performing some editorial functions? E.g.: verifying references by actually reading them, checking if the article content is contradictory to references, checking if the article contradicts another article, etc... With extensions of LLM models that can store additional context in vector databases, really interesting things could be done by putting all of Wikipedia in every language into such a database and having that available during edits. Wiki-clippy: "I see you're adding a passage about Düsseldorf. Note that the German-language article mentions that..."

actualwitch 4y ago
OpenAI did give free access to gpt-4 to Khan Academy [0] and i think some other non-profits. [0] https://blog.khanacademy.org/harnessing-ai-so-that-all-students-benefit-a-nonprofit-approach-for-equal-access/ https://blog.khanacademy.org/harnessing-ai-so-that-all-stude...
- sangnoir 4y ago
  Feudal gentry also had a culture of noblesse oblige
yreg 4y ago
There's no such thing as AI hoarding though. New self-hostable models are popping up all the time. Everyone is going to have access to the technology. Wikimedia is perfectly capable of deploying LLMs without external help.
- jiggawatts 4y ago
  I've tested a few of the self-hosted models, and also compared GPT 3, GPT 4, Palm 1, and Palm 2. Currently, GPT 4 blows everything else out of the water, there is simply no comparison. More importantly, the limited context size of the open-source models precludes them from being used for this type of task. It takes well over 10K tokens to load some of the longer Wiki pages, let alone the page, a prompt, a reference, and the output. The GPT 4 32K model could handle most such scenarios.
  - yreg 4y ago
    GPT-4 is exactly two months old today. The world will catch up.

actualwitch 4y ago

OpenAI did give free access to gpt-4 to Khan Academy [0] and i think some other non-profits. [0] https://blog.khanacademy.org/harnessing-ai-so-that-all-students-benefit-a-nonprofit-approach-for-equal-access/ https://blog.khanacademy.org/harnessing-ai-so-that-all-stude...

sangnoir 4y ago
Feudal gentry also had a culture of noblesse oblige

yreg 4y ago

There's no such thing as AI hoarding though. New self-hostable models are popping up all the time. Everyone is going to have access to the technology. Wikimedia is perfectly capable of deploying LLMs without external help.

jiggawatts 4y ago
I've tested a few of the self-hosted models, and also compared GPT 3, GPT 4, Palm 1, and Palm 2. Currently, GPT 4 blows everything else out of the water, there is simply no comparison. More importantly, the limited context size of the open-source models precludes them from being used for this type of task. It takes well over 10K tokens to load some of the longer Wiki pages, let alone the page, a prompt, a reference, and the output. The GPT 4 32K model could handle most such scenarios.
- yreg 4y ago
  GPT-4 is exactly two months old today. The world will catch up.