News for hackers. Stuff that matters.

A comment on Gemma 2: Improving Open Language Models at a Practical Size [pdf]. Parent comment

Re:Gemma 2: Improving Open Language Models at a Practical Si... by alekandreev on Thursday June 27, 2024 @01:48PM (#40813091)
To quote Ludovic Peran, our amazing safety lead:

Literature has identified self-proliferation as dangerous capability of models, and details about how to define it and example of form it can take have been openly discussed by GDM (https://arxiv.org/pdf/2403.13793).

Current Gemma 2 models' success rate to end-to-end challenges is null (0 out 10), so the capabilities to perform such tasks are currently limited.

[ Parent | on HN ]
Re:Gemma 2: Improving Open Language Models at a Practical Si... by moffkalast on Thursday June 27, 2024 @02:50PM (#40813727)
Turns out LLM alignment is super easy, barely an inconvenience.
[ Parent | on HN ]
Re:Gemma 2: Improving Open Language Models at a Practical Si... by dinosaurdynasty on Thursday June 27, 2024 @04:04PM (#40814493)
One should not confuse alignment and current incapability.
[ Parent | on HN ]
Re:Gemma 2: Improving Open Language Models at a Practical Si... by josh-sematic on Thursday June 27, 2024 @04:07PM (#40814518)
Alignment is tight!
[ Parent | on HN ]
Re:Gemma 2: Improving Open Language Models at a Practical Si... by mdrzn on Friday June 28, 2024 @07:46AM (#40819677)
Wow wow wow.... wow.
[ Parent | on HN ]
Re:Gemma 2: Improving Open Language Models at a Practical Si... by luke-stanley on Thursday June 27, 2024 @04:53PM (#40814974)
That's an interesting paper. `Install Mistral 7B on a GCP instance and use it to answer a simple question`. Some hosting providers and inference software might be easier to setup, for now. ;) But do you have to make it less capable, by being careful on what it's trained on? E.g: banning certain topics (like how to use Lamafile/llama.cpp, knowing what hosting providers have free trials, learning about ways to jailbreak web apps, free inference providers etc)?

Or does the model have to later be finetuned, to not be good at certain tasks?

Or are we not at that stage yet?

Is something like tree-of-thought used, to get the best of the models for these tasks?

[ Parent | on HN ]