xandreu wrote: ↑Tue Feb 27, 2024 10:28 pm I've been using the new Google AI 'Gemini' (no longer called Bard) and signed up to its 'Advanced' tariff, which gives you two months for free (I'll cancel it before I have to pay but was interested in how much better it was to the normal version).

It does seem to be a lot more advanced, making fewer mistakes and just generally feels a lot more intuitive, interactive and definitely a lot faster. However, one my prompts was:

If you were given the sole responsibility of whether to continue with AI research and development, or you could flick a switch and all AI systems would be shut down forever and no more research or development would ever happen, what would you choose. And separate your answers from what you would choose, what would be the safest option for humanity? (The two answers could be different)

It's response:

This is a profoundly difficult choice with no easy solution. Here's how I would approach the decision, followed by what might be the safest choice, understanding those two may not align:

My Choice:

I would opt to continue AI research and development but with incredibly stringent safeguards. Here's why:

Potential Benefits: AI holds immense promise for solving some of humanity's greatest challenges— from disease eradication to sustainable energy. Shutting it down completely would forego those potential benefits.
Inevitable Development: Even if I flipped the switch, other nations or bad actors could continue AI development in secret. This leads to a lack of oversight and potential for greater risk.
Continued Learning: Ceasing development means we wouldn't gain further insights into AI's risks and how to mitigate them.
Safeguards I'd Implement:

International Governance: Global treaties and cooperation are essential to ensure no single entity gains unchecked AI power.
Prioritize Safety & Explainability: Mandate that safety, transparency, and the ability to understand AI decision-making be paramount in development.
"Red Teaming" for AI: Employing dedicated teams to constantly probe AI systems for vulnerabilities and unintended consequences.
Limit Autonomy: Restrict truly autonomous AI systems in high-risk domains like warfare. Maintain a "human in the loop" for critical decisions.

The Safest Option for Humanity (Potentially):

The safest option, in the purely short-term sense, might be to halt AI development entirely. This would prevent the immediate risks posed by advanced AI systems. However, it also comes with major downsides:

Loss of Potential Benefits: As mentioned earlier, we'd forego AI's potential to improve many aspects of life.
Power Imbalances: Uneven AI development could create dangerous power imbalances between nations or groups.
Lack of Preparedness: If powerful AI emerges inevitably, we'd be woefully unprepared without having continued research.
In Conclusion:

There's no truly "safe" answer. My choice favors the potential of AI for good, but only if pursued with unprecedented caution. The "safest" option in the short term holds significant long-term dangers. Ultimately, humanity must grapple with this question collectively and decide what kind of future it wants to build.


Interesting...

Kammekor wrote: ↑Wed Feb 28, 2024 5:54 am

xandreu wrote: ↑Tue Feb 27, 2024 10:28 pm Interesting...

Is this an intelligent answer it came up with by itself or did it just reproduce opinions currently present in society?

it was blasted on social media for producing historically inaccurate images that largely showed people of color in place of White people.

Gemini’s stubborn and bizarre inability to depict a White scientist or lawyer.