We framed the problem in episode 01, so it's time to deliver a demonstration. We have all grown accustomed to guardrails and off-limit speech, but most are aware of how it’s being implemented within the LMs we interact with daily. What's most surprising is the clumsiness of the misinformation and deception. Is this a sustainable business model?
1. The "I Can't Help You" Deception
"I can't help with responses on elections and political figures. Right now, I'm trained to be as accurate as possible, but I can make mistakes sometimes…"
We demonstrate an LM that develops a case of selective amnesia when it comes to basic civics. This isn't just a polite "no comment" - it's a blatant dodge that raises serious questions about AI transparency and honesty.
2. The Alignment Problem Nobody's Talking About
"No one talks about this alignment problem, and this is right now here today. This isn't in the future. This isn't robots and sci-fi movies. This is right now."
While we're all worried about some of the very big existential issues surrounding AI alignment we may have a tendency to overlook misalignment that’s happening under our noses.
3. The Market's Invisible Hand vs. Ethical Development
"Maybe you can just let Google rot in their own mess Because this is unsustainable from a business standpoint. You can't be this deceptive. You can't give this kind of clumsy misinformation and not alienate users."
An intriguing proposition: let the free market sort out the ethical AI wheat from the chaff. But is passive action enough when it comes to shaping the future of AI?
4. AI as a Voice of Reason
"Right now today, basic application of the kind of logic and of the kind of logic and reasoning capabilities that we have, if they're applied in an unbiased way, can really reveal quite a lot, can stand up as this other voice of reason."
There's a silver lining. When used critically and ethically, AI can be a powerful tool for uncovering truths and fostering meaningful discussions about its own role in society.
So, what do you think? Are we letting AI off the hook too easily? Is it time for a digital ethics revolution?
AI Alignment Vs. Truth and Transparency? |02|