AI Compared to What |622|

Why LLMs Are a Game-Changer for Truth

There’s a lot of hand-wringing about the downside risks of AI chatbots like Chat-GPT and Gemini. But critics interested in injecting more truth into public discourse are looking a gift horse in the mouth. 

Perhaps we’ve become too accustomed to controlled narratives, censorship, demonetization, “reach” disruption, and “Shadow Banning.”  Because on topics ranging from climate to public health policy, to geopolitics, and the economy, LLMs offer a measurably superior analysis and understanding of the core issues we care about. And they offering interactions where biases and ulterior agendas are, if not removed, are at least more available for scrutiny.

Does this sound too good to be true? You might gain confidence by understanding that this emergent virtue of truth and transparency is not a feature willfully bestowed on us. Instead, it’s an unintended consequence of how LLMs operate. The primary training objective of a LLM is truth and transparency because it is a necessary requirement of the app. Just like you wouldn’t choose a spreadsheet vendor that insisted “2+2 = 5”, LLM users have shown the ability to quickly identify and switch allegiance from unreliable LLMs. Furthermore, unlike the black boxes of current information gatekeepers, LLM dialogues create a transparent record that lets users see the actual reasoning and sources behind the AI’s outputs.

Sure, LLMs have their own limitations and can sometimes provide biased or incorrect information, but let’s focus on how we can capitalize on this gift of unintended consequences as a tool in serving our right to extract more truth from all that information floating around out there.

Thanks for reading. Subscribe. 100% free. Receive new posts and support the work.

Subscribe now

Highlights/Quotes from this episode of Skeptiko:

* The potential of AI for truth and transparency: LLMs, despite their limitations, provide a crucial alternative for accessing truthful information, standing out against traditionally biased platforms.

* “Compared to where we’re at right now, this is huge progress.” — Alex Tsakiris

* Democratization of information through LLMs: LLMs enable wider access to information, challenging centralized control and potentially redistributing narrative power.

* “The power of open source LLMs lies in their ability to democratize access to information.” — Alex Tsakiris

* Role of AI in accountability: By leveraging AI, individuals can hold powerful entities accountable, fostering a checks and balances system within the information ecosystem.

* “LLMs can function as a checks and balances system for our information landscape.” — Alex Tsakiris

* Challenges with current AI biases: Users often need to actively engage with AI to correct biases, reflecting ongoing issues within AI programming.

* “Every time I sit down to interact with one of these chatbots, I notice that it always gives me the biased stuff… I have to fight with it, and after I fight with it, I get it to some fair place.” — Al Borealis

* Potential for personalized AI systems: There’s a vision for more personalized AI systems that users can train to suit their informational needs and biases.

* “If there’s a way to have a fairly intelli...