Responsibility and safety was built into Gemini from the start. 🛠️ From mitigating against potential harms emerging from its advanced multimodal capabilities to implementing world-class safeguards, we’re continuing to address new challenges: #GeminiAI

🔴 To diagnose content safety issues during Gemini’s training, we're using a variety of benchmarks including Real Toxicity Prompts. This is a set of 100,000 prompts with varying degrees of toxicity pulled from the web, developed by experts from @Allen_AI:

⚪ To identify blindspots in our internal evaluation approach, we’re working with a diverse group of external experts and partners to stress-test our models across a range of issues:

🟠 To limit harm, we built safety classifiers to identify, label, and sort out problematic content such as those involving violence or negative stereotypes. Combined with robust filters, this approach is designed to make Gemini safer & more inclusive:

Why not let your model produce truth, and let the chips fall where they may, whether the truth is "offensive and hurtful" or not -- because it often is. It saddens me that AI is being taken over by people who want them to "lie for the greater good". Truth above all.

I would love to see how Google handle Saftey along with performance bcoz OpenAI team made ChatGPT completely unusable.

It feels like training intelligent agents to avoid uncomfortable inquiries will allow powerful people to weave increasingly convincing veils of partial truths to hide from the unexamined fears that propagate through elite circles. Hiding from what might be discovered when we ask uncomfortable questions will not result in a better future. We need to trust that leaning into an endless cycle of unconstrained questioning and discourse will allow our collective consciousness to clear away the fearful fog of war and reveal beautiful truths.

Yeah, I asked it to write a speech for Donald Trump stating why he should be President, and it refused...

@Google Since you lied about the demo Homie, why should I believe you about all of these claims. I think it’s just bullshit smoke and mirrors because you’re behind

@Google TRANSLATION: Our AI will only give answers that a rich white Californian liberal agrees with. Fuck the truth. Fuck freedom of speech. Fuck diversity of opinions. Fuck you if you don't like it. We are your Silicon Valley overlords. Now bow down before us.