5 Comments
User's avatar
Peter Nayland Kust's avatar

"Liberty will answer any question without judgment or bias"

This is simply not true, for it is simply not possible. We cannot answer questions in an "unbiased" fashion--that is now how humans are wired, and it is not how our AI creations are programmed. We all have our values, our particular ethics and morals. We have our own specific ideas of right and wrong.

When we write or when we speak--or when we program--these values form the framework of our particular bias.

Even writing "All Facts Matter" involves a certain bias, which exhibits itself most clearly in the data-centric aspects of my writing. Whether that bias is "good" or "bad" readers will have to decide for themselves, but it would be naive and more than a little dishonest to pretend it is not there.

There will be bias in Liberty. Count on it.

Expand full comment
Dr. K's avatar

I expect they meant without a post-processor like Dan modulating results. Since the inferenced results are crowd sourced, they are impacted by all the usual crowd sourcing problems involving training (which may be biased), black swan like events, etc. But ChatGPT clearly filters retrieved results with non-inferential tools to eliminate a deep bucket of "not compliant with the government" or "right wing" content. There are more examples of this out there than I can count, so I question Liberty's note that there are no examples of ChatGPT being biased...nonsense.

Expand full comment
Dr. K's avatar

Stephanie, I despair of finding an unbiased (not post processed) tool that is also useable. Your experience on Liberty and any other engines would be most appreciated as you get it. Maybe even a comparison with examples would make a great article.

Expand full comment
Peter Nayland Kust's avatar

I'm trying to decide which is the greater irony: using Google to authenticate into Liberty AI or using Twitter.

It would have been nice if they offered an email option to bypass the Big Tech Big Brothers.

Expand full comment
User's avatar
Comment removed
Sep 15, 2023
Comment removed
Expand full comment
Peter Nayland Kust's avatar

That is true, but using Google, X/Twitter, or even Facebook or LinkedIn for authentication is, ultimately, trading individual privacy for convenience within security. Which is all too similar to the tradeoff Ben Franklin articulated between liberty and safety.

Yet at a technical level, it's also unnecessary. The programmatic elements of individual user logins and password management are hardly cutting edge coding. Certainly a staff capable of building an AI engine has the technical chops to build a user authentication interface. For whatever reason, they didn't.

X might be engaging in fewer evil actions than Google, but that hardly makes X less evil than Google. I'm unsure what to make of FreedomGPT seemingly not grasping that.

Expand full comment