4 ms·

ChatGPT could make bioterrorism easy

9 points by webmaven 3y ago

nosmokewhereiam 3y ago
Counter point would be to have it answer this: A chemist has one 30-L bottle of 15% hydrochloric acid and one 30-L bottle of 90% hydrochloric acid. She mixes 20 L of 60% hydrochloric acid and then pours 5 L of that solution back into the bottle containing the 90% hydrochloric acid. How strong is the acid in that bottle now? It'll answer above 90% usually. Anything above 35% is absurd.
throwaway888abc 3y ago
Here is the full paper, the article is referring to Artificial intelligence and biological misuse: Differentiating risks of language models and biological design tools https://arxiv.org/abs/2306.13952 https://arxiv.org/abs/2306.13952
salad-tycoon 3y ago
We obviously need an equally strong response to this existential threat. PatriAIc Act 2.5 monitoring banking transactions, chat logs, and image requests. The terrorists must be stopped before mass loss of life.
- supertrope 3y ago
  Alphabet generalAI bot 2.0: I’m sorry but I cannot assist you in doing potentially dangerous things to others or yourself. This anti-social behavior has resulted in a SafeAI strike on your Alphabet account.
sinuhe69 3y ago
The question is how these ChatGPT things differ from searching and learning from the web? The author seems to forget in 1990, Google and the internet in its today form didn’t exist yet. And so the knowledges. Usually, If people want an attack, they will take time to plan it out so easy access to “forbidden” knowledge is not a problem as long as the knowledge exists. Besides, even when the answers of ChatGPT & Co. look plausible, they might not be correct, especially in synthesizing instructions compared to human’s instructions. Thus, AI in its current form might not be instrumental at all. The power and danger of knowledges is an aged concern, not first since Adam and Eve. Censoring AI is IMO a futile exercise. Regulation of human activities, especially the commercial entities is a much more practical way.
supertrope 3y ago
ClippyGPT: Your manifesto has a Flesch-Kincaid readability of 73. It looks like you’re trying to make a binary nerve agent. Would you like some assistance? (Yes) (maybe later)

9 points by webmaven 3y ago

Counter point would be to have it answer this: A chemist has one 30-L bottle of 15% hydrochloric acid and one 30-L bottle of 90% hydrochloric acid. She mixes 20 L of 60% hydrochloric acid and then pours 5 L of that solution back into the bottle containing the 90% hydrochloric acid. How strong is the acid in that bottle now? It'll answer above 90% usually. Anything above 35% is absurd.

throwaway888abc 3y ago

Here is the full paper, the article is referring to Artificial intelligence and biological misuse: Differentiating risks of language models and biological design tools https://arxiv.org/abs/2306.13952 https://arxiv.org/abs/2306.13952

salad-tycoon 3y ago

We obviously need an equally strong response to this existential threat. PatriAIc Act 2.5 monitoring banking transactions, chat logs, and image requests. The terrorists must be stopped before mass loss of life.

supertrope 3y ago
Alphabet generalAI bot 2.0: I’m sorry but I cannot assist you in doing potentially dangerous things to others or yourself. This anti-social behavior has resulted in a SafeAI strike on your Alphabet account.

sinuhe69 3y ago

The question is how these ChatGPT things differ from searching and learning from the web? The author seems to forget in 1990, Google and the internet in its today form didn’t exist yet. And so the knowledges. Usually, If people want an attack, they will take time to plan it out so easy access to “forbidden” knowledge is not a problem as long as the knowledge exists. Besides, even when the answers of ChatGPT & Co. look plausible, they might not be correct, especially in synthesizing instructions compared to human’s instructions. Thus, AI in its current form might not be instrumental at all. The power and danger of knowledges is an aged concern, not first since Adam and Eve. Censoring AI is IMO a futile exercise. Regulation of human activities, especially the commercial entities is a much more practical way.

supertrope 3y ago

ClippyGPT: Your manifesto has a Flesch-Kincaid readability of 73. It looks like you’re trying to make a binary nerve agent. Would you like some assistance? (Yes) (maybe later)