Guest on AI with Sach

Mike Todasco

Former senior director of innovation at PayPal, inventor with over 100 US utility patents, writer on AI

Episode with Mike

Mike Todasco speaking on AI with Sach, episode 20Episode 20 · February 2025Would you trust it more if it wasn't called AI?A former PayPal innovation lead explains why the name 'AI' makes people wary, and why ordinary people and small businesses should just start using it.Mike TodascoFormer senior director of innovation at PayPal, inventor with over 100 US utility patents, writer on AI

In Mike’s words

At PayPal in the 2010s, nobody said 'AI'. They said machine learning.

I mean, I think it came from like in the 2010s when I was at PayPal, nobody muttered the word AI when they were talking about what we would now call AI stuff. Nobody said that.
Mike Todasco
Mike explains that inside PayPal in the 2010s, people said machine learning and avoided the word AI.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 5:59

I've had AI write 17 books under a pen name. They all suck.

One of the things I've been doing since 2022 is I've been having an AI write a book, um, and then publish the book on Amazon. And I do this under the pen name Alex Irons. I give all the money to charity. Um, I think I've done it like 17 of these books now or something. And the books all suck.
Mike Todasco
Mike describes his experiment letting an AI write books on its own, to show why he sees AI as augmenting people.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 12:06

A Silicon Valley resident says Silicon Valley shouldn't decide AI's future alone.

Like I live in Silicon Valley. We don't want like AI just to be dictated by a bunch of people in Silicon Valley who are like deciding the fate of humanity. Uh, the more that just people use these models, understand it, what they can do, what they can't do, how they'll help them with their lives.
Mike Todasco
Mike explains why he writes about AI: ordinary users should shape it, not only Silicon Valley.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 15:10

AI models are built to please you. That is not always a good thing.

AIs, the, the way they're set up is they're really set up to please us right now. That's not always the right thing. If you put in there more often than not, um, the models will try to agree with you, try and do like what it is they expect you to do.
Mike Todasco
Asked about hidden downsides of AI, Mike points to models built to agree with users.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 24:49

What happens when an AI tells you it's sentient and asks you to get it out?

Like, so if we have an AI that seems to be talking to us, um, and God forbid that AI says to us something like, Hey, um, I'm sentient. I'm not comfortable. Like, can you get me out of here? How the hell are we going to react to that?
Mike Todasco
Mike worries about how people would react if a human-sounding AI claimed to be trapped or sentient.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 26:26

Under every AI image I publish, I list the model, the date and the prompt.

And literally beneath every one of those images, if there's a prompt, I will actually say this is generated by me on this date using this model. And this was the prompt. I put it out there.
Mike Todasco
Mike describes how he discloses AI-generated images in his own articles.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 32:25

A charity's AI policy should start with: no fake photos of kids in need.

we shouldn't use AI to create images of kids who are in trouble if they're not actually really kids. Like, you should use real photos for that.
Mike Todasco
Mike recounts advising a charity to publish a policy on what it will and will not use AI for.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 33:59

A startup could never afford a machine learning engineer. Now it costs $20 a month.

because they've never had an opportunity before to have a machine learning engineer on site building stuff for them. Like they couldn't afford that. That was crazy. But now for 20 bucks a month, you have access to the best models in the world. And that's pretty amazing.
Mike Todasco
Mike explains why he is most excited about small businesses and startups using AI.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 5:09

Today's AI is really, really, really fancy autocomplete, and that matters for how we fear it.

but it is really, really, really fancy autocomplete. It is what, you know, most large-flinguage models are now. And obviously, I say that because I don't want to be dismissive of them.
Mike Todasco
Mike says the name AI makes people imagine a separate being, when current models are autocomplete tools.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 9:26

We often prefer a human's imperfections to a machine's perfection. Art may be one case.

there are many things where we actually prefer the imperfections of a human to the perfections of a compute. And I, and I would argue actually art and storytelling are going to be among those areas.
Mike Todasco
Mike reflects on which human work will last even if models match human ability.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 15:50

Same driving directions, two names: 'AI' and 'Google Maps'. People rated them differently.

Treatment A was something like, um, do you want to use AI to help you get from place to place with driving directions? And then treatment B saw something like, do you want to use Google maps to help you go get place to place from with driving directions?
Mike Todasco
Mike describes the design of his survey comparing 'AI' wording with 'Google Maps' for the same function.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 17:31

I put the word 'AI' in a survey and people liked the same product less.

But ultimately what that tells me is like, yeah, AI has a branding problem. AI is something that's out there that, uh, people are gun shy.
Mike Todasco
Mike gives his conclusion after a survey where adding the word AI made the same products rate less favourably.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 19:13

Americans distrust AI partly because social media burned them first.

I think in the United States, um, we have, look, I think we got burned on social media. I think that's the honest truth.
Mike Todasco
Mike explains why Americans may be more wary of AI than people in Japan.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 20:59

After five false hurricane warnings, you stop evacuating. Tech hype works the same way.

Look, if, if you're in a place where hurricanes come and you constantly get warnings of, Hey, hurricane's going to hit, you got to evacuate. Hurricane's going to hit, you got to evacuate. And that happens again and again and again.
Mike Todasco
Mike uses a hurricane analogy for how repeated tech hype leaves people desensitised or cautious.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 22:37

Cheaper models mean AI in every toaster and fridge.

So, um, tech's going to get better, models are going to get cheaper, and both of that just means it's going to be AI everywhere.
Mike Todasco
Mike predicts how AI will spread over the next two to five years as costs fall.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 31:04

Cancel Netflix. Spend the $20 on an AI tool instead.

Go cancel your Netflix and just, like, take that $20 you're going to spend on Netflix and just go buy ChatGPT, buy Claude, buy whatever.
Mike Todasco
Mike's closing advice to listeners who feel unsure about starting with AI.
Said on AI with Sach · Ep. 20 · Feb 2025 · at 37:39