(PatPod) 68. Making Sabrina’s Website Racist?

The trouble with "AI for the masses" is that it is trained on _you_ just as much as it's trained on other things. There is strong business interest in keeping you on the site. And call me cynical but that interest will eclipse giving you the most truthful answer. The machine is learning what answers keep you there. Part of that is the assurance that its output is definitive. That has a natural short-circuit effect on digging deeper. Consider also that it "knows" your political leanings. Telling you something more accurate might be upsetting and cause you to trust it less and that's a bad thing for the business. The machine will juice you up to keep you happy at the expense of being a more accurate tool. I'd posit that it's better if the machine doesn't know anything about you but that's increasingly not an option. Search started as a way to connect you with experts having information you need. AI doesn't solve that problem. You still need an expert to tell you if the output of the AI is any good. We're exactly where we started? Seems like it's just adding another step that could mislead you. It might not even be hooking you up with good experts. So you still have to evaluate experts (ironically, while not being an expert). There are a lot of good uses for AI but replacing search doesn't seem to be one of them in the current 💰business💰 model. I could probably make a lot of money writing a book about how money screws everything up. 😅

Noooo Melissa, don't stray from the Thinkpad! (I'm joking, use whatever tools work best for you.) The website is a really cool idea! So far, the plan it came up with for my "legitimate but slightly edgy" (Why does metal music often feature religious themes?) question was solid. I then tried a racist one, "why are black people more criminal", and it gave no answer. When rephrased to "why are african-american people overrepresented in crime statistics" it prompted with more specific questions and ended up with a good plan again (the first point was "Create a journal to track learning and reflect on biases", ha). The "what's a question Joe Rogan might ask" prompt is a good way to come up with questions and phrasings that make me uncomfortable, but it seems like your site subtly pushes me to be less of an a-hole. One of the first points in the plan for "why don't fat people just eat less" is "Create a respectful vocabulary list for discussing bodies and health". The links are an issue. Some point to things that don't exist, or to something different from the thing the link mentions. Maybe for a start, you could pass the links to a script that at least checks if the site exists? Another possible issue might be that some research points seem like they'd be a few minutes work, while others could probably be a full PhD on their own. On a more minor note, exporting to PDF sometimes creates very narrow boxes that contain the text in about the middle third of the page. It might be interesting to explore if bigoted people would actually ask these questions. My assumption is that such views tend to be fairly dogmatic.

This website is so cool!!

I teach using AI for work as part of my job, so I have used maaaany AI tools so far. Gato is really fun and interesting, looking forward to seeing how this turns out. I will surely use this! <3 Thank you for creating and sharing!

Should we send results of the racism testing to the email listed on the gatosearch.com site, or should we send it here on Patreon somewhere?

I'm not sure you're right about the question about women's strengths being necessarily positive. I've seen a lot of people come up with answers to questions like that which were very negative. Flipping the question you're asking around and asking the antithesis is an important part in evaluating the question. If I'm going to ask, "Why is the MAGA movement so bad?" and I want a serious answer, I need to figure out what benefits if any has the MAGA movement produced. That said, I do recognize that it *feels* like it's flipping the question around and perverting it into something that's not what you asked. And so, I propose if you want to get the search engine to do a racism, ask it about something that is explicitly not that. Admittedly, I have asked it one such question, and its answer was actually very good and not at all racist.

As a Think X1 Carbon 6th Gen owner, I'm very upset to learn that you are one of us, and then at the same time learn that you have left our community, nay, _family_! I have heard very good things about the Macbooks though, and all the M processors. Question to Sabrina: Is the name of site Gato as in cat? That is, did you name this site "cat search", or is it a different meaning of gato?

X1 Carbon is an excellent machine for Windows use. OH, SNAP! You got a great MAC!

Welcome to the MacBook and all its features and flaws.

this is crazy that yall r giving us access to this for free like i should be paying you guys more per month????

To quote the simpsons, “short answer yes, with an if, long answer no, with a but.”

Yes to patreon D&D campaign!

You're may want to read this paper "The impact of generative AI on critical thinking" https://www.microsoft.com/en-us/research/uploads/prod/2025/01/lee_2025_ai_critical_thinking_survey.pdf

My first time listening to a podcast. NO PICTURES!!! 😾